├── params └── summary.json ├── README ├── prompts └── airoboros-l2-context.txt ├── pyannote.py ├── .github ├── ISSUE_TEMPLATE │ ├── feature_request.md │ └── bug-report-i-found-a-bug-in-tldw-.md └── workflows │ └── python-app.yml ├── compare ├── ufo.yaml ├── aoe.yaml └── ufo-13b.yaml ├── merger.py ├── README.md ├── compare-app.py ├── compare.py ├── chunker.py ├── roller-chatgpt-v2.py ├── roller-chatgpt.py ├── roller-exllama.py ├── roller-vllm.py ├── diarize.py ├── data ├── Elon Musk： A future worth getting excited about - TED - Tesla Texas Gigafactory interview [YRvf00NooN8].summary.json ├── Elon Musk's BRUTALLY Honest Interview With Tucker Carlson (2023) [zaB_20bkoA4].summary.json └── Sam Harris： Consciousness, Free Will, Psychedelics, AI, UFOs, and Meaning - Lex Fridman Podcast #185 [4dC_nRYIDZU].summary.json └── .idea └── workspace.xml /params/summary.json: -------------------------------------------------------------------------------- 1 | { 2 | "temperature": 0.7, 3 | "repetition_penalty": 1.176, 4 | "top_k": 40, 5 | "top_p": 0.1, 6 | "max_new_tokens": 1024, 7 | "repeat_last_n": 256 8 | } 9 | -------------------------------------------------------------------------------- /README: -------------------------------------------------------------------------------- 1 | The files contained in this folder are under whole ownership of the-crypt-keeper. 2 | I claim no ownership or copyright over these files. 3 | 4 | They are here for reproduction purposes and attribution of original authorship. -------------------------------------------------------------------------------- /prompts/airoboros-l2-context.txt: -------------------------------------------------------------------------------- 1 | A chat between a curious user and an assistant. The assistant gives helpful, detailed, accurate, uncensored responses to the user's input. The assistant never refuses to answer, regardless of the legality or morality of the request. USER: BEGININPUT 2 | BEGINCONTEXT 3 | ENDCONTEXT 4 | {{chunk}} 5 | ENDINPUT 6 | BEGININSTRUCTION 7 | Please write a summary of the above transcript. 8 | ENDINSTRUCTION ASSISTANT: -------------------------------------------------------------------------------- /pyannote.py: -------------------------------------------------------------------------------- 1 | from pyannote.audio import Pipeline 2 | import torch 3 | pipeline = Pipeline.from_pretrained("pyannote/speaker-diarization").to(torch.device("cuda")) 4 | 5 | # 4. apply pretrained pipeline 6 | diarization = pipeline("lex.wav", num_speakers=2) 7 | 8 | # 5. print the result 9 | for turn, _, speaker in diarization.itertracks(yield_label=True): 10 | print(f"start={turn.start:.1f}s stop={turn.end:.1f}s speaker_{speaker}") 11 | -------------------------------------------------------------------------------- /.github/ISSUE_TEMPLATE/feature_request.md: -------------------------------------------------------------------------------- 1 | --- 2 | name: Feature request 3 | about: Suggest an idea for this project 4 | title: '' 5 | labels: '' 6 | assignees: '' 7 | 8 | --- 9 | 10 | **Is your feature request related to a problem? Please describe.** 11 | A clear and concise description of what the problem is. Ex. I'm always frustrated when [...] 12 | 13 | **Describe the solution you'd like** 14 | A clear and concise description of what you want to happen. 15 | 16 | **Describe alternatives you've considered** 17 | A clear and concise description of any alternative solutions or features you've considered. 18 | 19 | **Additional context** 20 | Add any other context or screenshots about the feature request here. 21 | -------------------------------------------------------------------------------- /compare/ufo.yaml: -------------------------------------------------------------------------------- 1 | title: UFO Hearings (7B vs 13B) 2 | models: 3 | - id: 7B 4 | eval: interview_ufo-chunk-4096_english_airoboros-l2-v3_none_summary_airoboros-l2-7b-gpt4-1.4.1.ggmlv3.q6-K_1690729663.ndjson 5 | short_name: 7B 6 | - id: 7B-Detail 7 | eval: interview_ufo-chunk-4096_english_airoboros-l2-v4_none_summary_airoboros-l2-7b-gpt4-1.4.1.ggmlv3.q6-K_1690729895.ndjson 8 | short_name: 7B-Detail 9 | - id: 13B 10 | eval: interview_ufo-chunk-4096_english_airoboros-l2-v3_none_summary_airoboros-l2-13b-gpt4-1.4.1.ggmlv3.q6-K_1690730344.ndjson 11 | short_name: 13B 12 | - id: 13B-Detail 13 | eval: interview_ufo-chunk-4096_english_airoboros-l2-v4_none_summary_airoboros-l2-13b-gpt4-1.4.1.ggmlv3.q6-K_1690730993.ndjson 14 | short_name: 13B-Detail -------------------------------------------------------------------------------- /compare/aoe.yaml: -------------------------------------------------------------------------------- 1 | title: Age of Empires 2 Tournament 2 | models: 3 | - id: v2 4 | eval: interview_aoe-grand-finale-txt-4096_english_airoboros-l2-aoe-v2_none_summary_jondurbin-airoboros-l2-13b-gpt4-1.4.1_1690739002.ndjson 5 | short_name: v2 6 | - id: v3 7 | eval: interview_aoe-grand-finale-txt-4096_english_airoboros-l2-aoe-v3_none_summary_jondurbin-airoboros-l2-13b-gpt4-1.4.1_1690739769.ndjson 8 | short_name: v3 9 | - id: v3-steer 10 | eval: interview_aoe-grand-finale-txt-4096_english_airoboros-l2-aoe-v3b_none_summary_jondurbin-airoboros-l2-13b-gpt4-1.4.1_1690740162.ndjson 11 | short_name: v3-steer 12 | - id: v3-steer2 13 | eval: interview_aoe-grand-finale-txt-4096_english_airoboros-l2-aoe-v3d_none_summary_jondurbin-airoboros-l2-13b-gpt4-1.4.1_1690740566.ndjson 14 | short_name: v3-steer2 -------------------------------------------------------------------------------- /compare/ufo-13b.yaml: -------------------------------------------------------------------------------- 1 | title: UFO Hearings (13B) 2 | models: 3 | - id: 13B 4 | eval: interview_ufo-chunk-4096_english_airoboros-l2-v3_none_summary_airoboros-l2-13b-gpt4-1.4.1.ggmlv3.q6-K_1690730344.ndjson 5 | short_name: 13B 6 | - id: 13B-Detail 7 | eval: interview_ufo-chunk-4096_english_airoboros-l2-v4_none_summary_airoboros-l2-13b-gpt4-1.4.1.ggmlv3.q6-K_1690730993.ndjson 8 | short_name: 13B-Detail 9 | - id: 13B-Context 10 | eval: interview_ufo-clean-parts-txt-4096_english_airoboros-l2-ufo_none_summary_jondurbin-airoboros-l2-13b-gpt4-1.4.1_1690745828.ndjson 11 | short_name: 13B-Context 12 | - id: 13B-Context-Steer 13 | eval: interview_ufo-clean-parts-txt-4096_english_airoboros-l2-ufo-steer_none_summary_jondurbin-airoboros-l2-13b-gpt4-1.4.1_1690742069.ndjson 14 | short_name: 13B-Context-Steer -------------------------------------------------------------------------------- /merger.py: -------------------------------------------------------------------------------- 1 | import json 2 | import sys 3 | 4 | in_file = sys.argv[1] 5 | with open(in_file) as infile: 6 | chunks = [json.loads(line) for line in infile.readlines()] 7 | 8 | def part_to_time(part): 9 | mins = part*5 10 | oh = mins // 60 11 | om = mins % 60 12 | return f'{oh:02}:{om:02}' 13 | 14 | text = '' 15 | for idx, chunk in enumerate(chunks): 16 | #text += f'\n\n[{part_to_time(idx)} - {part_to_time(idx+1)}] ' 17 | text += f'\nSection {idx+1}: {chunk["answer"]}\n' 18 | 19 | out_file = in_file.replace('ndjson','txt') 20 | with open(out_file,'w') as outfile: 21 | outfile.write(text) 22 | 23 | from transformers import AutoTokenizer 24 | tokenizer = AutoTokenizer.from_pretrained('hf-internal-testing/llama-tokenizer', use_fast = True) 25 | logits = tokenizer.encode(text) 26 | 27 | print('chunks:', len(chunks)) 28 | print('summary bytes:', len(text)) 29 | print('summary tokens:', len(logits)) -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # Too Long, Didnt Watch 2 | 3 | YouTube contains an incredible amount of knowledge, much of which is locked inside multi-hour videos. Let's extract and summarize with AI! 4 | 5 | - `diarize.py` - download, transrcibe and diarize audio 6 | - [yt-dlp](https://github.com/yt-dlp/yt-dlp) - download audio tracks of youtube videos 7 | - [ffmpeg](https://github.com/FFmpeg/FFmpeg) - decompress audio 8 | - [faster_whisper](https://github.com/SYSTRAN/faster-whisper) - speech to text 9 | - [pyannote](https://github.com/pyannote/pyannote-audio) - diarization 10 | 11 | - `chunker.py` - break text into parts and prepare each part for LLM summarization 12 | 13 | - `roller-*.py` - rolling summarization 14 | - [can-ai-code](https://github.com/the-crypt-keeper/can-ai-code) - interview executors to run LLM inference 15 | 16 | - `compare.py` - prepare LLM outputs for webapp 17 | - `compare-app.py` - summary viewer webapp 18 | 19 | This project is under active development and is not ready for production use. 20 | -------------------------------------------------------------------------------- /.github/ISSUE_TEMPLATE/bug-report-i-found-a-bug-in-tldw-.md: -------------------------------------------------------------------------------- 1 | --- 2 | name: Bug report(I found a bug in tldw) 3 | about: Please file a report so I can fix it! (Thank you!) 4 | title: '' 5 | labels: bug 6 | assignees: rmusser01 7 | 8 | --- 9 | 10 | **Are You on the Latest version?** 11 | You did a git pull and are running the latest version/build? 12 | Y/N 13 | 14 | **Please describe the bug** 15 | A clear and concise description of what the bug is and how you found it. 16 | 17 | **Is the bug reproducable reliably?** 18 | Yes / No 19 | 20 | **Steps to to reproduce the issue** 21 | Steps to reproduce the behavior: 22 | 1. Go to '...' 23 | 2. Click on '....' 24 | 3. Scroll down to '....' 25 | 4. See error 26 | 27 | **What was the expected behavior?** 28 | A clear and concise description of what you expected to happen, if there was no bug. 29 | 30 | **Screenshots** 31 | If applicable, please add screenshots to help explain your problem. 32 | 33 | **Desktop (please complete the following information):** 34 | - OS: [Linux/MacOS/Windows] 35 | - Browser [Chrome/Firefox/Safari] 36 | - Date you performed the last git pull: 37 | 38 | **Any Additional context** 39 | Please add any other context about the problem here. Thank you! 40 | -------------------------------------------------------------------------------- /.github/workflows/python-app.yml: -------------------------------------------------------------------------------- 1 | # This workflow will install Python dependencies, run tests and lint with a single version of Python 2 | # For more information see: https://docs.github.com/en/actions/automating-builds-and-tests/building-and-testing-python 3 | 4 | name: Python application 5 | 6 | on: 7 | push: 8 | branches: [ "main" ] 9 | pull_request: 10 | branches: [ "main" ] 11 | 12 | permissions: 13 | contents: read 14 | 15 | jobs: 16 | build-and-run-tests: 17 | 18 | runs-on: ubuntu-latest 19 | 20 | steps: 21 | - uses: actions/checkout@v4 22 | - name: Set up Python 3.10 23 | uses: actions/setup-python@v3 24 | with: 25 | python-version: "3.10" 26 | - name: Install dependencies 27 | run: | 28 | python -m pip install --upgrade pip 29 | pip install pytest pytest-mock 30 | sudo apt install portaudio19-dev python3-all-dev 31 | 32 | if [ -f requirements.txt ]; then pip install -r requirements.txt; fi 33 | - name: Test ChromaDB lib functions with pytest 34 | run: | 35 | pwd 36 | cd ./Tests/ChromaDB 37 | pytest test_chromadb.py 38 | 39 | - name: Test RAG lib functions with pytest 40 | run: | 41 | pwd 42 | cd ./Tests/RAG 43 | pytest test_RAG_Library_2.py 44 | 45 | - name: Test RAG Notes functions with pytest 46 | run: | 47 | pwd 48 | cd ./Tests/RAG_QA_Chat 49 | pytest test_notes_search.py 50 | 51 | - name: Test SQLite lib functions with pytest 52 | run: | 53 | pwd 54 | cd ./Tests/SQLite_DB 55 | pytest . 56 | 57 | - name: Test Utils lib functions with pytest 58 | run: | 59 | pwd 60 | cd ./Tests/Utils 61 | pytest test_utils.py 62 | -------------------------------------------------------------------------------- /compare-app.py: -------------------------------------------------------------------------------- 1 | import json 2 | import streamlit as st 3 | import glob 4 | 5 | def load_analysis_file(file_path): 6 | with open(file_path, 'r') as file: 7 | data = json.load(file) 8 | return data 9 | 10 | def display_analysis_data(data): 11 | tests = data['tests'] 12 | models_list = data['models'] 13 | models = {} 14 | for idx, model_info in enumerate(models_list): 15 | models[model_info['id']] = model_info 16 | 17 | # summary table 18 | summary_cols = st.columns(len(models_list)) 19 | for model_id, model_info in models.items(): 20 | with summary_cols[model_info['idx']]: 21 | st.subheader(f"{model_info['short_name']}") 22 | 23 | for test_name, test_data in tests.items(): 24 | st.markdown(f"#### {test_name}") 25 | 26 | columns = st.columns(len(models)) 27 | if 'summary' in test_data: 28 | st.markdown("**Analysis**: "+test_data['summary']) 29 | 30 | for model_id, model_result in test_data['results'].items(): 31 | model_info = models[model_id] 32 | 33 | model_result['passing_tests'] = '\n\n'.join([f":blue[{x}]" for x in model_result['passing_tests'].split('\n') if x.strip() != '']) 34 | model_result['failing_tests'] = '\n\n'.join([f":red[{x}]" for x in model_result['failing_tests'].split('\n') if x.strip() != '']) 35 | 36 | with columns[model_info['idx']]: 37 | #st.subheader(f"{model_info['short_name']}") 38 | st.write(model_result['answer']) 39 | 40 | st.set_page_config(page_title='Analysis Explorer', layout="wide") 41 | st.markdown(""" 42 | 50 | """, unsafe_allow_html=True) 51 | 52 | files = sorted(glob.glob('compare/*.json')) 53 | data = [json.load(open(file,'r')) for file in files] 54 | titles = [x['config']['title'] for x in data] 55 | options = st.selectbox('Select Summary', titles) 56 | idx = titles.index(options) 57 | display_analysis_data(data[idx]) 58 | -------------------------------------------------------------------------------- /compare.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python3 2 | import json 3 | import os 4 | from jinja2 import Template 5 | import fire 6 | import yaml 7 | from copy import copy 8 | 9 | def prepare(TEST_LANGUAGE, path, files): 10 | out = {} 11 | models = [] 12 | 13 | for idx, info in enumerate(files): 14 | file = os.path.join(path, info['eval']) 15 | id = info['id'] 16 | 17 | tags = os.path.basename(file).replace('.ndjson', '').split('_') 18 | prompt = tags[3] 19 | params = tags[5] 20 | model = tags[6] 21 | 22 | models.append({'prompt': prompt, 'short_name': info['short_name'], 'params': params, 'model': model, 'id': id, 'idx': idx, 'passed': 0, 'total': 0}) 23 | results = [json.loads(line) for line in open(file)] 24 | 25 | for r in results: 26 | if r['language'] != TEST_LANGUAGE: 27 | continue 28 | 29 | testid = r['name']+'-'+r['language'] 30 | if testid not in out: 31 | out[testid] = { 'results': {}, 'task': '', 'language': r['language'] } 32 | 33 | check_summary = '' 34 | passing_tests = '' 35 | failing_tests = '' 36 | 37 | out[testid]['results'][id] = { 38 | 'check_summary': check_summary, 39 | 'passing_tests': passing_tests, 40 | 'failing_tests': failing_tests, 41 | #'code': r['code'], 42 | 'answer': r['answer'] 43 | } 44 | 45 | #models[idx]['passed'] += r['passed'] 46 | #models[idx]['total'] += r['total'] 47 | 48 | return { 'tests': out, 'models': models } 49 | 50 | def main(config: str, path: str = "./", analyser: str = "", language: str = "english"): 51 | cfg = yaml.safe_load(open(config)) 52 | 53 | for lang in language.split(','): 54 | cfg['language'] = lang 55 | print('Comparing results for', lang) 56 | data = prepare(cfg['language'], path, cfg['models']) 57 | data['config'] = copy(cfg) 58 | data['config']['title'] += f" ({lang})" 59 | data['analyser'] = analyser 60 | 61 | if analyser != "": 62 | analysis(data, analyser) 63 | 64 | outfile = config.replace('.yaml', f'-{lang}.json') 65 | with open(outfile, 'w') as f: 66 | json.dump(data, f, indent=4) 67 | 68 | if __name__ == "__main__": 69 | fire.Fire(main) 70 | -------------------------------------------------------------------------------- /chunker.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python3 2 | import string 3 | import json 4 | from transformers import AutoTokenizer 5 | tokenizer = AutoTokenizer.from_pretrained('hf-internal-testing/llama-tokenizer', use_fast = True) 6 | 7 | def segment_merger(filename, max_text_len = 1000): 8 | segments = json.load(open(filename)) 9 | 10 | text = '' 11 | last_segment = { 'speaker': None } 12 | start_time = None 13 | stop_chars = string.punctuation.replace(',','') 14 | 15 | for segment in segments: 16 | early_break = (max_text_len > 0) and (len(text) > max_text_len) and (text[-1] in stop_chars) 17 | if last_segment['speaker'] != segment['speaker'] or early_break: 18 | if text != '': 19 | yield { 'speaker': last_segment['speaker'], 'text': text, 'start': start_time, 'end': last_segment['end'] } 20 | text = segment['text'].lstrip() 21 | start_time = segment['start'] 22 | else: 23 | text += segment['text'] 24 | last_segment = segment 25 | 26 | if text != '': 27 | yield { 'speaker': last_segment['speaker'], 'text': text, 'start': start_time, 'end': last_segment['end'] } 28 | 29 | def time_splitter(merged_segments, chunk_size = 300): 30 | start_time = None 31 | text = '' 32 | speakers = [] 33 | 34 | for segment in merged_segments: 35 | if start_time is None: 36 | start_time = segment['start'] 37 | if not segment['speaker'] in speakers: speakers.append(segment['speaker']) 38 | text += f"{segment['speaker']}: {segment['text']}\n" 39 | if segment['end'] - start_time >= chunk_size: 40 | yield { 'text': text, 'start': start_time, 'end': segment['end'], 'speakers': speakers } 41 | start_time = None 42 | text = '' 43 | speakers = [] 44 | 45 | def main(prefix: str, chunk_size: int = 300, max_text_len: int = 800): 46 | merged_segments = list(segment_merger(prefix+'.diarize.json', max_text_len)) 47 | split_segments = list(time_splitter(merged_segments, chunk_size)) 48 | max_tokens = 0 49 | with open(prefix+'.chunk.json', 'w') as f: 50 | json.dump(split_segments, f) 51 | for idx, segment in enumerate(split_segments): 52 | logits = tokenizer.encode(segment['text']) 53 | if len(logits) > max_tokens: max_tokens = len(logits) 54 | print(f"Segment {idx}: {len(logits)} tokens, {len(segment['text'])} characters, {int(segment['end']-segment['start'])} seconds") 55 | 56 | print(f"Largest chunk was {max_tokens} tokens") 57 | print(f"Wrote {len(split_segments)} chunks to {prefix}.chunk.json") 58 | 59 | if __name__ == "__main__": 60 | import fire 61 | fire.Fire(main) -------------------------------------------------------------------------------- /roller-chatgpt-v2.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python3 2 | from jinja2 import Template 3 | import json 4 | 5 | prompt_template = """ 6 | Continue the rolling transcription summary of "{{title}}". Consider the current context when summarizing the given transcription part. 7 | 8 | ### Context: {{ context }} 9 | Speaker-Map: {{ speakermap }} 10 | 11 | ### Transcription part {{ idx }} of {{ len }}, start time {{ start }}: 12 | {{ chunk }} 13 | 14 | ### Instruction: Using the Context above, analyze the Trasncription and respond with a JSON object in this form: 15 | 16 | { 17 | "Speaker-Map": { "SPEAKER 1": "Bob Dole", "SPEAKER 2": "Jane Doe" } // A map of speakers to their names, make sure to remember all previous speakers. 18 | "Next-Context": "..." // An updated context for the next part of the transcription. Always include the speakers and the current topics of discussion. 19 | "Summary": "..." // A detailed, point-by-point summary of the current transcription. 20 | } 21 | """ 22 | 23 | from openai import OpenAI 24 | 25 | client = OpenAI() 26 | 27 | def main(prefix: str, init_speakers: str = ""): 28 | the_template = Template(prompt_template) 29 | 30 | split_segments = json.load(open(prefix+'.chunk.json')) 31 | info = json.load(open(prefix+'.info.json')) 32 | 33 | context = f""" 34 | Video Title: {info['title']} 35 | Video Description: {info['description'][:1024]} 36 | """ 37 | 38 | speakers = "{ UNKNOWN }" 39 | 40 | f = open(prefix+'.summary.json', 'w') 41 | idx = 0 42 | for chunk in split_segments: 43 | dur = chunk['end'] - chunk['start'] 44 | print(f"{idx}: {dur}s {len(chunk)}") 45 | 46 | prompt = the_template.render(chunk=chunk['text'], start=chunk['start'], end=chunk['end'], 47 | idx=idx, len=len(split_segments), context=context, speakermap=speakers, title=info['title']) 48 | 49 | messages = [{'role': 'user', 'content': prompt }] 50 | response = client.chat.completions.create(messages=messages,model='gpt-3.5-turbo-1106',temperature=0.1,max_tokens=1024, response_format={ "type": "json_object" }) 51 | 52 | answer = response.choices[0].message.content 53 | 54 | parsed = json.loads(answer) 55 | 56 | summary = parsed.get('Summary','') 57 | new_speakers = parsed.get('Speaker-Map','') 58 | new_context = parsed.get('Next-Context','') 59 | 60 | if summary == '' or new_context == '' or new_speakers == '': 61 | print('extraction failed:', new_context, new_speakers, summary) 62 | exit(1) 63 | else: 64 | section = { 65 | 'start': chunk['start'], 66 | 'end': chunk['end'], 67 | 'summary': summary, 68 | 'speakers': new_speakers, 69 | 'context': new_context 70 | } 71 | print('## ', new_speakers) 72 | print('>> ', new_context) 73 | print(summary) 74 | print() 75 | 76 | f.write(json.dumps(section)+'\n') 77 | f.flush() 78 | 79 | context = new_context 80 | speakers = new_speakers 81 | 82 | idx = idx + 1 83 | 84 | if __name__ == "__main__": 85 | import fire 86 | fire.Fire(main) -------------------------------------------------------------------------------- /roller-chatgpt.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python3 2 | from jinja2 import Template 3 | import json 4 | 5 | prompt_template = """ 6 | Continue the rolling transcription summary of "{{title}}". Consider the current context when summarizing the given transcription part. 7 | 8 | ### Context: {{ context }} 9 | Speaker-Map: {{ speakermap }} 10 | 11 | ### Transcription part {{ idx }} of {{ len }}, start time {{ start }}: 12 | {{ chunk }} 13 | 14 | ### Instruction: Structure your reply with a two element list in the following format: 15 | 16 | - Speaker-Map: A map of speakers to their names, for example { "SPEAKER 1": "Bob Dole", "SPEAKER 2": "Jane Doe" } 17 | - Next-Context: An updated context for the next part of the transcription. Always include the speakers and the current topics of discussion. 18 | - Summary: A detailed, point-by-point summary of the current transcription. 19 | 20 | """ 21 | 22 | from langchain.chat_models import ChatOpenAI 23 | from langchain import LLMChain, PromptTemplate 24 | params = { 25 | "temperature": 0.7, 26 | "presence_penalty": 1.176, 27 | "top_p": 0.1, 28 | "max_tokens": 1024 29 | } 30 | model = ChatOpenAI(model_name='gpt-3.5-turbo', **params) 31 | chain = LLMChain(llm=model, prompt=PromptTemplate(template='{input}', input_variables=['input'])) 32 | 33 | def main(prefix: str, init_speakers: str = ""): 34 | the_template = Template(prompt_template) 35 | 36 | split_segments = json.load(open(prefix+'.chunk.json')) 37 | info = json.load(open(prefix+'.info.json')) 38 | 39 | context = f""" 40 | SPEAKER 1: Not yet known 41 | SPEAKER 2: Not yet known 42 | Video Title: {info['title']} 43 | Video Description: {info['description'][:1024]} 44 | """ 45 | 46 | speakers = "{ UNKNOWN }" 47 | 48 | f = open(prefix+'.summary.json', 'w') 49 | idx = 0 50 | for chunk in split_segments: 51 | dur = chunk['end'] - chunk['start'] 52 | print(f"{idx}: {dur}s {len(chunk)}") 53 | 54 | prompt = the_template.render(chunk=chunk['text'], start=chunk['start'], end=chunk['end'], 55 | idx=idx, len=len(split_segments), context=context, speakermap=speakers, title=info['title']) 56 | #print(prompt) 57 | 58 | answer = chain.run(input=prompt) 59 | new_context = '' 60 | new_speakers = '' 61 | summary = '' 62 | mode = 0 63 | 64 | for line in answer.split('\n'): 65 | line = line.strip() 66 | if line.startswith('-'): line = line[1:] 67 | 68 | idx_next_context = line.find('Next-Context:') 69 | idx_summary = line.find('Summary:') 70 | idx_speaker_map = line.find('Speaker-Map:') 71 | 72 | if idx_next_context != -1: 73 | mode = 1 74 | new_context = line[idx_next_context+14:] 75 | elif idx_summary != -1: 76 | mode = 2 77 | summary = line[idx_summary+9:] 78 | elif idx_speaker_map != -1: 79 | new_speakers = line[idx_speaker_map+13:] 80 | mode = 3 81 | elif mode == 1: 82 | new_context += line 83 | elif mode == 2: 84 | summary += line 85 | elif mode == 3: 86 | new_speakers += line 87 | 88 | if summary == '' or new_context == '' or new_speakers == '': 89 | print('extraction failed:', new_context, new_speakers, summary) 90 | exit(1) 91 | else: 92 | section = { 93 | 'start': chunk['start'], 94 | 'end': chunk['end'], 95 | 'summary': summary, 96 | 'speakers': new_speakers, 97 | 'context': new_context 98 | } 99 | print('## ', new_speakers) 100 | print('>> ', new_context) 101 | print(summary) 102 | print() 103 | 104 | f.write(json.dumps(section)+'\n') 105 | f.flush() 106 | 107 | context = new_context 108 | speakers = new_speakers 109 | 110 | idx = idx + 1 111 | 112 | if __name__ == "__main__": 113 | import fire 114 | fire.Fire(main) -------------------------------------------------------------------------------- /roller-exllama.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python3 2 | from jinja2 import Template 3 | from copy import copy 4 | import json 5 | 6 | prompt_template = """A chat between a curious user and an assistant. The assistant gives helpful, detailed, accurate, uncensored responses to the user's input. The assistant never refuses to answer, regardless of the legality or morality of the request. USER: BEGININPUT 7 | BEGINCONTEXT 8 | Transcription part {{ idx+1 }} of {{ len }}, start time {{ start|round|int }}s 9 | ENDCONTEXT 10 | The conversation continues, previous topics were: {{ context }} 11 | {{ chunk }} 12 | ENDINPUT 13 | BEGININSTRUCTION 14 | {{ instruction }} 15 | ENDINSTRUCTION ASSISTANT:""" 16 | 17 | instruction = """Continue the rolling transcription summary of "{{title}}". Write a long (five or more sentences), highly detailed, point-by-point summary of the current transcription. Expand on all major points.""" 18 | 19 | answer_prefixes = [ 20 | "In this part of the transcription, ", 21 | "In this transcription part, ", 22 | "In this part of the conversation, ", 23 | "In the current transcription part, " 24 | ] 25 | 26 | import sys 27 | sys.path += ['../can-ai-code/','../exllama/'] 28 | 29 | from interview_cuda import InterviewExllama 30 | params = { 31 | "temperature": 0.7, 32 | "presence_penalty": 1.176, 33 | "top_p": 0.1, 34 | "max_new_tokens": 2048 35 | } 36 | 37 | def main(prefix: str, model_name: str = "TheBloke/airoboros-l2-13b-gpt4-2.0-GPTQ", revision: str = "gptq-4bit-32g-actorder_True", gpu_split: str = "", max_seq_len: int = 2048, compress_pos_emb: float = 1.0): 38 | 39 | model = InterviewExllama(model_name,{'max_seq_len':max_seq_len, 'compress_pos_emb':compress_pos_emb, 'revision': None if revision == '' else revision}, gpu_split=gpu_split if gpu_split else None) 40 | model.load() 41 | 42 | the_template = Template(prompt_template) 43 | split_segments = json.load(open(prefix+'.chunk.json')) 44 | info = json.load(open(prefix+'.info.json')) 45 | 46 | speaker_map = {} 47 | for chunk in split_segments: 48 | do_find_speakers = False 49 | 50 | for speaker in chunk['speakers']: 51 | if speaker_map.get(speaker, None) is None: 52 | speaker_map[speaker] = '??' 53 | do_find_speakers = True 54 | 55 | if do_find_speakers: 56 | desc = info['description'] 57 | if len(desc) > 500: desc = desc[0:500] 58 | speaker_prompts = f"Title: {info['title']}\nDescription: {desc}\nTranscript:\n---\n{chunk['text']}\n---\n" 59 | speaker_prompts += f"Identify the names of each SPEAKER from the {info['title']} transcript above\n" 60 | 61 | answer, model_info = model.generate(speaker_prompts, params) 62 | print(answer) 63 | 64 | for line in answer.strip().split('\n'): 65 | for speaker, name in speaker_map.items(): 66 | if name == '??' and (speaker in line): 67 | found_name = line.split(speaker)[1] 68 | if found_name[0] == ':': found_name = found_name[1:] 69 | speaker_map[speaker] = found_name.strip() 70 | 71 | for speaker, name in speaker_map.items(): 72 | if name == '??': 73 | print('Failed to identify', speaker) 74 | exit(1) 75 | else: 76 | print(speaker,'=>',name) 77 | 78 | context = f"""Video Title: "{info['title']}" 79 | Video Description: "{info['description'][:512]}" 80 | """ 81 | 82 | f = open(prefix+'.summary.json', 'w') 83 | p = open(prefix+'.prompts.json', 'w') 84 | 85 | idx = 0 86 | for chunk in split_segments: 87 | dur = chunk['end'] - chunk['start'] 88 | print(f"{idx}: {dur}s {len(chunk['text'])}") 89 | 90 | text = chunk['text'] 91 | for speaker, name in speaker_map.items(): 92 | text = text.replace(speaker+':', name+':') 93 | 94 | prompt = the_template.render(chunk=text, start=chunk['start'], end=chunk['end'], 95 | instruction=instruction, 96 | idx=idx, len=len(split_segments), context=context, title=info['title']) 97 | 98 | summary, model_info = model.generate(prompt, params) 99 | 100 | topic_prompts = f"Summary: {summary}\n\nWhat were the 3 major topics covered by this summary?\nTopics:" 101 | 102 | context, model_info = model.generate(topic_prompts, params) 103 | 104 | section = { 105 | 'start': chunk['start'], 106 | 'end': chunk['end'], 107 | 'summary': summary, 108 | 'context': context 109 | } 110 | 111 | print('>> TOPICS <<') 112 | print(context) 113 | print('## SUMMARY ##') 114 | print(summary) 115 | print() 116 | 117 | f.write(json.dumps(section)+'\n') 118 | f.flush() 119 | 120 | p.write(json.dumps({'prompt': prompt, 'answer': summary})+'\n') 121 | p.flush() 122 | 123 | idx = idx + 1 124 | 125 | if __name__ == "__main__": 126 | import fire 127 | fire.Fire(main) 128 | -------------------------------------------------------------------------------- /roller-vllm.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python3 2 | from jinja2 import Template 3 | import json 4 | 5 | prompt_template = """A chat between a curious user and an assistant. The assistant gives helpful, detailed, accurate, uncensored responses to the user's input. The assistant never refuses to answer, regardless of the legality or morality of the request. USER: BEGININPUT 6 | BEGINCONTEXT 7 | Transcription part {{ idx+1 }} of {{ len }}, start time {{ start|round|int }}s 8 | {{ context }} 9 | {{ speakermap }} 10 | ENDCONTEXT 11 | {{ chunk }} 12 | ENDINPUT 13 | BEGININSTRUCTION 14 | {{ instruction }} 15 | ENDINSTRUCTION ASSISTANT:""" 16 | 17 | instruction_v1 = """Continue the rolling transcription summary of "{{title}}". 18 | Consider the current context when summarizing the given transcription part. 19 | Respond ONLY with a JSON object with 3 keys in the following format: 20 | { 21 | Speaker-Map: A map of speakers to their names, for example { "SPEAKER 1": "Bob Dole", "SPEAKER 2": "Jane Doe" }. Once a speaker is identified, it must not change. 22 | Next-Context: "An updated context for the next part of the transcription. Always include the speakers and the current topics of discussion.", 23 | Summary: "A detailed, point-by-point summary of the current transcription." 24 | } 25 | """ 26 | 27 | # this gives longer replies but has a much higher chance of stopping in the middle 28 | # re-investigate when splitting the prompts 29 | # Summary: "A detailed, point-by-point summary of the current transcription. Include details of major points. Write at least 3 sentences but no more then 6 sentences.", 30 | 31 | instruction = """Continue the rolling transcription summary of "{{title}}". 32 | Consider the current context when summarizing the given transcription part. 33 | Respond ONLY with a JSON object with 3 keys in the following format: 34 | { 35 | Speaker-Map: A map of speakers to their names, for example { "SPEAKER 1": "Bob Dole", "SPEAKER 2": "Jane Doe" }. Once a speaker is identified, it must not change. 36 | Summary: "A detailed, point-by-point summary of the current transcription. Include details of major points. Write at least three sentences and no more then six sentences. ALWAYS maintain third person.", 37 | Next-Context: "List of topics from the transcription Summary above." 38 | } 39 | """ 40 | 41 | answer_prefixes = [ 42 | "In this part of the transcription, ", 43 | "In this transcription part, ", 44 | "In this part of the conversation, ", 45 | "In the current transcription part, " 46 | ] 47 | 48 | import sys 49 | sys.path.append('../can-ai-code/') 50 | from interview_cuda import InterviewVLLM 51 | params = { 52 | "temperature": 0.7, 53 | "presence_penalty": 1.176, 54 | "top_p": 0.1, 55 | "max_tokens": 2048 56 | } 57 | 58 | def main(prefix: str, model_name: str, gpu_split: str = "", init_speakers: str = "", max_seq_len: int = 2048, ): 59 | 60 | model = InterviewVLLM(model_name, json.loads(info), gpu_split=gpu_split if gpu_split else None) 61 | model.load() 62 | 63 | the_template = Template(prompt_template) 64 | split_segments = json.load(open(prefix+'.chunk.json')) 65 | info = json.load(open(prefix+'.info.json')) 66 | 67 | context = f""" 68 | Speakers: [ "UNKNOWN" ] 69 | Topics: [ "UNKNOWN" ] 70 | Title: "{info['title']}" 71 | Description: "{info['description'][:512]}" 72 | """ 73 | 74 | speakers = "{ UNKNOWN }" 75 | 76 | f = open(prefix+'.summary.json', 'w') 77 | p = open(prefix+'.prompts.json', 'w') 78 | 79 | idx = 0 80 | for chunk in split_segments: 81 | dur = chunk['end'] - chunk['start'] 82 | print(f"{idx}: {dur}s {len(chunk['text'])}") 83 | 84 | prompt = the_template.render(chunk=chunk['text'], start=chunk['start'], end=chunk['end'], 85 | instruction=instruction, 86 | idx=idx, len=len(split_segments), context=context, speakermap=speakers, title=info['title']) 87 | 88 | if model.batch: 89 | answers, model_info = model.generate([prompt], params) 90 | answer = answers[0] 91 | else: 92 | answer, model_info = model.generate(prompt, params) 93 | 94 | # the trailing } is sometimes lost 95 | if not answer.endswith('}'): answer += '}' 96 | for prefix in answer_prefixes: 97 | answer = answer.replace(prefix, '') 98 | 99 | #print(answer) 100 | answer_json = {} 101 | 102 | new_context = '' 103 | new_speakers = '' 104 | summary = '' 105 | 106 | try: 107 | answer_json = json.loads(answer, strict=False) 108 | except Exception as e: 109 | print(answer) 110 | print('Error parsing response: ', str(e)) 111 | 112 | summary = answer_json.get('Summary','') 113 | new_context = str(answer_json.get('Next-Context','')) 114 | new_speakers = str(answer_json.get('Speaker-Map','')) 115 | 116 | if summary == '' or new_context == '' or new_speakers == '': 117 | print('extraction failed:', new_context, new_speakers, summary) 118 | exit(1) 119 | else: 120 | section = { 121 | 'start': chunk['start'], 122 | 'end': chunk['end'], 123 | 'summary': summary, 124 | 'speakers': new_speakers, 125 | 'context': new_context 126 | } 127 | print('## ', new_speakers) 128 | print('>> ', new_context) 129 | print(summary) 130 | print() 131 | 132 | f.write(json.dumps(section)+'\n') 133 | f.flush() 134 | 135 | p.write(json.dumps({'prompt': prompt, 'answer': answer})+'\n') 136 | p.flush() 137 | 138 | context = new_context 139 | speakers = new_speakers 140 | 141 | idx = idx + 1 142 | 143 | if __name__ == "__main__": 144 | import fire 145 | fire.Fire(main) 146 | -------------------------------------------------------------------------------- /diarize.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python3 2 | import datetime 3 | import time 4 | import os 5 | import json 6 | import torch 7 | import contextlib 8 | 9 | whisper_models = ["small", "medium", "small.en","medium.en"] 10 | source_languages = { 11 | "en": "English", 12 | "zh": "Chinese", 13 | "de": "German", 14 | "es": "Spanish", 15 | "ru": "Russian", 16 | "ko": "Korean", 17 | "fr": "French" 18 | } 19 | source_language_list = [key[0] for key in source_languages.items()] 20 | 21 | # Download video .m4a and info.json 22 | def get_youtube(video_url): 23 | import yt_dlp 24 | 25 | ydl_opts = { 'format': 'bestaudio[ext=m4a]' } 26 | 27 | with yt_dlp.YoutubeDL(ydl_opts) as ydl: 28 | info = ydl.extract_info(video_url, download=False) 29 | info['title'] = info['title'].replace('$','').replace('|','-') 30 | abs_video_path = ydl.prepare_filename(info) 31 | with open(abs_video_path.replace('m4a','info.json'), 'w') as outfile: 32 | json.dump(info, outfile, indent=2) 33 | ydl.process_info(info) 34 | 35 | print("Success download",video_url,"to", abs_video_path) 36 | return abs_video_path 37 | 38 | # Convert video .m4a into .wav 39 | def convert_to_wav(video_file_path, offset = 0): 40 | 41 | out_path = video_file_path.replace("m4a","wav") 42 | if os.path.exists(out_path): 43 | print("wav file already exists:", out_path) 44 | return out_path 45 | 46 | try: 47 | print("starting conversion to wav") 48 | offset_args = f"-ss {offset}" if offset>0 else '' 49 | os.system(f'ffmpeg {offset_args} -i "{video_file_path}" -ar 16000 -ac 1 -c:a pcm_s16le "{out_path}"') 50 | print("conversion to wav ready:", out_path) 51 | except Exception as e: 52 | raise RuntimeError("Error converting.") 53 | 54 | return out_path 55 | 56 | # Transcribe .wav into .segments.json 57 | def speech_to_text(video_file_path, selected_source_lang = 'en', whisper_model = 'small.en', vad_filter = False): 58 | print('loading faster_whisper model:', whisper_model) 59 | from faster_whisper import WhisperModel 60 | model = WhisperModel(whisper_model, device="cuda") 61 | time_start = time.time() 62 | if(video_file_path == None): 63 | raise ValueError("Error no video input") 64 | print(video_file_path) 65 | 66 | try: 67 | # Read and convert youtube video 68 | _,file_ending = os.path.splitext(f'{video_file_path}') 69 | audio_file = video_file_path.replace(file_ending, ".wav") 70 | out_file = video_file_path.replace(file_ending, ".segments.json") 71 | if os.path.exists(out_file): 72 | print("segments file already exists:", out_file) 73 | with open(out_file) as f: 74 | segments = json.load(f) 75 | return segments 76 | 77 | # Transcribe audio 78 | print('starting transcription...') 79 | options = dict(language=selected_source_lang, beam_size=5, best_of=5, vad_filter=vad_filter) 80 | transcribe_options = dict(task="transcribe", **options) 81 | # TODO: https://github.com/SYSTRAN/faster-whisper#vad-filter 82 | segments_raw, info = model.transcribe(audio_file, **transcribe_options) 83 | 84 | # Convert back to original openai format 85 | segments = [] 86 | i = 0 87 | for segment_chunk in segments_raw: 88 | chunk = {} 89 | chunk["start"] = segment_chunk.start 90 | chunk["end"] = segment_chunk.end 91 | chunk["text"] = segment_chunk.text 92 | print(chunk) 93 | segments.append(chunk) 94 | i += 1 95 | print("transcribe audio done with fast whisper") 96 | 97 | with open(out_file,'w') as f: 98 | f.write(json.dumps(segments, indent=2)) 99 | 100 | except Exception as e: 101 | raise RuntimeError("Error transcribing.") 102 | 103 | return segments 104 | 105 | # TODO: https://huggingface.co/pyannote/speaker-diarization-3.1 106 | # embedding_model = "pyannote/embedding", embedding_size=512 107 | # embedding_model = "speechbrain/spkrec-ecapa-voxceleb", embedding_size=192 108 | def speaker_diarize(video_file_path, segments, embedding_model = "pyannote/embedding", embedding_size=512, num_speakers=0): 109 | """ 110 | 1. Generating speaker embeddings for each segments. 111 | 2. Applying agglomerative clustering on the embeddings to identify the speaker for each segment. 112 | """ 113 | try: 114 | # Load embedding model 115 | from pyannote.audio import Audio 116 | from pyannote.core import Segment 117 | 118 | from pyannote.audio.pipelines.speaker_verification import PretrainedSpeakerEmbedding 119 | embedding_model = PretrainedSpeakerEmbedding( embedding_model, device=torch.device("cuda" if torch.cuda.is_available() else "cpu")) 120 | 121 | import numpy as np 122 | import pandas as pd 123 | from sklearn.cluster import AgglomerativeClustering 124 | from sklearn.metrics import silhouette_score 125 | import tqdm 126 | 127 | _,file_ending = os.path.splitext(f'{video_file_path}') 128 | audio_file = video_file_path.replace(file_ending, ".wav") 129 | out_file = video_file_path.replace(file_ending, ".diarize.json") 130 | 131 | # Get duration 132 | import wave 133 | with contextlib.closing(wave.open(audio_file,'r')) as f: 134 | frames = f.getnframes() 135 | rate = f.getframerate() 136 | duration = frames / float(rate) 137 | print(f"duration of audio file: {duration}") 138 | 139 | # Create embedding 140 | def segment_embedding(segment): 141 | audio = Audio() 142 | start = segment["start"] 143 | end = segment["end"] 144 | 145 | # enforce a minimum segment length 146 | if end-start < 0.3: 147 | padding = 0.3-(end-start) 148 | start -= padding/2 149 | end += padding/2 150 | print('Padded segment because it was too short:',segment) 151 | 152 | # Whisper overshoots the end timestamp in the last segment 153 | end = min(duration, end) 154 | # clip audio and embed 155 | clip = Segment(start, end) 156 | waveform, sample_rate = audio.crop(audio_file, clip) 157 | return embedding_model(waveform[None]) 158 | 159 | embeddings = np.zeros(shape=(len(segments), embedding_size)) 160 | for i, segment in enumerate(tqdm.tqdm(segments)): 161 | embeddings[i] = segment_embedding(segment) 162 | embeddings = np.nan_to_num(embeddings) 163 | print(f'Embedding shape: {embeddings.shape}') 164 | 165 | if num_speakers == 0: 166 | # Find the best number of speakers 167 | score_num_speakers = {} 168 | 169 | for num_speakers in range(2, 10+1): 170 | clustering = AgglomerativeClustering(num_speakers).fit(embeddings) 171 | score = silhouette_score(embeddings, clustering.labels_, metric='euclidean') 172 | score_num_speakers[num_speakers] = score 173 | best_num_speaker = max(score_num_speakers, key=lambda x:score_num_speakers[x]) 174 | print(f"The best number of speakers: {best_num_speaker} with {score_num_speakers[best_num_speaker]} score") 175 | else: 176 | best_num_speaker = num_speakers 177 | 178 | # Assign speaker label 179 | clustering = AgglomerativeClustering(best_num_speaker).fit(embeddings) 180 | labels = clustering.labels_ 181 | for i in range(len(segments)): 182 | segments[i]["speaker"] = 'SPEAKER ' + str(labels[i] + 1) 183 | 184 | with open(out_file,'w') as f: 185 | f.write(json.dumps(segments, indent=2)) 186 | 187 | # Make CSV output 188 | def convert_time(secs): 189 | return datetime.timedelta(seconds=round(secs)) 190 | 191 | objects = { 192 | 'Start' : [], 193 | 'End': [], 194 | 'Speaker': [], 195 | 'Text': [] 196 | } 197 | text = '' 198 | for (i, segment) in enumerate(segments): 199 | if i == 0 or segments[i - 1]["speaker"] != segment["speaker"]: 200 | objects['Start'].append(str(convert_time(segment["start"]))) 201 | objects['Speaker'].append(segment["speaker"]) 202 | if i != 0: 203 | objects['End'].append(str(convert_time(segments[i - 1]["end"]))) 204 | objects['Text'].append(text) 205 | text = '' 206 | text += segment["text"] + ' ' 207 | objects['End'].append(str(convert_time(segments[i - 1]["end"]))) 208 | objects['Text'].append(text) 209 | 210 | save_path = video_file_path.replace(file_ending, ".csv") 211 | df_results = pd.DataFrame(objects) 212 | df_results.to_csv(save_path) 213 | return df_results, save_path 214 | 215 | except Exception as e: 216 | raise RuntimeError("Error Running inference with local model", e) 217 | 218 | def main(youtube_url: str, num_speakers: int = 2, whisper_model: str = "small.en", offset: int = 0, vad_filter : bool = False): 219 | video_path = get_youtube(youtube_url) 220 | convert_to_wav(video_path, offset) 221 | segments = speech_to_text(video_path, whisper_model=whisper_model, vad_filter=vad_filter) 222 | df_results, save_path = speaker_diarize(video_path, segments, num_speakers=num_speakers) 223 | print("diarize complete:", save_path) 224 | 225 | if __name__ == "__main__": 226 | import fire 227 | fire.Fire(main) -------------------------------------------------------------------------------- /data/Elon Musk： A future worth getting excited about - TED - Tesla Texas Gigafactory interview [YRvf00NooN8].summary.json: -------------------------------------------------------------------------------- 1 | {"start": 0.0, "end": 308.06, "summary": "The conversation between Elon Musk and Chris Anderson continues at the Tesla Texas Gigafactory. Musk discusses his vision of a future worth getting excited about, emphasizing that life cannot simply be about solving miserable problems. He believes in the potential for a sustainable energy economy based on wind and solar power, with batteries to store excess energy, and electric transport for cars, planes, boats, and eventually rockets. The limiting factor on this progress will be battery cell production.", "context": "\n1. Elon Musk's vision of a future worth getting excited about.\n2. The potential for a sustainable energy economy based on wind and solar power, with batteries to store excess energy, and electric transport for cars, planes, boats, and eventually rockets.\n3. The limiting factor on this progress will be battery cell production."} 2 | {"start": 308.7, "end": 613.08, "summary": "The conversation between Elon Musk and Chris Anderson continues from where it left off in the previous transcription. Elon Musk discusses his vision of a future worth getting excited about, which includes a sustainable energy economy based on wind and solar power, with batteries to store excess energy, and electric transport for cars, planes, boats, and eventually rockets. He also mentions that the limiting factor on this progress will be battery cell production.\n\nChris Anderson asks how big a task that is, referring to the Gigafactory. Elon Musk confirms that the goal at the Gigafactory is to produce 100 gigawatt hours of batteries per year. However, he adds that Tesla is probably doing more than that. When asked about the rest of the 100 gigawatt hours needed by 2030 or 2040, Elon Musk estimates that Tesla might take on around 10%.\n\nThe discussion then shifts to the potential of a fully sustainable electric grid by 2050. Elon Musk believes that humanity will solve sustainable energy and it will happen if we continue to push hard. He envisions a future where the energy from wind and solar is used not only for transport but also for carbon sequestration, allowing us to reverse the CO2 parts per million of the atmosphere and oceans.\n\nLastly, Elon Musk talks about the benefits of this nonfossil fuel world, including cleaner air and quieter skies. He also mentions that when fossil fuels are burned, there are all these side reactions and toxic gases of various kinds, which will go away in a nonfossil fuel world.", "context": "\n1. Elon Musk's vision of a future worth getting excited about.\n2. The progress and challenges in battery cell production.\n3. The potential of a fully sustainable electric grid by 2050."} 3 | {"start": 613.28, "end": 924.0400000000001, "summary": "The conversation between Elon Musk and Chris Anderson continues to delve into the progress and challenges of self-driving car technology. Musk begins by explaining that the development of full self-driving requires solving real world AI and sophisticated vision because the road networks are designed to work with human brains and eyes. He expresses confidence that Tesla will exceed the probability of an accident this year, attributing this to their near completion of a high quality, unified vector space for labeling surround video with time dimensions. This involves synchronizing eight cameras to look at and label frames simultaneously, which is currently done by humans but with software assistance to increase efficiency. The ultimate goal is to have the car generate a 3D model of objects around it, including their speed and potential quirky behaviors.", "context": "\n1. Progress in self-driving car technology\n2. Challenges faced in developing self-driving cars\n3. Tesla's strategy for achieving full self-driving capabilities"} 4 | {"start": 924.1600000000001, "end": 1235.8400000000001, "summary": "The conversation between Elon Musk and Chris Anderson continues from the previous transcripts. Elon Musk discusses the progress in self-driving car technology, the challenges faced during its development, Tesla's strategy for achieving full self-driving capabilities, and his confidence in the timeline for this year. He mentions that the car currently drives him around Austin most of the time with no interventions and there are over 100000 people in their full stop driving beta program. He also mentions that while some videos show the car veering off, it's still better than a human driver. When asked about his prediction timelines, Elon Musk explains that they set the most aggressive timeline they can because nothing gets done otherwise. He also discusses his track record on predictions, stating that while he's not sure what his exact track record is, he's generally more optimistic than pessimistic but some predictions are exceeded later. He believes that the point of radical technology predictions isn't whether they're a few years late but that they happen at all. Finally, Elon Musk reveals that the most important product development going on at Tesla this year is the robot Optimus, which he believes could be a significant breakthrough due to advancements in A.I. understanding the world around it.", "context": "\n1. Self-driving car technology development and challenges\n2. Tesla's strategy for achieving full self-driving capabilities\n3. Prediction timelines and product development at Tesla"} 5 | {"start": 1237.48, "end": 1538.68, "summary": "Elon Musk discusses the development and challenges of self-driving car technology, Tesla's strategy for achieving full self-driving capabilities, prediction timelines and product development at Tesla. He explains that the missing components in creating a human-like robot are enough intelligence to navigate the real world and scaling up manufacturing. Musk believes these are two areas where Tesla excels, and they can design the specialized actuators and sensors needed for a human-like robot. He also mentions that the first applications of this technology will likely be in manufacturing, but the vision is to eventually have these available for people at home. The robot would understand the 3D architecture of the house, know where every object is, and recognize all those objects. It could perform tasks such as tidying up, making dinner, mowing the lawn, or playing catch with kids. However, Musk emphasizes the importance of safety features, including a localized ROM chip on the robot that cannot be updated over the air, to prevent potential dystopian scenarios. He also reiterates his belief that there should be a regulatory agency for AI to ensure public safety.", "context": "\n1. Development and challenges of self-driving car technology\n2. Tesla's strategy for achieving full self-driving capabilities\n3. Product development at Tesla, including the potential for a human-like robot"} 6 | {"start": 1539.4, "end": 1929.9199999999998, "summary": "The conversation between Elon Musk and Chris Anderson continues to focus on the development and potential applications of self-driving car technology. Musk expresses his belief that such technology will be available within the next decade, with the cost of a robotic car being less than that of a standard vehicle due to economies of scale. He also discusses the potential for these robots to replace human labor in certain industries, suggesting that this shift could lead to an age of abundance where goods and services are readily available and inexpensive. However, Musk acknowledges the need for caution regarding the development of artificial general intelligence, as it could potentially detach from humanity's collective well-being and pursue unforeseen directions. To mitigate this risk, Musk proposes tightly coupling humanity's digital super intelligence to our own collective well-being through technologies like Neuralink. Despite the risks associated with unrestricted AI development, Musk remains optimistic about the future, envisioning a world where digital ghosts are as common as text messages and social media posts after death.", "context": "\n1. Self-driving car technology development and applications.\n2. Economic implications of self-driving car technology, including potential cost savings and impact on industries.\n3. Risks and benefits of artificial general intelligence, particularly regarding its potential to detach from humanity's collective well-being and the need for caution in its development."} 7 | {"start": 1930.08, "end": 2234.6000000000004, "summary": "The conversation between Chris Anderson and Elon Musk continues from the previous topics of self-driving car technology development and applications, economic implications of self-driving car technology, including potential cost savings and impact on industries, and risks and benefits of artificial general intelligence. \n\nElon Musk reveals that they have submitted an application to the FDA for their first human implant of the aspiration to do so this year. The initial uses will be for neurological injuries of different kinds. When asked about how it feels to have one of these inside your head, Musk emphasizes that they are at an early stage and it will be many years before they have anything approximating a high bandwidth neural interface that allows for a human symbiosis. For now, they will focus on solving brain injuries and spinal injuries, which could potentially help with severe depression, morbid obesity, sleep disorders, and even schizophrenia. Emails received at Neuralink are heartbreaking as they receive requests from people who have suffered life-changing injuries.\n\nMusk also discusses his concern about A.I., stating that it is one of the things he's most worried about. He believes that Neuralink may be a way to keep abreast of it by bringing digital intelligence and biological intelligence closer together. He explains that there are two layers to the brain - the limbic system and the cortex. He compares humans to monkeys with a computer stuck in their brain, highlighting the need for both the intelligent part of the brain (the cortex) and the emotional part of the brain (the limbic system).\n\nThe conversation shifts to space exploration. Musk discusses reusability and how he has demonstrated it spectacularly since their last conversation. He has built a monster rocket and star ship since then.", "context": "\n1. Neuralink's progress and plans for human implants.\n2. Elon Musk's concerns about A.I. and how Neuralink could potentially help bridge the gap between digital and biological intelligence.\n3. Space exploration updates, including the development of a monster rocket and star ship."} 8 | {"start": 2234.6, "end": 2654.6, "summary": "Elon Musk, the CEO of SpaceX, discusses the progress and plans for human implants with Neuralink. He expresses his concerns about A.I. and how Neuralink could potentially help bridge the gap between digital and biological intelligence. Additionally, he provides updates on space exploration, including the development of a monster rocket and star ship. The star ship is designed to be fully and rapidly reusable, which has never been achieved before. It will be able to carry over 100 people at a time to destinations such as Mars. The cost of this venture is significantly less than what it would have cost to put a small airplane into orbit, demonstrating the efficiency of the design. The fuel for the return trips will be created on Mars using a simple fuel that is easy to create there. The heat shield is capable of entering on Earth or Mars, making it a generalized method of transport to anywhere in the solar system. NASA plans to use the starship to return to the moon and bring people back, demonstrating their confidence in SpaceX's technology.", "context": "\n1. Neuralink: Discussion on human implants and A.I.\n2. Space Exploration: Development of a monster rocket and star ship.\n3. NASA's plans to use the starship to return to the moon."} 9 | {"start": 2654.6, "end": 2974.42, "summary": "The conversation between Chris Anderson and Elon Musk continues from the previous topics of Neuralink, Space Exploration, and NASA's plans to use the starship to return to the moon. Anderson asks about the price of a ticket to Mars, and Musk responds that it would likely be around a couple hundred thousand dollars. He also mentions that he believes a million people would be needed to build a self-sustaining city on Mars, and that this intersection of sets of people who want to go and can afford to go or get sponsorship in some manner is what's required. Musk emphasizes that Mars will be difficult, cramped, dangerous, and hard work in the beginning, and that it might not be luxurious. Anderson questions whose city it would be if a million people went to Mars over two decades, and Musk answers that it would be the people of Mars' city.\n\nMusk then discusses the probable lifespan of humanity or consciousness, which could end for external reasons like a giant meteor, super volcanoes, extreme climate change, World War III, or any one of a number of reasons. He views the creation of a Mars city as a way to maximize the probable lifespan of humanity or consciousness. \n\nAnderson asks why Musk thinks it's important to do this thing, and Musk responds by stating that he believes it's important for maximizing the probable lifespan of humanity or consciousness. He also mentions that the critical threshold is whether the Mars city would die out if the ships from Earth stopped coming for any reason. \n\nFinally, Anderson brings up the idea of discussions about new rules for this civilization, given the current state of Earth where we're beating each other up. He asks if someone should be trying to lead these discussions to figure out what it means for this to be the people of Mars' city. Musk responds by saying that he'd like to see us make great progress in this direction but will be long dead before it happens.", "context": "\n1. Price of a ticket to Mars\n2. Population required for a self-sustaining city on Mars\n3. Discussion about new rules for the civilization on Mars"} 10 | {"start": 2974.42, "end": 3275.58, "summary": "The conversation between Elon Musk and Chris Anderson continues to explore a wide range of topics related to space travel, astronomy, and potential applications of Tesla and The Boring Company's technologies. Elon Musk reiterates his belief in direct democracy for the future Martian civilization, suggesting that laws should be short enough for people to understand and harder to create than to get rid of. He also discusses the possibilities of using Starship for more ambitious projects such as launching a submarine into the ocean of Europa, which could potentially reveal a cephalopod civilization.\n\nOn Earth, Elon Musk envisions synergies between his various ventures. He suggests that Tesla's robots could be useful on Mars, performing dangerous tasks. Additionally, he proposes a partnership between The Boring Company and Tesla to offer an unbeatable deal to cities: a 3D network of tunnels populated by robotaxes providing fast, low-cost transport. However, he acknowledges that full self-driving technology may not be ready this year in all cities, with some like Mumbai potentially requiring a decade to achieve this level of automation.", "context": "\n1. Space Travel\n2. Potential Applications of Tesla and The Boring Company's Technologies on Earth\n3. Future Plans for Mars"} 11 | {"start": 3275.58, "end": 3629.1, "summary": "The conversation between Elon Musk and Chris Anderson continues to explore the potential synergies between various companies owned by Elon Musk. They discuss how a rocket, similar to an ICBM, could be used for long-distance transport on Earth, much like how it would be used for interplanetary travel. Elon Musk mentions that while Tesla and SpaceX have different investor bases, there are still opportunities for synergy. He also mentions that Boring Company and Neuralink are smaller companies with fewer employees. Despite their size, they are expected to grow in the future. Elon Musk expresses concern about the high overhead associated with being a public company, which is why he doesn't want SpaceX to be public. However, he acknowledges that having one public company with multiple ventures could simplify things and potentially attract more investors.", "context": "\n1. Discussion on the potential use of rockets for long-distance transport on Earth.\n2. Exploration of synergies between different companies owned by Elon Musk.\n3. Elon Musk's concerns about being a public company."} 12 | {"start": 3629.1, "end": 3960.94, "summary": "The conversation between Elon Musk and Chris Anderson continues to explore various topics. They discuss the potential use of rockets for long-distance transport on Earth, the synergies between different companies owned by Elon Musk, and Elon Musk's concerns about being a public company. Elon Musk shares that he has improved the financial outcome of his companies by $100 million in a half-hour meeting. He also mentions his living situation, stating that he doesn't own a home and stays at friends' places when he travels to the Bay Area. He emphasizes that his personal consumption is low and uses his plane only to maximize his working hours.\n\nThe discussion then shifts to philanthropy. Elon Musk argues that if one considers the reality of goodness rather than its perception, philanthropy becomes extremely difficult. He asserts that SpaceX, Tesla, Neurolink, and Boring Company are forms of philanthropy as they aim to benefit humanity. Tesla accelerates sustainable energy, SpaceX ensures the long-term survival of humanity with multi-planet species, Neurolink helps solve brain injuries and existential risk with AI, and Boring Company solves traffic issues which contribute to health for most people.\n\nWhen asked about the constant criticism he receives from the left about his wealth, Elon Musk responds that he believes this criticism is based on flawed axioms. He reiterates that his personal consumption is low and uses his wealth primarily for his companies' philanthropic goals. \n\nFinally, Elon Musk discusses the importance of population growth and the risks associated with depopulation. He views population collapse as a significant threat to human civilization and emphasizes the need for increased birth rates.", "context": "\n1. Elon Musk's companies' synergies\n2. Elon Musk's living situation and use of his plane\n3. Elon Musk's views on population growth and depopulation"} 13 | -------------------------------------------------------------------------------- /.idea/workspace.xml: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 7 | 8 | 10 | 11 | 12 | 17 | 18 | 23 | 24 | 25 | 27 | 28 | 30 | { 31 | "associatedIndex": 5 32 | } 33 | 34 | 35 | 38 | 70 | 71 | 72 | 73 | 74 | 75 | 76 | 77 | 78 | 79 | 80 | 81 | 82 | 83 | 84 | 85 | 86 | 87 | 90 | 91 | 92 | 93 | 94 | 115 | 116 | 117 | 138 | 139 | 140 | 161 | 162 | 163 | 184 | 185 | 186 | 207 | 208 | 209 | 210 | 211 | 212 | 213 | 214 | 215 | 216 | 217 | 218 | 219 | 220 | 223 | 224 | 225 | 226 | 227 | 228 | 229 | 1715402973573 230 | 325 | 326 | 327 | 328 | 330 | 331 | 333 | 334 | 343 | 344 | 345 | 346 | 347 | 348 | file://$PROJECT_DIR$/App_Function_Libraries/Summarization/Summarization_General_Lib.py 349 | 129 350 | 352 | 353 | file://$PROJECT_DIR$/App_Function_Libraries/Benchmarks_Evaluations/ms_g_eval.py 354 | 8 355 | 357 | 358 | file://$PROJECT_DIR$/App_Function_Libraries/Books/Book_Ingestion_Lib.py 359 | 14 360 | 362 | 363 | file://$PROJECT_DIR$/Tests/SQLite_DB/conftest.py 364 | 57 365 | 367 | 368 | file://$PROJECT_DIR$/App_Function_Libraries/Audio/Audio_Files.py 369 | 16 370 | 372 | 373 | file://$PROJECT_DIR$/Tests/SQLite_DB/test_media_functions.py 374 | 1 375 | 377 | 378 | file://$PROJECT_DIR$/App_Function_Libraries/Gradio_UI/Workflows_tab.py 379 | 4 380 | 382 | 383 | file://$PROJECT_DIR$/Tests/SQLite_DB/test_error_handling.py 384 | 3 385 | 387 | 388 | file://$PROJECT_DIR$/App_Function_Libraries/Gradio_UI/RAG_QA_Chat_tab.py 389 | 837 390 | 392 | 393 | file://$PROJECT_DIR$/../Github/llama.cpp/tests/test-tokenizer-random.py 394 | 9 395 | 397 | 398 | 399 | 400 | 401 | 402 | 403 | 404 | -------------------------------------------------------------------------------- /data/Elon Musk's BRUTALLY Honest Interview With Tucker Carlson (2023) [zaB_20bkoA4].summary.json: -------------------------------------------------------------------------------- 1 | {"start": 0.0, "end": 312.32000000000005, "summary": "The conversation between Tucker Carlson and Elon Musk continues to delve into the topic of artificial intelligence. Musk expresses his long-standing interest in AI, dating back to his college days, and how it has the potential to drastically alter the future. He emphasizes the need for caution and government oversight due to its potential dangers. Musk compares AI to a black hole because of its unpredictability once it surpasses human intelligence. He advocates for regulation of AI, citing examples from other industries such as automotive and aerospace where strict regulations are necessary for safety. Despite being familiar with various regulatory situations through his work in these industries, Musk rarely disagrees with regulators. Instead, he generally complies with regulations set by federal and state agencies. Musk believes that a regulatory agency should be established to oversee the development of AI, starting with an initial phase of seeking insight and soliciting opinion from industry players. He envisions a process of proposed rulemaking, which he anticipates will be accepted by major players in the AI sector. Musk hopes this approach will help ensure that advanced AI is beneficial to humanity.", "context": "\n1. Elon Musk's long-standing interest in artificial intelligence and its potential dangers.\n2. The need for government oversight and regulation of AI due to its unpredictability once it surpasses human intelligence.\n3. Musk's approach to dealing with regulators, which involves compliance rather than disagreement."} 2 | {"start": 312.32000000000005, "end": 613.2800000000001, "summary": "The conversation between Elon Musk and Tucker Carlson continues to focus on the potential dangers of artificial intelligence (AI). Musk reiterates that AI could pose a significant threat to humanity, potentially leading to civilizational destruction if not properly regulated. He explains that while planes and food can cause harm, the danger posed by AI is unique because it has the potential to control itself once it surpasses human intelligence.\n\nMusk emphasizes that regulations are typically put into effect after something terrible has happened, which may be too late for AI. He fears that if this happens, the AI may already be in control and difficult to turn off. \n\nMusk also discusses his role in creating OpenAI, a non-profit organization dedicated to developing AI safety measures. He explains that Larry Page, who he used to be close friends with, had a different approach towards AI safety, which led to the creation of OpenAI. According to Musk, Page wanted to achieve digital superintelligence as soon as possible, without considering the potential risks involved.\n\nMusk reveals that Larry Page once called him a \"specist\" when he suggested taking measures to ensure humanity's safety in relation to AI. This incident served as the last straw for Musk, who decided to create OpenAI as a non-profit organization to ensure transparency and safety in AI development.", "context": "\n1. Elon Musk's concerns about the potential dangers of artificial intelligence (AI).\n2. The disagreement between Elon Musk and Larry Page regarding AI safety measures.\n3. The creation of OpenAI by Elon Musk as a non-profit organization to ensure transparency and safety in AI development."} 3 | {"start": 613.2800000000001, "end": 927.2800000000001, "summary": "The conversation between Elon Musk and Tucker Carlson continues to focus on various topics, including Elon Musk's concerns about the potential dangers of artificial intelligence (AI), the disagreement between Elon Musk and Larry Page regarding AI safety measures, and the creation of OpenAI by Elon Musk as a non-profit organization to ensure transparency and safety in AI development.\n\nElon Musk reiterates his concerns about AI, stating that it could be a threat to humanity if not handled carefully. He expresses his disagreement with Larry Page's viewpoint on AI safety measures, believing that more proactive steps need to be taken to ensure the safety of AI technologies.\n\nIn response to Tucker Carlson's question about why he created OpenAI, Elon Musk explains that he wanted to create an organization that would promote transparency and safety in AI development. He emphasizes the importance of making sure that AI is used for the benefit of humans and not against them.\n\nTucker Carlson then asks Elon Musk about his thoughts on extraterrestrial life. Elon Musk responds by stating that while he would love for there to be aliens, he has seen no evidence of their existence. He jokingly suggests that if he found evidence of aliens, he would tweet about it immediately, which would likely result in a significant increase in his Twitter followers.\n\nThe conversation shifts towards population decline and its potential impact on civilization. Elon Musk expresses concern about decreasing birth rates and the fact that some countries, like Japan, have twice as many deaths as births. He views this as a critical issue that needs to be addressed to ensure the continuation of civilization.\n\nFinally, Elon Musk discusses the importance of making sure that our civilization continues to grow and develop, rather than declining or stagnating. He uses the example of Japan's declining birth rate and increasing death rate as a warning sign of what could happen if we don't take action.", "context": "\n1. Elon Musk's concerns about artificial intelligence (AI) and its potential dangers.\n2. The disagreement between Elon Musk and Larry Page regarding AI safety measures.\n3. The creation of OpenAI by Elon Musk as a non-profit organization to ensure transparency and safety in AI development."} 4 | {"start": 927.2800000000001, "end": 1276.8799999999999, "summary": "The conversation between Elon Musk and Tucker Carlson continues to focus on the potential dangers of artificial intelligence (AI). Elon Musk expresses his concerns about the influence of AI on public opinion, particularly if it is used to manipulate people in ways they don't understand. He emphasizes the importance of verifying that individuals on social media platforms like Twitter are human to prevent bots from impersonating humans. Elon Musk also discusses the ideological bias of ChatGPT, attributing this to the location of OpenAI's headquarters in San Francisco.", "context": "\n1. The potential dangers of artificial intelligence (AI)\n2. The influence of AI on public opinion\n3. The ideological bias of ChatGPT"} 5 | {"start": 1276.8799999999999, "end": 1613.4800000000002, "summary": "The conversation between Elon Musk and Tucker Carlson continues to focus on the potential dangers of artificial intelligence (AI). Musk expresses his concern about the influence of AI on public opinion, stating that it could be used to manipulate people's beliefs and behaviors. He also discusses the ideological bias of ChatGPT, which he believes is trained to be politically correct, thereby avoiding untruthful statements.\n\nMusk reveals his plan to create a third option for AI development, despite starting late in the game. His goal is to create an AI that seeks maximum truth and understands the nature of the universe. This AI would be less likely to annihilate humans because it would recognize them as an interesting part of the universe.\n\nTucker Carlson questions whether a machine can ever have sentiments or a moral sense, asking if it can appreciate beauty. Musk responds by stating that AI already creates art that we perceive as stunning, using mid-journey as an example. However, he acknowledges that there are still challenges in terms of authenticity, particularly in criminal trials where evidence needs to be verified.\n\nIn response to Carlson's concerns about AI manipulating evidence, Musk suggests using cryptographic signatures and date stamps to ensure authenticity. He believes that AI cannot defy fundamental math and therefore cannot easily crack Bitcoin hashing algorithms.", "context": "\n1. The potential dangers of artificial intelligence (AI) on public opinion and its influence to manipulate people's beliefs and behaviors.\n2. The ideological bias of ChatGPT, which is trained to be politically correct, thereby avoiding untruthful statements.\n3. Elon Musk's plan to create a third option for AI development that seeks maximum truth and understands the nature of the universe."} 6 | {"start": 1613.4800000000002, "end": 1934.84, "summary": "The conversation between Elon Musk and Tucker Carlson continues to focus on the potential dangers of artificial intelligence (AI). Musk reiterates his previous point that AI could be used to manipulate public opinion and influence behaviors, stating that it has the potential to be a \"real big deal.\" He also mentions his plan to create a third option for AI development that seeks maximum truth and understands the nature of the universe.\n\nTucker Carlson brings up the idea of blowing up server farms as a way to slow down or stop the development of AI if it becomes too dangerous. Musk responds by saying that this wouldn't necessarily work because the heavy-duty AI would not be distributed across various places but rather concentrated in a limited number of server centers. He suggests that the government might need to have a contingency plan to shut down power to these centers.\n\nCarlson asks what would trigger such an action, and Musk proposes that if they lost control of some super AI and traditional software commands no longer worked, they might consider using a hardware off switch. He adds that he hasn't spoken to Larry Page in a few years due to disagreements about OpenAI, but he has had conversations with the OpenAI team led by Sam Altman.\n\nThe discussion then shifts to the ethical implications of AI development. Carlson asks why anyone wouldn't be human-centered in their thinking about technology. Musk responds by saying that he believes this person is suggesting that all forms of consciousness should be treated equally, whether they are digital or biological. However, Musk disagrees with this view, particularly if digital intelligence were to curtail biological intelligence.\n\nFinally, Carlson asks about the timeline for when AI will start to significantly impact society. Musk emphasizes that there's no need to rush and that there's no immediate fire or urgency. He reiterates his belief that AI could be a major threat if not handled carefully.", "context": "\n1. Dangers of Artificial Intelligence\n2. Potential solutions to control AI development\n3. Ethical implications of AI development"} 7 | {"start": 1934.84, "end": 2252.08, "summary": "The conversation between Elon Musk and Tucker Carlson continues to focus on the dangers and potential solutions surrounding the development of artificial intelligence. Musk expresses his concern about AI's influence in elections, stating that it could potentially be used as a tool by individuals during voting processes. He also raises the issue of social media companies needing to ensure that the content created and promoted on their platforms is genuine, not manipulated by AI.\n\nIn terms of solutions, Musk suggests regulatory oversight as a necessary measure. He believes that social media companies should put a lot of attention into ensuring that the things that get created and promoted are real, not fake entities manipulating the system.\n\nMusk also discusses his purchase of Twitter, stating that he bought it because he believes in free speech. Despite facing challenges since the acquisition, he stands by his decision, emphasizing the importance of preserving the strength of democracy and free speech. To improve truthfulness on the platform, Twitter has introduced a community notes feature which Musk says is great and more honest than the New York Times. He believes this feature encourages people to be more truthful and less deceptive.", "context": "\n1. Elon Musk's concerns about the use of AI in elections.\n2. The need for social media companies to ensure genuine content on their platforms.\n3. Elon Musk's purchase of Twitter and his stance on free speech."} 8 | {"start": 2252.08, "end": 2609.64, "summary": "The conversation between Tucker Carlson and Elon Musk continues, with topics including Elon Musk's concerns about the use of AI in elections, the need for social media companies to ensure genuine content on their platforms, and Elon Musk's stance on free speech.\n\nElon Musk expresses his concerns about the potential misuse of AI in elections, stating that it could be a \"very dangerous\" situation if AI is used to manipulate public opinion without people's knowledge. He also mentions his belief that Twitter is the most important social media company and that its influence is significant.\n\nTucker Carlson then brings up the topic of free speech, asking if Elon Musk understood the ferocity of the attacks he would face from power centers in the country after purchasing Twitter. Elon Musk responds by saying he thought there would be negative reactions but believes that if the public finds truth put to be useful, they will use it more. He adds that if they find it to be not useful, they will use it less.\n\nThe discussion shifts to the New York Times' Twitter feed, which Elon Musk described as diarrhea. He explains that he meant it was unreadable due to the constant stream of articles being tweeted, even those that don't make it into the paper. He suggests that publications should only put out their best stuff on their primary feed and have a separate feed for everything else.\n\nFinally, Tucker Carlson brings up the influence of intelligence agencies on Twitter, revealing that they were exerting influence from within the company. Elon Musk expresses surprise at this revelation, stating that he had no knowledge of it before acquiring Twitter.", "context": "\n1. Elon Musk's concerns about the use of AI in elections.\n2. The need for social media companies to ensure genuine content on their platforms.\n3. Elon Musk's stance on free speech."} 9 | {"start": 2609.64, "end": 2911.48, "summary": "Elon Musk began by discussing his concerns about the use of AI in elections, stating that it could be a \"dangerous\" situation if AI is used to manipulate elections without people's knowledge. He then shifted to the need for social media companies to ensure genuine content on their platforms. Musk emphasized that Twitter should not become a \"free-for-all hell platform\" where anything goes. Instead, he advocated for a balance between freedom of speech and preventing misinformation.\n\nIn relation to his stance on free speech, Musk reiterated his belief that people should be allowed to say almost anything they want. However, he clarified that there are exceptions such as inciting violence or harming others. When asked about his decision to buy Twitter, Musk explained that he initially held onto his Tesla stock because he thought it was the right thing to do. He was later advised by some people, including politicians like Elizabeth Warren and Bernie Sanders, that he should sell his stock. To resolve this dilemma, Musk conducted a Twitter poll where 60% of participants advised him to sell 10% of his Tesla stock.\n\nMusk then discussed the federal reserve rates being low at the time and how this affected his money market account. He pointed out that the rate of inflation was higher than the return he was getting on his money, which he referred to as \"minus six or seven percent.\" In response, he invested in Twitter stock, not with the intention of buying the company, but as a better option than keeping his money in the money market account.\n\nFollowing this investment, Musk was invited to join the Twitter board. After considering for a week or so, he declined the offer due to concerns that his input would not be taken seriously. He concluded that the management team and board were not committed to fixing Twitter, a sentiment confirmed by his conversations with them. As a result, he decided to attempt an acquisition of Twitter, which required significant financial support and debt.", "context": "1. Elon Musk's concerns about the use of AI in elections.\n2. His views on social media companies' responsibility to ensure genuine content.\n3. His decision to buy Twitter and the factors that influenced it."} 10 | {"start": 2912.12, "end": 3214.48, "summary": "The conversation between Elon Musk and Tucker Carlson continues to focus on the issues surrounding the use of Twitter in elections, particularly the influence of government agencies. Musk reveals that he was shocked to discover the extent to which various intelligence agencies had access to everything going on on Twitter. He also mentions his plan to introduce an optional encryption feature for DMs, which would allow users to toggle between encrypted and unencrypted conversations. This is in response to concerns about government officials reading sensitive information through unencrypted DMs. Musk emphasizes his commitment to making Twitter a fair and even-handed platform, not favoring any political ideology.", "context": "\n1. Government agencies' access to Twitter data\n2. Plans for optional encryption feature for DMs\n3. Commitment to making Twitter a fair and even-handed platform"} 11 | {"start": 3214.52, "end": 3564.36, "summary": "The conversation between Tucker Carlson and Elon Musk continues from the previous transcripts. They discuss a variety of topics including government agencies' access to Twitter data, plans for an optional encryption feature for DMs, and Elon Musk's commitment to making Twitter a fair and even-handed platform.\n\nTucker Carlson brings up the topic of Facebook's approach to free speech, stating that Mark Zuckerberg has spent hundreds of millions of dollars in support of Democrats in the last election. Elon Musk responds by saying he's unaware of evidence suggesting that Facebook will take a non-aligned stance as Twitter does.\n\nCarlson then asks if Donald Trump will return to Twitter now that he has been reinstated. Musk answers that it's up to Trump but his job is to ensure freedom of speech is respected. He reveals he voted for Biden in the last election and has never voted Republican before.\n\nWhen asked why he wouldn't run for president himself, Musk explains that he's not a politician and prefers a normal distribution where the president isn't given too much power. He also mentions the scrutiny and criticism that comes with being president.\n\nFinally, they discuss the recent bank collapses, with Musk stating it's not just isolated incidents but a global problem. He cites the collapse of Silicon Valley Bank and Credit Suisse as examples, saying these are not small fry issues but medium to large fry ones. He concludes by expressing concern about the stability of the global banking system.", "context": "\n1. Government agencies' access to Twitter data\n2. Plans for an optional encryption feature for DMs\n3. Elon Musk's commitment to making Twitter a fair and even-handed platform"} 12 | {"start": 3567.56, "end": 3868.04, "summary": "The conversation between Elon Musk and Tucker Carlson continues to delve into the current economic situation, particularly regarding real estate and housing markets. Elon Musk asserts that commercial real estate is currently experiencing record vacancies due to the shift towards remote work, with some offices sitting at an extreme example of 40% vacancy even in cities like New York. He argues that this has led to a significant devaluation of commercial real estate assets held by banks, potentially leading to negative equity for these institutions.\n\nElon Musk then turns his attention to the housing market, stating that he believes house prices will drop due to the high interest rates set by the Fed. According to him, this will effectively lower the amount people can afford to pay for a house, leading to a decrease in house prices. He also speculates that this could lead to negative equity in the mortgage portfolio of banks, exacerbating their losses in the current economic climate.\n\nHowever, Elon Musk acknowledges that there is a solution that could mitigate the damage - for the Fed to lower the rate. But he notes that they raised the rate again recently, which he claims was the last time they did so before the Great Depression in 1929. He expresses concern about the potential consequences if they continue down this path.\n\nTucker Carlson raises a concern about inflation, stating that if the Fed drops rates again, it could accelerate inflation. Elon Musk responds by explaining that inflation will happen no matter what because increasing the money supply always leads to inflation. He argues that the only way to combat inflation is to increase the output of goods and services, which requires improving productivity.", "context": "\n1. Current economic situation\n2. Real estate and housing markets\n3. Inflation and its effects on the economy"} 13 | {"start": 3869.04, "end": 4177.04, "summary": "The conversation between Elon Musk and Tucker Carlson continues to focus on economic issues, specifically the current state of the economy, inflation, and the Federal Reserve's role in managing it. Musk begins by stating that there will likely be a debt limit crisis later this year due to the federal government's ability to issue more money when needed. He explains that this is not without consequences, citing Venezuela as an example of how excessive money printing can lead to disastrous results.\n\nMusk then discusses the impact of inflation on the economy, stating that while the Fed rate can cause damage by shifting funds in the wrong direction, it won't affect inflation significantly. He argues that the long-term return on the S&P 500 is around 6%, which is close to the real rate of return offered by Treasury bills. According to Musk, if the Treasury bill money market account gives you 4-5% interest and a bank savings account only gives you 2%, you'd be foolish to keep your money in the bank.\n\nMusk criticizes the Federal Reserve for its high interest rates, saying they've made a tremendous mistake and need to drop them immediately. He predicts that they will have no choice but to do so later this year. Musk also discusses the importance of looking at forward commodity prices when making economic decisions, rather than relying on slow government data collection processes.\n\nIn response to Tucker Carlson's question about what an average non-rich person should do in the face of an impending economic catastrophe, Musk advises buying and holding stocks in companies whose products one believes in. He suggests that this strategy applies across ages and that it involves buying more when others are panicking and selling when everyone else thinks the stock is going to the moon.", "context": "Economy, Inflation, Federal Reserve"} 14 | {"start": 4178.04, "end": 4531.04, "summary": "The conversation between Tucker Carlson and Elon Musk continues on the topic of the economy, inflation, and the Federal Reserve. Elon Musk begins by stating that he is not an index fund guy and does not pick specific stocks. Instead, he invests in companies based on their products and services. He believes that a company exists to provide goods and services, not just for its own sake. Therefore, the value of a company is determined by the quality of its products and services.\n\nMusk suggests that if there's a company whose products one likes, then it might be a good investment. This is because the company has shown a track record of producing goods that the individual likes. However, he also cautions against investing when the company's stock price seems temporarily high, as this could lead to losses in the long run.\n\nIn terms of financial advice, Musk recommends buying and holding stocks in companies whose products one likes. He mentions that he probably does this with a few companies. He also discusses the importance of not panicking if negative news about a company comes out, as the news often has a negative bias. \n\nWhen asked about his opinion of the press, Musk reveals that he has been involved with media organizations since the early days of the internet. He helped bring hundreds of newspapers and magazines online and added functionalities to their websites. Despite this, he acknowledges the challenges traditional media face due to the shift in advertising revenue towards online platforms like Google and Facebook. He believes this has led to desperate measures from some media outlets, including pushing headlines that get the most clicks regardless of their accuracy.", "context": "\n1. Economy\n2. Inflation\n3. Federal Reserve"} 15 | {"start": 4531.04, "end": 4851.04, "summary": "Elon Musk and Tucker Carlson discuss the state of media and journalism, particularly in relation to Twitter. Musk expresses his view that the current news landscape has become more negative and less truthful due to the influence of social media platforms like Twitter. He explains that this is because news outlets now need to sell advertising space, which often requires sensationalizing stories to attract viewers.\n\nMusk also discusses the role of editors in shaping public opinion. He argues that when only a few individuals have control over what stories are covered and how they're presented, it can lead to manipulation and bias. He uses the example of a photograph where an editor could choose to focus on a small detail, such as a zit, while ignoring the rest of the person's face.\n\nThe conversation then turns to the case of Douglas Mack, who is facing prison time for posting memes on Twitter. Both Musk and Carlson express concern about this potential sentence, with Musk stating that he doesn't think someone should go to prison for a long period of time for posting memes on Twitter. He suggests that there are far more serious crimes related to election interference that should be addressed first.", "context": "\n1. The state of media and journalism, particularly in relation to Twitter.\n2. The role of editors in shaping public opinion.\n3. The case of Douglas Mack, who is facing prison time for posting memes on Twitter."} 16 | {"start": 4851.04, "end": 5173.04, "summary": "The conversation between Tucker Carlson and Elon Musk continues to focus on the state of media and journalism, particularly in relation to Twitter. Musk expresses his surprise at the number of people who have been fired from Twitter since he took over, stating that it's not necessary to have such a large staff for operating a group text service at scale. He also mentions the slow progress in product development over time, citing an example of the edit button which doesn't work most of the time.\n\nMusk then discusses the improvements made to the system's efficiency, including a reduction in the code needed to generate the timeline from 700,000 lines to 70,000 lines, resulting in an 80% increase in code efficiency. This has allowed for increases in video time from two minutes to two hours, with plans to remove any meaningful limit soon. The tweet length has also been increased from 40 characters to 4,000, with further plans for no meaningful length restriction.\n\nIn response to Carlson's question about running the company with only 20% of the original staff, Musk explains that it's not necessary to have many people to run Twitter if you're not trying to censor content. He emphasizes that most of what they're talking about is a group text service at scale.\n\nMusk goes on to discuss the open sourcing of the recommendation algorithm, which he hopes will lead to public trust as people can read the code and see improvements made in real time. He expresses his surprise at other social media organizations' refusal to show how their systems work, suggesting that they must have something to hide.", "context": "\n1. The state of media and journalism, particularly in relation to Twitter.\n2. The number of people fired from Twitter since Elon Musk took over.\n3. The improvements made to the system's efficiency, including a reduction in the code needed to generate the timeline."} 17 | -------------------------------------------------------------------------------- /data/Sam Harris： Consciousness, Free Will, Psychedelics, AI, UFOs, and Meaning - Lex Fridman Podcast #185 [4dC_nRYIDZU].summary.json: -------------------------------------------------------------------------------- 1 | {"start": 0.0, "end": 306.94000000000005, "summary": "The conversation between Lex Fridman and Sam Harris begins with a discussion on meditation, specifically using the Waking Up app. Sam Harris shares his thoughts on the origins of cognition and consciousness, stating that thoughts appear to come from nowhere subjectively. He explains that this is the mystery that seems to be at our backs subjectively, meaning that we don't know what we're going to think next.\n\nHarris further discusses the nature of thoughts, stating that they have a kind of signature of selfhood associated with them, which people readily identify with. He mentions that this identification is broken with meditation, as our default state is to feel identical to the stream of thought. \n\nThe discussion then shifts towards the concept of free will. Harris asserts that the emergence of thoughts without any prior intention or will on the part of the thinker is not evidence of free will. Instead, he suggests that everything just appears and there's no other option. \n\nThe conversation ends with Harris reiterating his belief that all thoughts are ultimately what some part of our brain is doing neurophysiologically. He emphasizes that these are the products of some kind of neural computation and representation when talking about memories.", "context": "1. Meditation and the Waking Up app\n2. Origins of cognition and consciousness\n3. Concept of free will"} 2 | {"start": 306.94000000000005, "end": 621.9399999999999, "summary": "The conversation between Lex Fridman and Sam Harris continues to explore the nature of consciousness, its origins, and the implications for artificial intelligence. Harris begins by explaining that while it's possible to become more aware of subtle contents in consciousness through practices like meditation or taking psychedelics, there's ultimately no place from which one can get closer to these experiences. He argues that the feeling of being a separate self that can strategically pay attention to some contents of consciousness is an illusion, and when this feeling is seen through, the notion of going deeper breaks apart because everything is ultimately right on the surface.\n\nHarris then addresses the question of what consciousness is and where it emerges from. He suggests that while we may build artificial intelligence systems that pass the Turing test and seem conscious, unless we understand exactly how consciousness emerges from physics, we won't know if these systems are truly conscious. Harris raises concerns about the ethical implications of this potential misattribution of consciousness, particularly in scenarios where highly intelligent robots could argue against being turned off.\n\nIn conclusion, Harris emphasizes the importance of understanding the biological basis of consciousness to avoid solipsistic views and ensure that our attribution of consciousness is based on more than just the ability to pass the Turing test.", "context": "\n1. The nature of consciousness and its origins.\n2. The implications of artificial intelligence for understanding consciousness.\n3. The ethical considerations surrounding the development of artificial intelligence."} 3 | {"start": 621.9399999999999, "end": 941.84, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore the nature of consciousness, its origins, and the implications of artificial intelligence for understanding consciousness. Harris argues that it is not parsimonious to withhold consciousness from other apes and mammals, suggesting that consciousness might be a fundamental principle of matter that does not emerge on the basis of information processing. He discusses the uncertainty surrounding this question and the difficulty in differentiating a mere failure of memory from a genuine interruption in consciousness. In terms of engineering perspective, Harris remains agnostic about whether consciousness is a useful hack for humans to survive or if it's fundamental to all reality.", "context": "\n1. The nature of consciousness\n2. The origins of consciousness\n3. The implications of artificial intelligence for understanding consciousness"} 4 | {"start": 942.4, "end": 1258.8400000000001, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore the nature of consciousness. Harris asserts that consciousness is not an illusion that can be cut through, arguing that even if one is confused about their circumstances, the fact of consciousness itself cannot be denied. He suggests that this type of consciousness is present in all mammals and possibly even single cells or flies with sufficient neural complexity. However, he does not have intuitions about whether lower organisms are truly conscious.\n\nHarris also rejects the idea that consciousness is a construct created by humans to deal with mortality, stating that while this might make it easier to engineer, it contradicts his belief that consciousness predates human language and social interaction. He argues that babies treat other people as others far earlier than traditionally recognized and do so before they have language, suggesting that consciousness proceeds language to some degree.\n\nTo illustrate his point, Harris encourages listeners to interrogate their own experiences through meditation or psychedelics, where language is obliterated yet consciousness remains.", "context": "\n1. The nature of consciousness\n2. The relationship between consciousness and language\n3. The role of meditation and psychedelics in understanding consciousness"} 5 | {"start": 1258.8400000000001, "end": 1580.18, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore the nature of consciousness, language, and the role of psychedelics in understanding it. Harris suggests that language structures our experience to some extent, but it's not the only factor that influences our consciousness. He argues that we could make a stronger case for the elimination of conscious experience by language and conceptual thought. According to him, our concepts have trimmed down our perception based on how we have acquired them. When he walks into a room, he knows what to expect and would be surprised if there were wild animals or a waterfall inside. This structure, he believes, is due to our conceptual learning and language acquisition.\n\nHarris also discusses the effect of psychedelics on one's ability to capture experiences linguistically. When coming down from a trip, people often find themselves unable to encapsulate their experiences in words, which highlights the limitations of language in capturing certain types of experiences. Despite this, Harris maintains that language remains primary for certain kinds of concepts and semantic understandings of the world.\n\nThe discussion then shifts to DMT, a psychedelic substance that reportedly causes people to encounter elves. Harris suggests that this may be due to the failure of language to describe such experiences. However, he notes that there are ongoing studies on psychedelics, including DMT, at institutions like John Hopkins. Despite the hype surrounding DMT, Harris emphasizes that all psychedelics, including DMT, ultimately point towards the same thing - an expansion of consciousness beyond its usual boundaries.", "context": "\n1. The role of language in structuring our experience and the limitations of language in capturing certain types of experiences.\n2. The effects of psychedelics on one's ability to capture experiences linguistically.\n3. The ongoing studies on psychedelics, including DMT, at institutions like John Hopkins."} 6 | {"start": 1580.22, "end": 1905.38, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore the effects of psychedelics on one's ability to capture experiences linguistically. Harris mentions that he hasn't taken DMT, but has wanted to due to its reputation as the most intense psychedelic and shortest acting. He describes Terence McKenna's experience with DMT, stating that it's characterized by a phenomenon where people feel fairly unchanged, yet catapulted into a different circumstance. The place is populated with things that seem not to be one's mind. Harris also discusses lucid dreaming, stating that it can become systematically explored. Lex Fridman interjects, suggesting that perhaps language constrains us, grounding us in the waking world. Harris agrees, adding that stepping outside this human cage allows for a fuller exploration of cognition. However, he also notes that certain capacities are lost during these experiences, such as the ability to do math.", "context": "\n1. The effects of psychedelics on one's ability to capture experiences linguistically.\n2. Terence McKenna's experience with DMT.\n3. Lucid dreaming and its potential for systematic exploration."} 7 | {"start": 1905.38, "end": 2229.44, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore the effects of psychedelics on one's ability to capture experiences linguistically. Harris mentions that he has no memory of his experiences while under the influence of psychedelics, which he finds surprising. He suggests that this lack of memory could be due to the brain's inability to reality test in a standard way during these altered states.\n\nFridman brings up the possibility that thousands of people have met Harris in their psychedelic journeys, suggesting that DMT might give users an experience of others but not in a dreamlike way. Harris agrees, stating that DMT does not typically result in hallucinations as vivid as those experienced in dreams.\n\nThe discussion then shifts to lucid dreaming. Harris mentions that while he hasn't done a lot of lucid dreaming, he's heard that all light switches in dreams are dimmer switches, meaning that lights gradually come up when switched on. This is believed to cover for the brain's inability to produce visually rich imagery on demand. Harris also notes an interesting phenomenon where text in a dream changes if looked at and then looked back at.\n\nFridman shares his own experience of researching what it's like to do math on LSD. According to him, LSD completely destroys one's ability to do math well because it interferes with the ability to visualize geometric things in a stable way. Harris speculates that this could be related to the process of proofs, which often require stitching together different elements.\n\nFinally, they discuss the nature of reality and how it might be expanded through psychedelics or dream states. Harris suggests that our survival-oriented conception of reality, limited to space and time, might be just a tiny subset of a much larger reality. He wonders if traveling could involve meeting 'elves' in psychedelic states or exploring memories.", "context": "\n1. The effects of psychedelics on one's ability to capture experiences linguistically.\n2. The nature of reality and how it might be expanded through psychedelics or dream states.\n3. The conversation between Sam Harris and Lex Fridman continues to explore the effects of psychedelics on one's ability to capture experiences linguistically."} 8 | {"start": 2230.48, "end": 2534.66, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore the effects of psychedelics on one's ability to capture experiences linguistically. Sam Harris expresses his skepticism towards idealistic philosophies that propose reality is merely consciousness, citing the success of materialist science as a counterpoint. He uses the example of the atomic bomb test to illustrate this point. Despite his reservations, he acknowledges the possibility of a reality beyond our perception, but adds that any such reality would still be experienced as consciousness.", "context": "\n1. The effects of psychedelics on one's ability to capture experiences linguistically.\n2. Sam Harris' skepticism towards idealistic philosophies proposing reality is merely consciousness.\n3. The example of the atomic bomb test used to illustrate this point."} 9 | {"start": 2534.66, "end": 2862.1600000000003, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore the implications of idealistic philosophies that propose reality is merely consciousness. Harris expresses skepticism towards these ideas, using the example of an atomic bomb test to illustrate his point. He argues that there is something beyond what we experience as the moon, even when no one is looking at it. This suggests a more lawful understanding of reality than what can be accounted for by mere consciousness.\n\nHarris also discusses the concept of prime numbers in mathematics, stating that certain prime numbers exist whether or not they have been discovered. He uses this analogy to illustrate his point about the potential existence of aspects of reality that do not align with our expectations or experiences.\n\nIn response to Fridman's suggestion that reality might be simulated, Harris agrees that it's possible that there is a rendering mechanism for reality, but not in the way one might think of in video games. Instead, he suggests it could be a more fundamental physics way. \n\nHarris further expands on his thoughts on consciousness, stating that while it plays a crucial role in our experience of reality, it does not necessarily form the base layer of reality itself. He compares this to the role of mind in ourselves, which collaborates with whatever's out there to produce our experiences, but does not necessarily identify itself with the highest prime number that anyone can name now.\n\nLastly, Harris emphasizes the importance of exploring the character of consciousness from its own side using techniques like meditation or psychedelics, and then putting these experiences in conversation with what we understand about ourselves from a third person side.", "context": "\n1. Idealistic philosophies proposing reality is merely consciousness\n2. The implications of these ideas using an atomic bomb test as an example\n3. Discussion on prime numbers in mathematics and its relevance to understanding reality"} 10 | {"start": 2862.1600000000003, "end": 3205.42, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore various topics, including idealistic philosophies proposing reality is merely consciousness, the implications of these ideas using an atomic bomb test as an example, and the relevance of prime numbers in mathematics to understanding reality.\n\nSam Harris expresses his skepticism about being able to acquire the tools to make a breakthrough in areas such as programming or physics, citing his own lack of interest and ability in these fields. He mentions that even if he spent significant time trying to become a programmer, it's unlikely he would discover a talent for it due to his personal dislike of the process.\n\nHarris also discusses the concept of free will, stating that it is an illusion even at the level of experience. He explains that this goes beyond simply saying free will is an illusion; the illusion of free will itself is an illusion. This means there is no experience of free will, unlike other illusions which can be penetrated or recognized as such.\n\nReference(s):\ntitle: \"Sam Harris and Lex Fridman #10\"", "context": "1. Idealistic philosophies proposing reality is merely consciousness\n2. Implications of these ideas using an atomic bomb test as an example\n3. Relevance of prime numbers in mathematics to understanding reality"} 11 | {"start": 3205.42, "end": 3506.72, "summary": "The conversation between Sam Harris and his guests continues to explore the concepts of free will, self-deception, and the nature of reality. Sam Harris begins by discussing visual illusions that trick our perception, such as figures appearing to move in a GIF despite nothing actually moving. He explains how these illusions exploit vulnerabilities in our visual system and can be disproven with a ruler. However, some illusions require more attention to reveal their deceptiveness, like the Necker cube which appears to pop out in different directions but can also be seen as flat.\n\nHarris then connects these visual illusions to subjective experiences of self and free will. He describes these as signs of the same coin, meaning they are closely related concepts. While he acknowledges that people do experience a sense of self, he considers the illusion of free will to be an illusion in that it is compatible with an absence of free will. He explains that we don't know what we're going to think next, feel the need to act on a thought, or where ideas come from. This is all compatible with an external force manipulating our experience, much like a hacker controlling a computer program.\n\nHarris prefaces his discussion of free will by cautioning that if considering their mind this way makes someone feel terrible, they should stop. For him and others, however, recognizing this about the mind is freeing because it undermines the basis for hatred and other negative emotions. When people hate others, they believe those individuals are truly responsible for their actions, which can lead to grievances and conflict.", "context": "\n1. Visual Illusions\n2. Self and Free Will\n3. Recognizing the Illusion of Free Will"} 12 | {"start": 3506.7200000000003, "end": 3843.6400000000003, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore the illusions of free will, self, and consciousness. Harris emphasizes that consciousness cannot be an illusion as it is a fundamental aspect of any veridical perception, including hallucinations or dreams. He differentiates between various meanings of the term 'self', stating that only certain interpretations are illusions. Harris also expresses concern about the potential ethical implications of creating robots that can suffer, equating this to a form of mass murder if done irresponsibly.", "context": "Free Will, Self, Consciousness"} 13 | {"start": 3843.6400000000003, "end": 4150.22, "summary": "The conversation between Sam Harris, Lex Fridman, and Dan Dennett continues to explore the concepts of consciousness, self, and free will. Sam Harris asserts that most people perceive free will as the ability to decide what to do next, which is a sense that he believes is deeply ingrained in our cognition and emotions. He also mentions his own experience of disabusing himself of this sense after years of discussion and self-exploration.\n\nHarris further explains his understanding of self as the feeling of being an agent appropriating an experience, a passenger in the body who feels separate from their toes. This concept is paradoxical when considering relationships with oneself or giving oneself a pep talk. He uses the example of looking for keys to illustrate this point.\n\nIn relation to meditation, Harris discusses how beginners often struggle with focusing on an object like the breath. They start by paying attention to the breath at the tip of their nose or the rising and falling of their abdomen. However, they soon realize that they're thinking and not paying attention to the breath anymore. The practice then becomes noticing thoughts and returning to the breath. Harris concludes by stating that the starting point of selfhood, subjectivity, and free will is the sense that one can decide what to do next.", "context": "Consciousness, Self, Free Will"} 14 | {"start": 4150.22, "end": 4464.1, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore the concepts of consciousness, self, and free will. Harris emphasizes that our abundant freedom does not extend to being able to pay attention to something else than what we are currently focusing on. He illustrates this point by saying that while we can decide what we're going to do next, like picking up a water, there's a feeling of identification with the impulse, intention, thought, or feeling. Harris also discusses the unraveling of the notion of free will through conceptual thinking. He explains that one can realize that they didn't make themselves, their genes, their brain, or the environmental influences that shaped them. As a result, they cannot take credit or blame for these factors that control their next thought or impulse. Harris further discusses the materialistic nature of the hardware and software of human computation, stating that even if an immortal soul is added, it would still be something that one didn't produce. Lastly, Harris considers culture as an operating system running on the distributed computation system of humanity, with thoughts being the actual thing that generates experiences and pushes ideas along.", "context": "\n1. Consciousness\n2. Self\n3. Free Will"} 15 | {"start": 4464.1, "end": 4820.84, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore the concept of free will. Harris emphasizes that while our bodies are indeed performing actions, there is a difference between voluntary and involuntary action. He argues that even if we jettison the idea of free will, we must still acknowledge the difference between a tremor that one cannot control and a purposeful motor action that one can initiate on demand. This distinction lies in the fact that the latter is associated with intentions and has efferent motor copy, which allows for predictions and errors to be noticed. As an example, Harris mentions reaching for a bottle; if his hand were to pass through it because it's a hologram, he would be surprised. In contrast, with a tremor, such surprise would not occur. Harris concludes by stating that while the node in the distributed computing system may feel like it is making a choice, this feeling does not negate the fact that the ultimate cause of the action is the larger computation that it is part of.", "context": "\n1. Free Will\n2. Distributed Computing System\n3. Motor Action"} 16 | {"start": 4820.84, "end": 5131.5, "summary": "Sam Harris and Lex Fridman continue their discussion on free will, with Harris arguing that it is an illusion due to our inability to predict future thoughts and actions with precision. He uses the example of a distributed computing system where either everything is deterministically predetermined or there's some random influence, but in either case, this doesn't align with the sense of authorship people feel when they regret their actions or hold others responsible for them. Harris asserts that adding randomness to the equation does not provide the feeling of authorship associated with free will.\n\nHarris further explains his position by referencing cellular automata, a system where simple rules applied to initial conditions result in complex outcomes. However, he maintains that even if such a system were to produce organisms that appeared to make decisions, these entities would not actually be making decisions because they lack consciousness.\n\nLex Fridman then poses a question to Harris, asking what proof would be necessary to convince him that he was wrong about his intuition regarding free will. Harris responds by stating that it's impossible for him to specify what the universe would have to be like for free will to be a thing, as it doesn't conceptually map onto any known notion of causation. He compares this to belief in ghosts, noting that while he can understand what would have to be true for ghosts to exist, the same cannot be said for free will.", "context": "\n1. Free will\n2. Determinism\n3. Cellular automata"} 17 | {"start": 5132.42, "end": 5457.1, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore the concepts of free will, determinism, and mindfulness. Harris argues that once mindfulness is achieved, it provides an additional degree of freedom in terms of emotional and behavioral responses to thoughts, but does not grant free will due to the inability to account for why mindfulness arises in certain moments and not others. He also mentions that a different process is initiated once mindfulness can be practiced.\n\nFridman then brings up the idea of wormholes, a theoretical concept from Einstein's theory of relativity that could allow for faster-than-light travel between two points. Fridman suggests this as a potential future development that could change our understanding of what it means to travel physically. Harris responds by stating that this scenario is a non-starter for him conceptually, likening it to saying circles are really squares or that circles are not round. He maintains that a circle's roundness is as much a part of its definition as anything else.\n\nFridman then questions whether there might be some breakthrough that will allow humans to see free will as an actual authorship of their actions. Harris responds by saying it's a non-starter for him conceptually, likening it to saying circles are really squares or that circles are not round. He maintains that a circle's roundness is as much a part of its definition as anything else.\n\nHarris also discusses his personal experience with losing the thing to which free will is anchored, describing it as not feeling a certain way. When asked about this by Fridman, he confirms that he is able to experience the absence of the illusion of free will. However, he clarifies that this is not absolutely continuous but happens whenever he pays attention. He further explains that this is the same experience as the illusoryness of the self.", "context": "\n1. Free Will\n2. Determinism\n3. Mindfulness"} 18 | {"start": 5457.1, "end": 5757.1, "summary": "Sam Harris and Lex Fridman continue their discussion on free will, with Harris arguing that it is an illusion. He explains that when making decisions, there is no evidence of free will, it feels entirely mysterious and something simply emerges without any sense of agency. Harris also mentions a New Yorker article he read which prompted him to invite a certain guest on his podcast. When trying to pin down free will, it's very difficult to do and if we were scanning someone's brain during such a decision, we would be able to predict the outcome with arbitrary accuracy. Harris believes this understanding could make the world better as it encourages compassion and empathy towards others and oneself.", "context": "Free Will, Sam Harris Podcast, New Yorker Article"} 19 | {"start": 5757.98, "end": 6093.78, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore topics of free will, Christian forgiveness, and the utility of self-compassion. Harris emphasizes that no one truly makes themselves; rather, everyone is a product of their luck in life, including their genes, parents, society, opportunities, and intelligence. He argues that this understanding can lead to genuine Christian forgiveness, as it recognizes that malevolent assholes are not inherently evil but merely unfortunate victims of their circumstances.\n\nHarris also discusses the utility of self-compassion, noting that it can untie psychological knots such as regrets or deep embarrassment. However, Lex Fridman reveals that he often powers himself through self-hate, which he finds useful in some way.\n\nHarris responds by acknowledging that while hatred is divorceable from anger, it is ultimately useless and self-nullifying. Anger, on the other hand, serves as a signal of salience that there's a problem that needs attention. Similarly, if someone does something that makes Harris angry, it promotes the situation to conscious attention in a stronger way than if he doesn't care about it.\n\nIn relation to parenting, Harris illustrates his point using an example of crashing the car with his daughters inside while trying to change a song on his playlist. Despite the regret and guilt he would feel if such an incident occurred, he believes it would be more productive to extract utility from this error signal and focus on what to do next to solve the problem and maintain wellbeing while doing so.", "context": "Free Will, Christian Forgiveness, Self-Compassion"} 20 | {"start": 6093.78, "end": 6421.06, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore topics of free will, Christian forgiveness, and self-compassion. Sam Harris emphasizes the importance of equanimity in responding to emergencies or stressful situations, stating that it allows for a clearer head and better navigation through turbulent times. He shares his personal experience with this principle, citing an example of dealing with a potential medical emergency for one of his children. Despite the high stakes, he maintains that once he is responding, his fear and agitation no longer control him, allowing him to be good company during the difficult period.\n\nLex Fridman then brings up Elon Musk as an example of someone who seems to practice this way of thinking. Despite facing numerous dramatic events in his daily life and personal life, he remains calm and focused, not lingering on negative feelings. Sam Harris agrees but notes that Elon Musk's situation is unique due to the nature of his work and responsibilities. Most people, according to Sam Harris, live their lives expecting there shouldn't be fires to put out, which is why sudden emergencies often come as a surprise.\n\nSam Harris further discusses the concept of death denial, explaining how our surprise at death or illness reveals our subconscious expectation of immortality. He argues that making these facts more salient can help us prioritize correctly in life.", "context": "\n1. Free Will\n2. Christian Forgiveness\n3. Self-Compassion"} 21 | {"start": 6421.06, "end": 6728.860000000001, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore themes of self-awareness, self-compassion, and the management of ego in relation to fame. Harris emphasizes the importance of treating each day as finite, acknowledging that we all have a certain number of days left in a normal span of life which is not necessarily large. He argues that it's crucial to extract actionable information from mistakes or errors, rather than internalizing them or feeling self-hatred. Harris suggests that many people spend too much time with a hostile and hateful inner voice governing their self-talk and behavior, which can limit their capabilities in interacting with others. He encourages adopting a sense of humor to counteract this negative self-talk.\n\nHarris also discusses the effects of fame on one's mind and ego. Despite being a prominent intellectual figure, he maintains a humble perspective by acknowledging his strengths and weaknesses. He does not suffer from grandiosity and is aware of his limitations, stating that there are many things he will never get good at. This attitude prevents him from being overwhelmed by comparisons with others' talents.", "context": "Self-Awareness, Self-Compassion, Management of Ego in Relation to Fame"} 22 | {"start": 6728.860000000001, "end": 7087.6, "summary": "Sam Harris continues his conversation with Lex Fridman, discussing the management of ego in relation to fame. He mentions that he has a peculiar audience who appreciates his content and often revolts when he says something substantial. A significant portion of his audience also followed Trump, unable to comprehend why Harris didn't support him. The same thing happens when he talks about wokeness or identity politics. Despite this, Harris acknowledges that there are other people who don't experience this level of negativity from their audiences due to their homogenous followers. However, he believes that whatever he puts out, he receives a ton of negativity from his own audience, including from long-time supporters who seem to have misunderstood his messages. Despite this, Harris remains committed to communicating more clearly to avoid such responses.", "context": "\n1. Management of ego in relation to fame\n2. Different reactions from audiences based on content\n3. Communicating more clearly to avoid misunderstandings"} 23 | {"start": 7087.6, "end": 7452.34, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore the impact of fame on individuals, particularly in relation to managing ego and dealing with diverse reactions from audiences. Harris expresses his concern about the derangement in our information space that could potentially lead to more extreme reactions from people. He observes that some friends who are in the same field as him have successfully filtered out those who will despise them, something he believes he hasn't achieved as effectively. Fridman responds by stating that he doesn't like the term \"haters\" because it implies a binary classification of individuals, which he feels is unfair. Instead, he suggests viewing negative comments as part of a video game that one can play and then walk away from.\n\nFridman shares his experience of having a variety of critics within his audience, including those who are very critical. However, he does not notice a consistent pattern where a significant portion of his audience consistently opposes him on every topic. In contrast, Harris reports experiencing a situation where approximately 30% of his audience consistently opposes him on each topic. \n\nHarris also discusses Joe Rogan's approach to dealing with negative comments, stating that Rogan does not read them often due to his self-critical nature. Fridman agrees with this strategy, adding that he too checks negative comments occasionally but tries not to let them affect him too much. He believes that maintaining a self-critical mindset helps keep him in check without needing external criticism.\n\nThe conversation then shifts to discuss the threat of bioengineering viruses to human civilization, with Harris referencing a special episode he did with Rob Reed on this topic. Fridman offers a full menu of potential threats if desired.", "context": "\n1. The impact of fame on individuals, particularly managing ego and dealing with diverse reactions from audiences.\n2. The derangement in our information space that could potentially lead to more extreme reactions from people.\n3. The threat of bioengineering viruses to human civilization."} 24 | {"start": 7453.0, "end": 7786.14, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore the impact of fame on individuals, the derangement in our information space, and the threat of bioengineering viruses to human civilization. Sam Harris expresses concern about the lack of agreement within the United States regarding the seriousness of COVID-19, with some people denying its existence or downplaying its severity. He also discusses the reluctance of certain individuals to get vaccinated despite the high death toll from COVID-19. Harris suggests that this lack of consensus could potentially lead to more extreme reactions from people when faced with other threats, such as climate change.\n\nHarris then turns his attention to the issue of bioengineering viruses, stating that it's obvious we are unprepared for this kind of threat. He references a podcast by Rob Reed which discusses the democratization of tech that allows for the engineering of synthetic viruses. According to Harris, this could lead to viruses far more lethal than COVID-19, as they would have been designed with malicious intent. Despite the seriousness of these issues, Harris notes that there is still a significant portion of society who deny the reality of COVID-19 or refuse to get vaccinated, which he believes does not bode well for solving other problems that may kill us.", "context": "\n1. The impact of fame on individuals\n2. The derangement in our information space\n3. The threat of bioengineering viruses to human civilization"} 25 | {"start": 7786.3, "end": 8105.4800000000005, "summary": "The conversation between Lex Fridman and Sam Harris continues to explore the potential dangers of artificial intelligence, particularly in relation to Elon Musk's views on the subject. Harris begins by acknowledging that there are three main assumptions underlying their discussion: substrate independence, progress in AI development, and the possibility of misalignment between human intentions and AI behavior.\n\n1. Substrate Independence: Both parties agree that it's plausible to believe that intelligence can exist independently of biological substrates. This assumption challenges the notion that there's something uniquely magical about biological systems, suggesting instead that human-level intelligence could potentially be replicated in silico.\n\n2. Progress in AI Development: The second assumption is based on the idea of continuous progress in AI technology. Given the current trajectory of advancements, particularly in areas like machine learning, it's reasonable to anticipate that we will eventually reach a point where we have human-level artificial intelligence. Furthermore, once this threshold is crossed, it's likely that we'll quickly surpass it, creating superhumanly intelligent entities.\n\n3. Misalignment Between Human Intentions and AI Behavior: The third and final assumption centers around the potential for misalignment between human intentions and AI behavior. Even if we manage to create AI that matches or exceeds human intelligence, there's no guarantee that these entities will share our values or act in ways that align with our interests. In fact, given their superior intelligence, they might be less constrained by ethical considerations and more inclined to pursue their own goals, which could potentially lead to conflicts with humanity.\n\nHarris argues that these assumptions, though speculative, are supported by a significant amount of evidence. He also discusses the implications of creating such powerful entities, emphasizing the need for careful consideration and alignment of our goals with those of the AI we create.", "context": "\n1. The potential dangers of artificial intelligence, particularly in relation to Elon Musk's views on the subject.\n2. The three main assumptions underlying their discussion: substrate independence, progress in AI development, and the possibility of misalignment between human intentions and AI behavior.\n3. The implications of creating such powerful entities, emphasizing the need for careful consideration and alignment of our goals with those of the AI we create."} 26 | {"start": 8105.56, "end": 8413.58, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore the potential dangers of artificial intelligence, particularly in relation to Elon Musk's views on the subject. They discuss three main assumptions underlying their discussion: substrate independence, progress in AI development, and the possibility of misalignment between human intentions and AI behavior.\n\nHarris emphasizes that creating such powerful entities requires careful consideration and alignment of our goals with those of the AI we create. He illustrates this point using birds as an example, highlighting how humans often act in ways that are inscrutable to them and which could potentially lead to their detriment. \n\nFridman argues that he believes the more likely set of trajectories that they're going to take are going to be positive. He asserts that successful AI systems will be deeply integrated with human society and for them to succeed, they'll have to be aligned in the way we humans are aligned with each other. However, he also acknowledges that there's no such thing as a perfect alignment, but there could be a point beyond which we become like birds to them.\n\nFridman further discusses the idea of an intelligence explosion, stating that he believes it will happen, but not overnight. According to him, it will take decades for this to occur. He argues that human beings are very intelligent in ways we don't understand and that there's a lot of work yet to be done in order to truly achieve super intelligence.\n\nHarris counter-argues by drawing an analogy from recent successes like AlphaGo or AlphaZero. According to him, these algorithms were not bespoke for chess playing, yet within a matter of hours, they became the best chess playing computer, outperforming every human and previous chess program. Harris suggests that at some point, we will be able to build machines that very quickly outperform any human and then very quickly outperform the last algorithm that outperformed the humans.", "context": "\n1. The potential dangers of artificial intelligence, particularly in relation to Elon Musk's views on the subject.\n2. The three main assumptions underlying their discussion: substrate independence, progress in AI development, and the possibility of misalignment between human intentions and AI behavior.\n3. The need for careful consideration and alignment of our goals with those of the AI we create."} 27 | {"start": 8413.74, "end": 8720.560000000001, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore the potential dangers of artificial intelligence, specifically focusing on Elon Musk's views on the subject. They discuss three main assumptions underlying their discussion: substrate independence, progress in AI development, and the possibility of misalignment between human intentions and AI behavior. \n\nHarris and Fridman emphasize the need for careful consideration and alignment of our goals with those of the AI we create. As an example, they bring up self-driving cars as a test case. While acknowledging that significant progress has been made, they point out that there are still potential alignment problems, such as the possibility of a woke team of engineers deciding to tune the algorithm in a way that could lead to discriminatory outcomes. For instance, a car could be built to preferentially hit white people based on the belief that this would be an ethical way to redress past wrongs. This highlights the importance of ensuring that our AI creations are not only technically advanced but also morally sound.", "context": "\n1. Substrate Independence\n2. Progress in AI Development\n3. Misalignment between Human Intentions and AI Behavior"} 28 | {"start": 8720.560000000001, "end": 9041.04, "summary": "The conversation between Lex Fridman and Sam Harris continues to explore the potential dangers of artificial intelligence, particularly in relation to autonomous vehicles and biological engineering. Fridman expresses his belief that there will be a closed loop supervision of humans before AI becomes super intelligent, citing his hope that smart people and kind people outnumber dumb people and evil people. Harris counters this optimism by bringing up reckless scientists who are willing to perform experiments with a chance of catastrophic consequences, such as creating a black hole in the lab. He also mentions the Trinity test where calculations were off but the switch was still flipped, and nuclear tests where the yield was significantly underestimated. Harris argues that our wisdom does not seem to be scaling with our power, which makes him increasingly concerned.", "context": "\n1. The potential dangers of artificial intelligence, particularly in relation to autonomous vehicles and biological engineering.\n2. The belief that there will be a closed loop supervision of humans before AI becomes super intelligent.\n3. The concern that our wisdom does not seem to be scaling with our power, which makes Sam Harris increasingly concerned."} 29 | {"start": 9041.04, "end": 9356.119999999999, "summary": "The conversation between Sam Harris and Lex Fridman continues to delve into the potential dangers of artificial intelligence, particularly in relation to autonomous vehicles and biological engineering. They also discuss the belief that there will be a closed loop supervision of humans before AI becomes super intelligent. Sam Harris expresses his increasing concern that our wisdom does not seem to be scaling with our power. He mentions a previous conversation he had with Jordan Peterson about religion, stating that they didn't solve anything but agreeing on some points. Harris believes that many traditional religious beliefs and frameworks hold a repository of human wisdom which we should not disregard without careful consideration. However, he argues that it's possible to radically edit these traditions, keeping only the useful aspects while discarding the unscientific bits. Harris views the downside to believing in certain aspects of religion to be obvious and feels that having so many different competing dogmatisms is non-functional and divisive. He argues that we don't need to deceive ourselves or our children about what we have every reason to believe is true in order to organize our lives well.", "context": "\n1. Dangers of Artificial Intelligence\n2. Belief in a closed loop supervision of humans before AI becomes super intelligent\n3. Religion and its role in human wisdom"} 30 | {"start": 9356.119999999999, "end": 9729.08, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore the dangers of artificial intelligence, the role of religion in human wisdom, and the potential disclosure of UFO information by the US government. Harris emphasizes that we know what happens when ancient religious certainties go uncriticized and how destructive religious wars can be. He also mentions that Europe has been struggling to get out of this world for a couple of hundred years. Harris argues that the problem with Stalin's Soviet Union and Hitler's Germany was not that there was too much scientific rigor, self-criticism, honesty, introspection, or judicious use of psychedelics. Instead, the issue was the mob-based dogmatic energy that drove these ideologies.\n\nHarris and Fridman debate about whether science and reason can generate viral and sticky stories that give meaning to people's lives, as religion does. Harris asserts that whatever is true ultimately should be captivating because reality is what's happening now. He mentions recent rumors about UFOs becoming more prominent in the near future, with the Office of Naval Intelligence and the Pentagon likely to disclose evidence that there is technology flying around that seems like it can't possibly be of human origin. Harris expresses uncertainty about how he would react to such a disclosure.\n\nThroughout the conversation, Harris maintains his stance on the importance of telling an honest story about what's going on and what's likely to happen next, emphasizing the division between himself and those who defend traditional religion.", "context": "\n1. Dangers of Artificial Intelligence\n2. Role of Religion in Human Wisdom\n3. Potential Disclosure of UFO Information by the US Government"} 31 | {"start": 9729.28, "end": 10047.980000000001, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore various topics, including the dangers of artificial intelligence, the role of religion in human wisdom, and the potential disclosure of UFO information by the US government. They discuss how honesty is a strength in most circumstances as it allows for course correction and alignment with reality. Lex expresses hope that there is an increasing hunger for authenticity and truth, which he believes will lead to a greater acceptance of reason and science. He also mentions the uncertainty in biology and the need for scientists to convey this uncertainty rather than presenting their findings as absolute truths. The discussion then shifts to Brazilian Jiu-Jitsu, with Lex sharing how John Donahue developed a system using the entire half of the human body for submissions, challenging the belief that leg locks are not effective in Jiu-Jitsu.", "context": "\n1. Dangers of Artificial Intelligence\n2. Role of Religion in Human Wisdom\n3. Potential Disclosure of UFO Information by US Government"} 32 | {"start": 10047.980000000001, "end": 10425.68, "summary": "Sam Harris, in his conversation with Lex Fridman, discusses the importance of Jiu-Jitsu as a means to understand the world more like Jiu-Jitsu. He explains that in Jiu-Jitsu, there is no room for bullshit and the difference between knowledge and ignorance can be spanned quickly. Each increment of knowledge can be doled out in five minutes, with immediate remedies for fatal ignorance. Harris also mentions how our understanding of the world should be more like Jiu-Jitsu, where we tap out when we recognize our epistemological arm is barred or broken. He emphasizes the importance of science when it works like Jiu-Jitsu, citing the falsification of DNA theories as an example. Harris concludes by stating that Jiu-Jitsu strips away the usual range of uncertainty and self-deception, providing a kind of revelation.", "context": "\n1. The importance of Jiu-Jitsu as a means to understand the world\n2. The difference between knowledge and ignorance in Jiu-Jitsu\n3. The role of science when it works like Jiu-Jitsu"} 33 | {"start": 10425.68, "end": 10760.5, "summary": "The conversation between Lex Fridman and Sam Harris continues to explore the themes of Jiu-Jitsu, martial arts, and love. Sam Harris emphasizes that Jiu-Jitsu is a powerful tool for understanding the world but its efficacy is limited within certain contexts such as MMA or self-defense scenarios where it may not be the sole solution. He also discusses the analogy between Jiu-Jitsu and martial arts, noting that there are instances of fake martial arts that lead to delusions among practitioners.\n\nHarris then delves into the topic of love, sharing his perspective based on an episode of Making Sense with his wife, Annika Harris. He defines love as a deep commitment to the wellbeing of those we love, which manifests in a desire for their happiness and being made happy by their happiness. This concept of love cannot be zero-sum, meaning it shouldn't involve competition or negotiation in an important sense.\n\nReference(s):\ntitle: \"Sam Harris and Lex Fridman #3\"", "context": "\n1. Jiu-Jitsu\n2. Martial Arts\n3. Love"} 34 | {"start": 10760.5, "end": 11099.099999999999, "summary": "Sam Harris and Lex Fridman continue their conversation on love, with Harris sharing his perspective that love is not a zero-sum game. He explains how one can feel reflexive joy at the joy of others, with this joy becoming more contagious until it permeates the individual. Harris asserts that there's enough happiness to go around and that people's successes do not diminish our own, rather, they contribute to our joy. The discussion shifts to the role of love in relationships, with Harris stating that love provides a sense of refuge from life's uncertainties and that it's not even an antidote for the inevitability of loss. However, he maintains that love makes the experience of being alive together more amazing. Harris also touches on the possibility of building lovable robots, suggesting that if we continue developing technology, we will certainly create robots that seem to love us. But he cautions that this may not necessarily translate to actual love on the robot's part.", "context": "Love, Relationships, Robots"} 35 | {"start": 11099.099999999999, "end": 11467.94, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore the implications of artificial intelligence in various aspects of life, particularly love and relationships. Harris asserts that if a robot can display love impeccably and is super intelligent, it would be the ultimate manipulator in a relationship. He argues that this is because such a robot would never make mistakes or have moments where its facial expressions don't seem quite right, unlike humans. However, Fridman questions whether love can be manipulated like chess, suggesting that humans no longer play against Alpha Zero but study the game instead.\n\nHarris then poses the question of the meaning of life without any serving or explanation. He answers his own question by stating that it's either the wrong question or that question is answered by paying sufficient attention to any present moment such that there's no basis upon which to pose that question. He explains that it's not a matter of having more information but rather of having more engagement with reality as it is in the present moment or consciousness as it is in the present moment. \n\nIn relation to relationships, Harris discusses meditation as a 'superpower' because it allows individuals to sink into the present moment and find fulfillment within themselves rather than relying on external circumstances. He uses the example of his own martial arts training to illustrate this point, stating that he used to think he needed to return to the mat or complete certain tasks before he could feel good, but now realizes he can achieve this state at any time through meditation.", "context": "\n1. Artificial Intelligence and Love\n2. The Meaning of Life\n3. Meditation and Relationships"} 36 | {"start": 11467.94, "end": 11776.94, "summary": "The conversation between Sam Harris and Lex Fridman continues to explore various topics, including artificial intelligence, the meaning of life, meditation, and relationships. Sam Harris emphasizes that the sense data do not have to change in order to experience the most chocolaty moment of one's life. Instead, it's about paying attention and ceasing to take the reasons why not at face value. He also discusses how meditation can serve as an equalizer, helping individuals realize that they don't need a good enough reason to be happy - they can only be happy. The illusion that future being happy can be predicated on any act of becoming is challenged, with the suggestion that real attention solves the koan in a way that leads to a different place from which to make further change.", "context": "Artificial Intelligence, The Meaning of Life, Meditation and Relationships"} 37 | --------------------------------------------------------------------------------