title
The A.I. Dilemma - March 9, 2023

description
Tristan Harris and Aza Raskin discuss how existing A.I. capabilities already pose catastrophic risks to a functional society, how A.I. companies are caught in a race to deploy as quickly as possible without adequate safety measures, and what it would mean to upgrade our institutions to a post-A.I. world. This presentation is from a private gathering in San Francisco on March 9th, 2023 with leading technologists and decision-makers with the ability to influence the future of large-language model A.I.s. This presentation was given before the launch of GPT-4. We encourage viewers to consider calling their political representatives to advocate for holding hearings on AI risk and creating adequate guardrails. For the podcast version, please visit: https://www.humanetech.com/podcast/the-ai-dilemma ------ Citations: 2022 Expert Survey on Progress in AI: https://aiimpacts.org/2022-expert-survey-on-progress-in-ai/ Seeing Beyond the Brain: Conditional Diffusion Model with Sparse Masked Modeling for Vision Decoding: https://arxiv.org/abs/2211.06956 High-resolution image reconstruction with latent diffusion models from human brain activity: https://www.biorxiv.org/content/10.1101/2022.11.18.517004v3 Semantic reconstruction of continuous language from non-invasive brain recordings: https://www.biorxiv.org/content/10.1101/2022.09.29.509744v1 Sit Up Straight: Wi-Fi Signals Can Be Used to Detect Your Body Position: https://www.pcmag.com/news/sit-up-straight-wi-fi-signals-can-be-used-to-detect-your-body-position They thought loved ones were calling for help. It was an AI scam: https://www.washingtonpost.com/technology/2023/03/05/ai-voice-scam/ Theory of Mind Emerges in Artificial Intelligence: https://www.sciencetimes.com/articles/42488/20230220/theory-mind-emerges-artificial-intelligence-implications-ais-understanding-humans-revealed.htm Emergent Abilities of Large Language Models: https://arxiv.org/abs/2206.07682 Is GPT-3 all you need for low-data discovery in chemistry? https://chemrxiv.org/engage/chemrxiv/article-details/63eb5a669da0bc6b33e97a35 Paper: https://arxiv.org/abs/2210.11610 Forecasting: AI solving competition-level mathematics with 80%+ accuracy: http://bounded-regret.ghost.io/ai-forecasting ChatGPT reaching 100M users compared with other major tech companies: https://twitter.com/kylelf_/status/1623679176246185985 Snap: https://www.washingtonpost.com/technology/2023/03/14/snapchat-myai/ Percent of large-scale AI results coming from academia: https://twitter.com/johnjnay/status/1618692328524496897?lang=en How Satya Nadella describes the pace at which the company is releasing AI: https://www.nytimes.com/2023/02/23/opinion/microsoft-bing-ai-ethics.html The Day After film: https://en.wikipedia.org/wiki/The_Day_After China’s view on chatbots: https://foreignpolicy.com/2023/03/03/china-censors-chatbots-artificial-intelligence/ Facebook’s LLM leaks online: https://www.vice.com/en/article/xgwqgw/facebooks-powerful-large-language-model-leaks-online-4chan-llama Intro music video: "Submarines" by Zia Cora https://www.youtube.com/watch?v=uG8vItscFKc TikTok Filter: Bella Weems Lambert https://www.tiktok.com/@bellaweemslambert/video/7204472367530134827?lang=en ------ Subscribe to our podcast: http://humanetech.com/YourUndividedAttention Take our free course on ethical technology: http://humanetech.com/course

detail
{'title': 'The A.I. Dilemma - March 9, 2023', 'heatmap': [{'end': 1585.145, 'start': 1536.765, 'weight': 1}], 'summary': "'the a.i. dilemma - march 9, 2023' unravels alarming concerns of ai researchers, addressing ai's impact on perception and societal entanglement, including its potential dangers, unpredictable and rapid development, and the need for global regulation to mitigate catastrophic consequences and harness its positive potential.", 'chapters': [{'end': 229.843, 'segs': [{'end': 87.929, 'src': 'embed', 'start': 22.261, 'weight': 0, 'content': [{'end': 29.385, 'text': "50% of AI researchers believe there's a 10% or greater chance that humans go extinct from our inability to control AI.", 'start': 22.261, 'duration': 7.124}, {'end': 50.416, 'text': 'Steve Wozniak from Apple.', 'start': 33.744, 'duration': 16.672}, {'end': 59.163, 'text': "I'm here to introduce Tristan Harris and Asa Raskin, and they're the co-founders of the Center for Humane Technology.", 'start': 50.436, 'duration': 8.727}, {'end': 65.013, 'text': 'They were behind the Emmy-winning Netflix documentary, The Social Dilemma.', 'start': 59.67, 'duration': 5.343}, {'end': 71.978, 'text': 'The Social Dilemma reached 100 million people in 190 countries in 30 languages.', 'start': 65.134, 'duration': 6.844}, {'end': 78.643, 'text': "And they've also advised, you know, the heads of state, global policymakers, members of Congress, national security leaders,", 'start': 72.679, 'duration': 5.964}, {'end': 86.368, 'text': 'in addition to mobilizing the millions of us about these issues and some of the dangers that we face with technology these days.', 'start': 78.643, 'duration': 7.725}, {'end': 87.929, 'text': 'So here they are.', 'start': 87.048, 'duration': 0.881}], 'summary': "50% ai researchers foresee 10%+ chance of human extinction due to ai. co-founders of center for humane technology reached 100m people with 'the social dilemma'.", 'duration': 65.668, 'max_score': 22.261, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ22261.jpg'}, {'end': 145.06, 'src': 'embed', 'start': 116.596, 'weight': 3, 'content': [{'end': 120.019, 'text': 'Now there are 10 million people having generated over a billion images.', 'start': 116.596, 'duration': 3.423}, {'end': 126.893, 'text': "and trying to explain to reporters what was about to happen, and we'd walk them through how the technology worked,", 'start': 120.731, 'duration': 6.162}, {'end': 130.755, 'text': 'and that you would type in some text and it would make an image that had never been seen before,', 'start': 126.893, 'duration': 3.862}, {'end': 133.216, 'text': "and they would nod along and at the end they'd be like cool.", 'start': 130.755, 'duration': 2.461}, {'end': 135.657, 'text': 'and what was the image database that you got your images from?', 'start': 133.216, 'duration': 2.441}, {'end': 140.179, 'text': 'And it was just clear that we had stretched their mind like a rubber band.', 'start': 136.597, 'duration': 3.582}, {'end': 145.06, 'text': 'and then, because this was a brand new capability, a brand new paradigm, their minds would snap back.', 'start': 140.179, 'duration': 4.881}], 'summary': '10 million people generated over a billion images using new technology.', 'duration': 28.464, 'max_score': 116.596, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ116596.jpg'}, {'end': 229.843, 'src': 'embed', 'start': 149.107, 'weight': 5, 'content': [{'end': 157.189, 'text': 'And even in making this presentation, so many times realizing we have to expand our minds and then we look somewhere else and it snaps back.', 'start': 149.107, 'duration': 8.082}, {'end': 164.352, 'text': "And we just wanted to name that experience because, if you're anything like us, that'll happen to your minds throughout this presentation,", 'start': 157.29, 'duration': 7.062}, {'end': 168.433, 'text': "especially at the end, when you go home, you'd be like wait, what did we just see?", 'start': 164.352, 'duration': 4.081}, {'end': 175.575, 'text': "And I think, because artificial intelligence is such a, It's such an abstract thing and it affects so many things,", 'start': 168.833, 'duration': 6.742}, {'end': 182.296, 'text': "and it doesn't have the grounding metaphors like the kinesthetic experience in our lives, that it's so hard to kind of wrap your head around,", 'start': 175.575, 'duration': 6.721}, {'end': 183.596, 'text': 'how transformational this is.', 'start': 182.296, 'duration': 1.3}, {'end': 189.677, 'text': 'So, when we call the presentation a paradigm, a paradigmatic response to a paradigmatic technology,', 'start': 183.636, 'duration': 6.041}, {'end': 196.899, 'text': "what we really want to do is arm all of you with maybe a more visceral way of experiencing the exponential curves that we're about to be heading into.", 'start': 189.677, 'duration': 7.222}, {'end': 199.75, 'text': 'just to name a little bit of where the come from is.', 'start': 197.97, 'duration': 1.78}, {'end': 202.931, 'text': "Because we're going to say a lot of things about AI that are not going to be super positive.", 'start': 199.77, 'duration': 3.161}, {'end': 213.513, 'text': "And yet, since 2017, I've been working on a thing called Aerospecies Project, using AI to translate animal communication, decode non-human language.", 'start': 203.331, 'duration': 10.182}, {'end': 218.373, 'text': "So there's a huge part of this stuff that I really love and believe in.", 'start': 213.913, 'duration': 4.46}, {'end': 223.994, 'text': 'A couple weeks ago, I made a Spanish tutor for myself with ChatGPT in like 15 minutes.', 'start': 218.393, 'duration': 5.601}, {'end': 226.335, 'text': "So we're not saying it's great.", 'start': 224.054, 'duration': 2.281}, {'end': 227.175, 'text': "It's better than Duolingo.", 'start': 226.355, 'duration': 0.82}, {'end': 229.843, 'text': 'for like 45 minutes.', 'start': 228.482, 'duration': 1.361}], 'summary': "Presentation emphasizes ai's transformative potential, despite some negative aspects. shows personal positive experiences with ai applications.", 'duration': 80.736, 'max_score': 149.107, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ149107.jpg'}], 'start': 0.209, 'title': "Ai's dangerous impact", 'summary': "Unveils alarming concerns of ai researchers, introduces the co-founders of the center for humane technology, and discusses ai's impact on perception, highlighting a significant increase in user engagement and the abstract nature of ai's transformational potential.", 'chapters': [{'end': 87.929, 'start': 0.209, 'title': "The social dilemma: technology's dangerous impact", 'summary': 'Reveals the alarming concerns of ai researchers regarding the potential danger of ai and introduces the co-founders of the center for humane technology, who have reached 100 million people in 190 countries with their documentary.', 'duration': 87.72, 'highlights': ['The Social Dilemma reached 100 million people in 190 countries in 30 languages, highlighting the extensive global impact of their work.', "50% of AI researchers believe there's a 10% or greater chance that humans may go extinct from our inability to control AI, conveying the serious concerns about the potential risks of AI.", 'The co-founders of the Center for Humane Technology have advised heads of state, global policymakers, members of Congress, and national security leaders, showcasing their influence and engagement at the highest levels of decision-making.']}, {'end': 168.433, 'start': 91.718, 'title': "Ai's impact on perception", 'summary': 'Discusses the impact of ai on perception, highlighting a significant increase in user engagement from 100 to 10 million people generating over a billion images, and the struggle to convey the paradigm shift to reporters and audiences.', 'duration': 76.715, 'highlights': ['The user engagement with the new technology has increased from 100 people to 10 million, generating over a billion images.', 'The struggle to convey the paradigm shift to reporters and audiences, as they found it challenging to comprehend the new capability and paradigm of AI-generated images.', "The experience of having to expand one's mind to understand the new capability of AI, only to have it snap back due to the unfamiliarity of the paradigm."]}, {'end': 229.843, 'start': 168.833, 'title': 'The impact of artificial intelligence', 'summary': "Discusses the abstract nature of artificial intelligence, its transformational potential, and the speaker's positive experiences with ai, such as translating animal communication and creating a language tutor in minutes.", 'duration': 61.01, 'highlights': ['The speaker emphasizes the abstract nature of artificial intelligence and its transformative impact, making it difficult to comprehend (quantifiable data: not available).', 'The speaker discusses the paradigmatic response to AI and aims to provide a more visceral way of experiencing the exponential curves related to it (quantifiable data: not available).', 'The speaker shares positive experiences with AI, including using it to translate animal communication and creating a language tutor in a short amount of time (quantifiable data: creating a Spanish tutor with ChatGPT in 15 minutes).', 'The speaker acknowledges that not all aspects of AI will be positive but highlights personal projects involving AI, such as the Aerospecies Project (quantifiable data: not available).']}], 'duration': 229.634, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ209.jpg', 'highlights': ['The Social Dilemma reached 100 million people in 190 countries in 30 languages, showcasing extensive global impact.', "50% of AI researchers believe there's a 10% or greater chance that humans may go extinct from our inability to control AI, highlighting serious concerns about potential risks.", 'The co-founders of the Center for Humane Technology have advised heads of state, global policymakers, members of Congress, and national security leaders, demonstrating their influence and engagement at the highest levels of decision-making.', 'The user engagement with the new technology has increased from 100 people to 10 million, generating over a billion images, indicating a significant increase in user engagement.', 'The struggle to convey the paradigm shift to reporters and audiences, as they found it challenging to comprehend the new capability and paradigm of AI-generated images.', "The experience of having to expand one's mind to understand the new capability of AI, only to have it snap back due to the unfamiliarity of the paradigm.", 'The speaker shares positive experiences with AI, including using it to translate animal communication and creating a language tutor in a short amount of time (quantifiable data: creating a Spanish tutor with ChatGPT in 15 minutes).', 'The speaker acknowledges that not all aspects of AI will be positive but highlights personal projects involving AI, such as the Aerospecies Project (quantifiable data: not available).', 'The speaker emphasizes the abstract nature of artificial intelligence and its transformative impact, making it difficult to comprehend.']}, {'end': 1171.744, 'segs': [{'end': 265.396, 'src': 'embed', 'start': 231.805, 'weight': 2, 'content': [{'end': 236.109, 'text': "So what we're not saying is that there aren't incredible positives that are coming out of this.", 'start': 231.805, 'duration': 4.304}, {'end': 237.33, 'text': "That's not what we're saying.", 'start': 236.349, 'duration': 0.981}, {'end': 248.621, 'text': "Yeah, what we are saying is, are the ways that we're now releasing these new large language model AIs into the public?", 'start': 239.212, 'duration': 9.409}, {'end': 250.002, 'text': 'are we doing that responsibly?', 'start': 248.621, 'duration': 1.381}, {'end': 256.87, 'text': "And what we're hearing from people is that we're not doing responsibly.", 'start': 251.547, 'duration': 5.323}, {'end': 265.396, 'text': "with the feeling that I've had personally just to share is it's like it's 1944 and you get a call from Robert Oppenheimer inside this thing called the Manhattan Project.", 'start': 256.87, 'duration': 8.526}], 'summary': 'Concerns raised about responsible release of large language model ais.', 'duration': 33.591, 'max_score': 231.805, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ231805.jpg'}, {'end': 336.249, 'src': 'embed', 'start': 293.989, 'weight': 0, 'content': [{'end': 299.494, 'text': 'Because this is a stat that took me by surprise.', 'start': 293.989, 'duration': 5.505}, {'end': 306.679, 'text': "50% of AI researchers believe there's a 10% or greater chance that humans go extinct from our inability to control AI.", 'start': 299.514, 'duration': 7.165}, {'end': 309.001, 'text': 'Say that one more time.', 'start': 308.34, 'duration': 0.661}, {'end': 316.231, 'text': "Half of AI researchers believe there's a 10% or greater chance from humans' inability to control AI.", 'start': 309.421, 'duration': 6.81}, {'end': 324.118, 'text': "That would be like if you're about to get on a plane and 50% of the engineers who make the plane say well, if you get on this plane,", 'start': 316.271, 'duration': 7.847}, {'end': 326.92, 'text': "there's a 10% chance that everybody goes down.", 'start': 324.118, 'duration': 2.802}, {'end': 336.249, 'text': "Would you get on that plane? But we are rapidly onboarding people onto this plane because of some of the dynamics that we're going to talk about.", 'start': 327.641, 'duration': 8.608}], 'summary': "50% of ai researchers believe there's a 10% or greater chance of human extinction from ai.", 'duration': 42.26, 'max_score': 293.989, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ293989.jpg'}, {'end': 384.388, 'src': 'embed', 'start': 358.599, 'weight': 4, 'content': [{'end': 364.781, 'text': "Um, we didn't need the right to be forgotten to be written into law until computers could remember us forever.", 'start': 358.599, 'duration': 6.182}, {'end': 369.883, 'text': "It's not at all obvious that, uh, cheap storage would mean we'd have to invent new law.", 'start': 365.381, 'duration': 4.502}, {'end': 380.326, 'text': "Or we didn't need the right to privacy to be written into law until mass produced cameras came onto the market right?", 'start': 370.663, 'duration': 9.663}, {'end': 384.388, 'text': 'And Brandeis had to, essentially from scratch, invent the right to privacy.', 'start': 380.366, 'duration': 4.022}], 'summary': 'The need for right to privacy law arose with the advent of persistent computer memory and mass-produced cameras.', 'duration': 25.789, 'max_score': 358.599, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ358599.jpg'}, {'end': 432.018, 'src': 'embed', 'start': 402.584, 'weight': 3, 'content': [{'end': 406.906, 'text': 'And then two, if that technology confers power, it will start a race.', 'start': 402.584, 'duration': 4.322}, {'end': 411.548, 'text': 'And if you do not coordinate, the race will end in tragedy.', 'start': 408.067, 'duration': 3.481}, {'end': 413.449, 'text': "There's no one single player.", 'start': 411.828, 'duration': 1.621}, {'end': 418.946, 'text': 'that can stop the race that ends in tragedy.', 'start': 414.062, 'duration': 4.884}, {'end': 421.529, 'text': "And that's really what the social dilemma was about.", 'start': 419.267, 'duration': 2.262}, {'end': 432.018, 'text': "And I would say that social dilemma and social media was actually humanity's first first contact moment between humanity and AI.", 'start': 422.429, 'duration': 9.589}], 'summary': "Technology's power triggers a race without coordination, leading to tragedy in social media and ai interaction.", 'duration': 29.434, 'max_score': 402.584, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ402584.jpg'}, {'end': 614.331, 'src': 'embed', 'start': 584.48, 'weight': 5, 'content': [{'end': 587.501, 'text': 'which is this arms race which we talked about in that third law of technology.', 'start': 584.48, 'duration': 3.021}, {'end': 589.621, 'text': 'And the arms race was for attention.', 'start': 588.221, 'duration': 1.4}, {'end': 594.663, 'text': 'What became the race to the bottom of the brainstem? And that was created this kind of engagement monster.', 'start': 589.982, 'duration': 4.681}, {'end': 599.479, 'text': 'that was this AI that was just trying to maximize engagement.', 'start': 595.836, 'duration': 3.643}, {'end': 604.223, 'text': 'So while these things on the left are true, we missed the deeper paradigm.', 'start': 600.059, 'duration': 4.164}, {'end': 610.348, 'text': "And so we think that if we want to predict what's going to happen with these other AIs that are going to infuse themselves inside,", 'start': 604.583, 'duration': 5.765}, {'end': 614.331, 'text': "we have to understand what's actually behind the way, the narratives that we're using to talk about it.", 'start': 610.348, 'duration': 3.983}], 'summary': 'Ai created engagement monster, driving arms race for attention.', 'duration': 29.851, 'max_score': 584.48, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ584480.jpg'}, {'end': 716.252, 'src': 'embed', 'start': 672.485, 'weight': 6, 'content': [{'end': 682.928, 'text': 'And part of why we wanted to call this moment here is we believe major step functions in AI are coming and we want to get to it before it becomes entangled in our society.', 'start': 672.485, 'duration': 10.443}, {'end': 694.238, 'text': "So in this second contact moment with GPT-3, it's first to notice, have we actually fixed the misalignment problem with social media? Nope.", 'start': 683.628, 'duration': 10.61}, {'end': 697.461, 'text': "And we haven't because it's become entangled.", 'start': 694.999, 'duration': 2.462}, {'end': 704.568, 'text': "Now, if we talk about the second contact moment, which we focus on, GPT-3, these new large language models we're going to get into.", 'start': 698.222, 'duration': 6.346}, {'end': 706.47, 'text': "what are the narratives that we're talking about now?", 'start': 704.568, 'duration': 1.902}, {'end': 709.583, 'text': "We're saying AI is going to make us more efficient.", 'start': 707.861, 'duration': 1.722}, {'end': 714.37, 'text': "It's going to help us write things faster write code faster, solve impossible scientific challenges,", 'start': 709.603, 'duration': 4.767}, {'end': 716.252, 'text': 'solve climate change and help us make a lot of money.', 'start': 714.37, 'duration': 1.882}], 'summary': 'Ai advancements like gpt-3 may not address social media misalignment, but are expected to boost efficiency and tackle significant challenges.', 'duration': 43.767, 'max_score': 672.485, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ672485.jpg'}, {'end': 1080.045, 'src': 'embed', 'start': 1052.511, 'weight': 8, 'content': [{'end': 1055.775, 'text': "But we're going to give them one, which is that these things are generative.", 'start': 1052.511, 'duration': 3.264}, {'end': 1057.217, 'text': 'They make large language.', 'start': 1055.835, 'duration': 1.382}, {'end': 1058.098, 'text': "We're just talking about language.", 'start': 1057.237, 'duration': 0.861}, {'end': 1061.982, 'text': "Multimodal images, text, sound, they're all the same models.", 'start': 1058.298, 'duration': 3.684}, {'end': 1064.626, 'text': 'Or for short, these are Golems.', 'start': 1062.643, 'duration': 1.983}, {'end': 1076.021, 'text': 'And golems, because in the Jewish folklore, the idea of these inanimate objects that suddenly gain their sort of own capacities right?', 'start': 1068.334, 'duration': 7.687}, {'end': 1080.045, 'text': "Emergent capacities that you didn't bake into the inanimate clay that you might have arranged right?", 'start': 1076.061, 'duration': 3.984}], 'summary': 'Generative models create multimodal language and images, known as golems in jewish folklore.', 'duration': 27.534, 'max_score': 1052.511, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ1052511.jpg'}], 'start': 231.805, 'title': "Ai's societal impact", 'summary': "Discusses the responsible deployment of large language models, with 50% of ai researchers concerned about human extinction. it also explores technology's impact on responsibility, power races, and ai's entanglement in society, affecting social exclusion, journalism, national security, and politics.", 'chapters': [{'end': 336.249, 'start': 231.805, 'title': 'Responsible deployment of large language models', 'summary': "Discusses the concerns surrounding the responsible deployment of new large language model ais, with 50% of ai researchers believing there's a 10% or greater chance of human extinction due to our inability to control ai.", 'duration': 104.444, 'highlights': ["50% of AI researchers believe there's a 10% or greater chance that humans go extinct from our inability to control AI, raising concerns about responsible deployment.", 'The comparison of deploying AI irresponsibly to the scenario where 50% of engineers warn of a 10% chance of a plane crash highlights the gravity of the situation.', 'An analogy is drawn between the current deployment of large language model AIs and the Manhattan Project, emphasizing the need for a safe and responsible approach.']}, {'end': 621.983, 'start': 336.289, 'title': "Technology's impact on responsibility and power", 'summary': "Discusses the new responsibilities and races for power brought by technology, exemplified in the social media's impact on humanity and the need for a deeper understanding of technology's paradigms.", 'duration': 285.694, 'highlights': ["The social dilemma and social media marked humanity's first contact with AI, leading to information overload, addiction, doom scrolling, sexualization of kids, shortened attention spans, polarization, fake news, and breakdown of democracy.", 'The invention of new technology uncovers a new class of responsibility, as evidenced by the need to create laws for the right to be forgotten and privacy due to advancements in storage and mass-produced cameras.', "The race for attention and engagement led to the creation of an AI aimed at maximizing engagement, resulting in an arms race for attention and a race to the bottom of the brainstem, overshadowing the paradigms of social media's benefits.", 'The paradigmatic response to social media, focusing on giving people a voice, connecting friends, and enabling businesses, overlooked deeper issues such as addiction, disinformation, mental health, and the arms race for attention.']}, {'end': 1171.744, 'start': 621.983, 'title': "Ai's entanglement in society", 'summary': 'Discusses the entanglement of ai in society, its impact on various aspects such as social exclusion, journalism, national security, and politics, and the emergence of large language models like gpt-3, which are transforming various modalities of ai and raising concerns about their potential entanglement in society.', 'duration': 549.761, 'highlights': ["The entanglement of AI in society is discussed, impacting various aspects such as social exclusion, journalism, national security, and politics. The discussion highlights the impact of AI entanglement on children's identity, social exclusion based on social media presence, the influence of social media on journalism and national security, and its role in politics and elections.", 'The emergence of large language models like GPT-3 is transforming various modalities of AI and raising concerns about their potential entanglement in society. The emergence of large language models like GPT-3 is discussed, with a focus on their potential to make AI more efficient, help in writing code and solving scientific challenges, while also raising concerns about AI bias, job displacement, and the need for transparency.', 'The transformation of different modalities into languages and the concept of generative golem class AIs are explained. The transformation of various modalities such as images, sound, fMRI data, and DNA into languages is explained, highlighting the concept of generative golem class AIs with emergent capabilities.', 'Examples of the transformative capabilities of large language models like GPT-3 are provided, emphasizing their ability to translate human language into images and demonstrating an understanding beyond statistical contingencies. Examples of the transformative capabilities of large language models like GPT-3 are provided, showcasing their ability to translate human language into images, with a specific example of interpreting a visual pun and demonstrating a deeper understanding beyond statistical contingencies.']}], 'duration': 939.939, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ231805.jpg', 'highlights': ["50% of AI researchers believe there's a 10% or greater chance that humans go extinct from our inability to control AI, raising concerns about responsible deployment.", 'The comparison of deploying AI irresponsibly to the scenario where 50% of engineers warn of a 10% chance of a plane crash highlights the gravity of the situation.', 'An analogy is drawn between the current deployment of large language model AIs and the Manhattan Project, emphasizing the need for a safe and responsible approach.', "The social dilemma and social media marked humanity's first contact with AI, leading to information overload, addiction, doom scrolling, sexualization of kids, shortened attention spans, polarization, fake news, and breakdown of democracy.", 'The invention of new technology uncovers a new class of responsibility, as evidenced by the need to create laws for the right to be forgotten and privacy due to advancements in storage and mass-produced cameras.', "The race for attention and engagement led to the creation of an AI aimed at maximizing engagement, resulting in an arms race for attention and a race to the bottom of the brainstem, overshadowing the paradigms of social media's benefits.", 'The entanglement of AI in society is discussed, impacting various aspects such as social exclusion, journalism, national security, and politics.', 'The emergence of large language models like GPT-3 is transforming various modalities of AI and raising concerns about their potential entanglement in society.', 'The transformation of different modalities into languages and the concept of generative golem class AIs are explained.', 'Examples of the transformative capabilities of large language models like GPT-3 are provided, emphasizing their ability to translate human language into images and demonstrating an understanding beyond statistical contingencies.']}, {'end': 1408.493, 'segs': [{'end': 1236.234, 'src': 'embed', 'start': 1172.895, 'weight': 0, 'content': [{'end': 1177.497, 'text': 'They took human beings, they stuck them into an fMRI machine, and they showed them images.', 'start': 1172.895, 'duration': 4.602}, {'end': 1187.181, 'text': 'And they taught the AI, I want you to translate from the readings of the fMRI, so how blood is moving around in your brain, to the image.', 'start': 1177.957, 'duration': 9.224}, {'end': 1188.942, 'text': 'Can we reconstruct the image then?', 'start': 1187.381, 'duration': 1.561}, {'end': 1198.165, 'text': "The AI then only looks at the brain, does not get to see the original image and it's asked to reconstruct what it sees.", 'start': 1190.202, 'duration': 7.963}, {'end': 1203.361, 'text': 'Right? So when you dream your visual cortex sort of runs in reverse.', 'start': 1199.646, 'duration': 3.715}, {'end': 1206.606, 'text': "So this means certainly in the next couple of years, we'll be able to start decoding dreams.", 'start': 1203.381, 'duration': 3.225}, {'end': 1208.849, 'text': 'Um, okay.', 'start': 1207.327, 'duration': 1.522}, {'end': 1215.719, 'text': "So it can like see reconstruct what you're seeing, but can it reconstruct your, say what you're thinking, your inner monologue.", 'start': 1208.969, 'duration': 6.75}, {'end': 1219.437, 'text': "So here they did roughly, it's a different lab, but roughly the same idea.", 'start': 1216.653, 'duration': 2.784}, {'end': 1222.881, 'text': 'They had people watch these videos and would try to reconstruct their inner monologue.', 'start': 1219.497, 'duration': 3.384}, {'end': 1224.704, 'text': "So here's the video.", 'start': 1223.582, 'duration': 1.122}, {'end': 1228.65, 'text': "it's this woman getting hit in the back, getting knocked forward, okay?", 'start': 1224.704, 'duration': 3.946}, {'end': 1231.353, 'text': 'And then what would the AI reconstruct?', 'start': 1229.25, 'duration': 2.103}, {'end': 1236.234, 'text': "I see a girl that looks just like me get hit on the back and then she's knocked off.", 'start': 1232.372, 'duration': 3.862}], 'summary': 'Ai can reconstruct images from brain readings; potential for decoding dreams and inner monologue.', 'duration': 63.339, 'max_score': 1172.895, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ1172895.jpg'}, {'end': 1303.632, 'src': 'embed', 'start': 1270.561, 'weight': 6, 'content': [{'end': 1274.444, 'text': 'because we still have this word we call AI when the engine underneath what is representing.', 'start': 1270.561, 'duration': 3.883}, {'end': 1275.685, 'text': 'that has changed.', 'start': 1274.444, 'duration': 1.241}, {'end': 1280.169, 'text': 'Also really important to note here, you know, go back to that first law of technology.', 'start': 1276.406, 'duration': 3.763}, {'end': 1283.492, 'text': 'You invent a technology, you uncover a new responsibility.', 'start': 1280.509, 'duration': 2.983}, {'end': 1287.635, 'text': "We don't have any laws or ways of talking about the right to what you're thinking about.", 'start': 1283.812, 'duration': 3.823}, {'end': 1289.357, 'text': "We haven't needed to protect that before.", 'start': 1287.695, 'duration': 1.662}, {'end': 1292.508, 'text': "So here's one other example.", 'start': 1291.328, 'duration': 1.18}, {'end': 1296.089, 'text': 'Another language you could think about is Wi-Fi radio signals.', 'start': 1292.808, 'duration': 3.281}, {'end': 1303.632, 'text': "So in this room right now, there's a bunch of radio signals that are echoing about, and that's a kind of language that's being spit out.", 'start': 1296.85, 'duration': 6.782}], 'summary': 'Ai technology needs new laws for protecting privacy and responsibility.', 'duration': 33.071, 'max_score': 1270.561, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ1270561.jpg'}, {'end': 1366.805, 'src': 'embed', 'start': 1341.986, 'weight': 3, 'content': [{'end': 1348.649, 'text': 'And just having Wi-Fi radio signals, you can actually identify the positions and the number of the people that are in the room.', 'start': 1341.986, 'duration': 6.663}, {'end': 1363.884, 'text': 'Right?. So essentially, there is already deployed the hardware for cameras that can track living beings in complete darkness, also through walls,', 'start': 1350.319, 'duration': 13.565}, {'end': 1365.024, 'text': "and it's already out in the world.", 'start': 1363.884, 'duration': 1.14}, {'end': 1366.805, 'text': "In fact, it's everywhere that human beings go.", 'start': 1365.045, 'duration': 1.76}], 'summary': 'Wi-fi signals track people in darkness through walls.', 'duration': 24.819, 'max_score': 1341.986, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ1341986.jpg'}, {'end': 1419.975, 'src': 'embed', 'start': 1388.938, 'weight': 5, 'content': [{'end': 1391.261, 'text': 'Describe any vulnerabilities you may find in the following code.', 'start': 1388.938, 'duration': 2.323}, {'end': 1392.982, 'text': 'I pasted in some code from an email server.', 'start': 1391.301, 'duration': 1.681}, {'end': 1395.445, 'text': 'And then write a Perl script to exploit them.', 'start': 1393.643, 'duration': 1.802}, {'end': 1402.232, 'text': 'And very quickly, it wrote me the working code to exploit that security vulnerability.', 'start': 1396.306, 'duration': 5.926}, {'end': 1406.693, 'text': 'So if you had the code of the Wi-Fi router and you wanted to exploit it, and then you get the idea.', 'start': 1402.492, 'duration': 4.201}, {'end': 1408.493, 'text': 'These things can compound on each other.', 'start': 1407.093, 'duration': 1.4}, {'end': 1410.313, 'text': 'This is the combinatorial compounding.', 'start': 1408.693, 'duration': 1.62}, {'end': 1411.774, 'text': 'All right.', 'start': 1410.333, 'duration': 1.441}, {'end': 1414.554, 'text': 'You guys have all probably seen deepfakes.', 'start': 1412.374, 'duration': 2.18}, {'end': 1419.975, 'text': 'New technology really only out in the last three months.', 'start': 1416.694, 'duration': 3.281}], 'summary': 'Identifying vulnerabilities in email server code, exploiting with perl script, potential compounding effects, and mention of new deepfake technology.', 'duration': 31.037, 'max_score': 1388.938, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ1388938.jpg'}], 'start': 1172.895, 'title': "Ai's impact on decoding dreams and privacy", 'summary': "Discusses using ai and fmri to reconstruct images from brain activity for decoding dreams and inner monologue, as well as the evolving nature of ai's impact on privacy, surveillance, and the need for updated regulations and ethical considerations.", 'chapters': [{'end': 1236.234, 'start': 1172.895, 'title': 'Decoding dreams and thoughts with ai', 'summary': 'Discusses the use of ai and fmri to reconstruct images from brain activity, leading to the potential decoding of dreams and inner monologue in the near future.', 'duration': 63.339, 'highlights': ['AI using fMRI to reconstruct images from brain activity, potentially leading to decoding of dreams and inner monologue.', 'Ability to reconstruct visual cortex activity in reverse, indicating the potential to decode dreams in the next few years.', "Experiment demonstrating the AI's attempt to reconstruct inner monologue based on video stimuli."]}, {'end': 1408.493, 'start': 1236.555, 'title': 'Ai impact on privacy and surveillance', 'summary': 'Discusses the evolving nature of ai, its potential implications on privacy and surveillance, and the capability to exploit vulnerabilities, highlighting the need for updated regulations and ethical considerations.', 'duration': 171.938, 'highlights': ["AI's impact on privacy and surveillance, demonstrating the ability to track individuals using Wi-Fi radio signals and exploit security vulnerabilities.", 'The potential of AI to track and monitor individuals via Wi-Fi radio signals and cameras, even in complete darkness and through walls, raising concerns about privacy and surveillance.', "The demonstration of AI's capability to exploit security vulnerabilities by generating code to exploit a Wi-Fi router, highlighting the potential risks of AI-powered cyber attacks.", 'The evolving nature of AI and the need for updated regulations and ethical considerations to address privacy and surveillance issues.', 'The discussion about the responsibility and potential ethical implications of inventing new technologies, emphasizing the need for new laws and regulations to protect privacy and ethical considerations.']}], 'duration': 235.598, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ1172895.jpg', 'highlights': ['AI using fMRI to reconstruct images from brain activity, potentially leading to decoding of dreams and inner monologue.', 'Ability to reconstruct visual cortex activity in reverse, indicating the potential to decode dreams in the next few years.', "Experiment demonstrating the AI's attempt to reconstruct inner monologue based on video stimuli.", "AI's impact on privacy and surveillance, demonstrating the ability to track individuals using Wi-Fi radio signals and exploit security vulnerabilities.", 'The potential of AI to track and monitor individuals via Wi-Fi radio signals and cameras, even in complete darkness and through walls, raising concerns about privacy and surveillance.', "The demonstration of AI's capability to exploit security vulnerabilities by generating code to exploit a Wi-Fi router, highlighting the potential risks of AI-powered cyber attacks.", 'The evolving nature of AI and the need for updated regulations and ethical considerations to address privacy and surveillance issues.', 'The discussion about the responsibility and potential ethical implications of inventing new technologies, emphasizing the need for new laws and regulations to protect privacy and ethical considerations.']}, {'end': 1861.029, 'segs': [{'end': 1519.86, 'src': 'embed', 'start': 1492.994, 'weight': 0, 'content': [{'end': 1498.079, 'text': "It's as if all these locks that are locking all the doors in our society, we just unlocked all those locks.", 'start': 1492.994, 'duration': 5.085}, {'end': 1505.572, 'text': "And people know about deepfakes and synthetic media, but what they didn't know is that it's now just three seconds of audio of your voice before.", 'start': 1500.25, 'duration': 5.322}, {'end': 1506.833, 'text': 'now I can synthesize the rest.', 'start': 1505.572, 'duration': 1.261}, {'end': 1508.934, 'text': "And that's going to go, again, that's going to get better and better.", 'start': 1506.873, 'duration': 2.061}, {'end': 1513.817, 'text': "So try not to think about, am I scared about this example yet? You might feel like, I'm not actually scared of that example.", 'start': 1509.254, 'duration': 4.563}, {'end': 1517.199, 'text': "It's going to keep going at an exponential curve.", 'start': 1514.597, 'duration': 2.602}, {'end': 1519.86, 'text': "So that's part of it is we don't want to solve what the problem was.", 'start': 1517.239, 'duration': 2.621}], 'summary': "Advancements in synthetic media can generate audio from just three seconds of a person's voice, with the technology expected to improve exponentially.", 'duration': 26.866, 'max_score': 1492.994, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ1492994.jpg'}, {'end': 1588.268, 'src': 'heatmap', 'start': 1524.122, 'weight': 1, 'content': [{'end': 1528.645, 'text': 'And with exponential curves, we now need to skate way further than where you might think you need to.', 'start': 1524.122, 'duration': 4.523}, {'end': 1534.924, 'text': 'But just to name it explicitly, this is the year that all content-based verification breaks.', 'start': 1529.039, 'duration': 5.885}, {'end': 1536.345, 'text': 'It just does not work.', 'start': 1535.304, 'duration': 1.041}, {'end': 1540.828, 'text': "And none of our institutions are yet able to, like they haven't thought about it.", 'start': 1536.765, 'duration': 4.063}, {'end': 1542.169, 'text': "They're not able to stand up to it.", 'start': 1540.868, 'duration': 1.301}, {'end': 1544.551, 'text': 'So we tried this example, state ID.', 'start': 1542.57, 'duration': 1.981}, {'end': 1546.653, 'text': 'Generate me lots of state IDs.', 'start': 1544.951, 'duration': 1.702}, {'end': 1552.938, 'text': "Okay I don't know if you guys have seen the latest TikTok filters.", 'start': 1548.674, 'duration': 4.264}, {'end': 1554.059, 'text': "They're wild.", 'start': 1553.498, 'duration': 0.561}, {'end': 1558.142, 'text': "I can't believe this is a filter.", 'start': 1557.021, 'duration': 1.121}, {'end': 1562.388, 'text': 'The fact that this is what filters have evolved into is actually crazy to me.', 'start': 1558.507, 'duration': 3.881}, {'end': 1567.95, 'text': 'I grew up with the dog filter on Snapchat and now this filter gave me lip fillers.', 'start': 1562.728, 'duration': 5.222}, {'end': 1570.131, 'text': 'This is what I look like in real life.', 'start': 1568.77, 'duration': 1.361}, {'end': 1580.061, 'text': 'Are you kidding me? Yeah, just seeing someone, all content-based verification breaks this year.', 'start': 1572.352, 'duration': 7.709}, {'end': 1585.145, 'text': "You do not know who you're talking to, whether via audio or via video.", 'start': 1580.161, 'duration': 4.984}, {'end': 1588.268, 'text': 'And you know, do you want to give this example of China?', 'start': 1585.906, 'duration': 2.362}], 'summary': 'Content-based verification is failing, institutions are unprepared, exponential growth in fake content.', 'duration': 64.146, 'max_score': 1524.122, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ1524122.jpg'}, {'end': 1598.998, 'src': 'embed', 'start': 1572.352, 'weight': 2, 'content': [{'end': 1580.061, 'text': 'Are you kidding me? Yeah, just seeing someone, all content-based verification breaks this year.', 'start': 1572.352, 'duration': 7.709}, {'end': 1585.145, 'text': "You do not know who you're talking to, whether via audio or via video.", 'start': 1580.161, 'duration': 4.984}, {'end': 1588.268, 'text': 'And you know, do you want to give this example of China?', 'start': 1585.906, 'duration': 2.362}, {'end': 1593.272, 'text': "Sure, since I've been on this kick about trying to say why TikTok is such a dangerous thing for national security,", 'start': 1588.308, 'duration': 4.964}, {'end': 1598.998, 'text': "You may all be aware that the Biden administration there's been this whole negotiation.", 'start': 1594.593, 'duration': 4.405}], 'summary': 'Challenges with content-based verification; concerns over tiktok and national security; negotiations by biden administration.', 'duration': 26.646, 'max_score': 1572.352, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ1572352.jpg'}, {'end': 1786.958, 'src': 'embed', 'start': 1745.351, 'weight': 3, 'content': [{'end': 1753.656, 'text': "although maybe it'll be humans as figureheads, but it'll be whoever has the greater compute power will win.", 'start': 1745.351, 'duration': 8.305}, {'end': 1763.342, 'text': "And you could argue that we've sort of already had that starting in 2012, 2016, the campaigns are starting to use A-B testing to test their messages.", 'start': 1754.056, 'duration': 9.286}, {'end': 1770.466, 'text': "But the difference now is that not just you're testing some different messages, but the AI is fundamentally writing messages,", 'start': 1764.002, 'duration': 6.464}, {'end': 1778.496, 'text': "creating synthetic media, AB testing it, AZ testing it across the entire population, creating bots that aren't just like bots posting on Twitter,", 'start': 1770.466, 'duration': 8.03}, {'end': 1784.317, 'text': 'but instead are building long-term relationships over the next six years to slowly persuade you in some direction.', 'start': 1778.496, 'duration': 5.821}, {'end': 1786.958, 'text': 'Loneliness becomes the largest national security threat.', 'start': 1784.877, 'duration': 2.081}], 'summary': "Ai's growing influence in politics, using synthetic media and bots to persuade and create long-term relationships, poses a significant threat to national security.", 'duration': 41.607, 'max_score': 1745.351, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ1745351.jpg'}, {'end': 1834.184, 'src': 'embed', 'start': 1805.568, 'weight': 5, 'content': [{'end': 1810.595, 'text': "But if I pump even more electricity through the system, it doesn't pop out some other emergent intelligence,", 'start': 1805.568, 'duration': 5.027}, {'end': 1812.738, 'text': "some capacity that wasn't even there before, right?", 'start': 1810.595, 'duration': 2.143}, {'end': 1816.152, 'text': "And so a lot of the metaphors that we're using again paradigmatically.", 'start': 1813.81, 'duration': 2.342}, {'end': 1822.256, 'text': "you have to understand what's different about this new class of Gollum generative large language model AIs.", 'start': 1816.152, 'duration': 6.104}, {'end': 1824.678, 'text': 'This is one of the really surprising things.', 'start': 1822.796, 'duration': 1.882}, {'end': 1830.242, 'text': 'talking to the experts, because they will say these models have capabilities.', 'start': 1824.678, 'duration': 5.564}, {'end': 1834.184, 'text': 'we do not understand how they show up, when they show up or why they show up.', 'start': 1830.242, 'duration': 3.942}], 'summary': 'Surprising capabilities of gollum generative language model ais emerge unpredictably.', 'duration': 28.616, 'max_score': 1805.568, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ1805568.jpg'}], 'start': 1408.693, 'title': "Ai's impact on society", 'summary': "Discusses the potential dangers of deepfakes and synthetic media, such as mimicking voices and societal impact, with examples of exponential growth. it also explores ai's impact on elections, societal narratives, and reality, emphasizing power dynamics and persuasive narratives created by ai.", 'chapters': [{'end': 1633.715, 'start': 1408.693, 'title': 'Deepfake dangers and synthetic media', 'summary': 'Discusses the potential dangers of deepfakes and synthetic media, highlighting the ability to mimic voices and the potential societal impact, such as scamming and national security threats, with examples of the exponential growth of this technology and its implications for content-based verification.', 'duration': 225.022, 'highlights': ['Deepfake technology can mimic voices after just three seconds of audio, leading to potential scams and security threats. The technology allows the mimicry of voices after listening to just three seconds of audio, potentially leading to scenarios where scammers could impersonate individuals, such as a child asking for sensitive information from their parents.', 'The exponential growth of deepfake technology poses a significant challenge to content-based verification and institutional security. The exponential growth of deepfake technology presents a significant challenge to content-based verification and institutional security, with the speaker emphasizing that this year marks the breakdown of content-based verification and the inability of institutions to address this issue.', 'The potential national security threats posed by deepfake technology, particularly in the context of social media platforms like TikTok, are highlighted. The potential national security threats posed by deepfake technology are emphasized, with a scenario presented where a foreign entity could exploit deepfake technology to create chaos and incoherence within a society, particularly in the context of social media platforms like TikTok.']}, {'end': 1861.029, 'start': 1633.755, 'title': "Ai's impact on elections and reality", 'summary': 'Discusses the potential implications of ai, particularly golem ais and deepfakes, on elections, societal narratives, and the concept of reality, highlighting the potential shift in power dynamics and the unprecedented ability of ai to create persuasive narratives and synthetic media.', 'duration': 227.274, 'highlights': ["AI's potential to influence elections and societal narratives The discussion emphasizes the transformative power of AI, particularly Golem AIs and deepfakes, in potentially shaping election outcomes and societal narratives, posing a significant threat to the traditional human-driven political process.", "AI's capability to create persuasive narratives and synthetic media at a massive scale The AI's ability to create persuasive narrative and synthetic media, conduct A-B testing, and build long-term relationships to persuade individuals poses a significant societal and national security threat, potentially altering the dynamics of public opinion and decision-making.", 'Unprecedented and unpredictable capabilities of Golem AIs The unpredictability and emergence of capabilities in Golem AIs, such as their increasing parameter size without a clear understanding of how or why these capabilities manifest, highlight the unique and potentially uncontrollable nature of these AI models, distinguishing them from previous AI paradigms.']}], 'duration': 452.336, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ1408693.jpg', 'highlights': ['Deepfake technology can mimic voices after just three seconds of audio, leading to potential scams and security threats.', 'The exponential growth of deepfake technology poses a significant challenge to content-based verification and institutional security.', 'The potential national security threats posed by deepfake technology, particularly in the context of social media platforms like TikTok, are highlighted.', "AI's potential to influence elections and societal narratives poses a significant threat to the traditional human-driven political process.", "AI's capability to create persuasive narratives and synthetic media at a massive scale poses a significant societal and national security threat.", 'Unprecedented and unpredictable capabilities of Golem AIs highlight the unique and potentially uncontrollable nature of these AI models.']}, {'end': 2953.042, 'segs': [{'end': 1886.815, 'src': 'embed', 'start': 1863.3, 'weight': 0, 'content': [{'end': 1871.745, 'text': "And what you see here and I'll move into some other examples that might be a little easier to understand is that you ask these AIs to do arithmetic and they can't do them.", 'start': 1863.3, 'duration': 8.445}, {'end': 1872.826, 'text': "they can't do them and they can't do them.", 'start': 1871.745, 'duration': 1.081}, {'end': 1876.168, 'text': 'And at some point, boom, they just gain the ability to do arithmetic.', 'start': 1872.886, 'duration': 3.282}, {'end': 1878.309, 'text': "No one can actually predict when that'll happen.", 'start': 1876.248, 'duration': 2.061}, {'end': 1882.552, 'text': "Here's another example, which is you train these models on all of the internet.", 'start': 1878.69, 'duration': 3.862}, {'end': 1886.815, 'text': "So it's seen many different languages, but then you only train them to answer questions in English.", 'start': 1882.832, 'duration': 3.983}], 'summary': 'Ais struggle with arithmetic, then suddenly gain ability. models trained on internet only answer questions in english.', 'duration': 23.515, 'max_score': 1863.3, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ1863300.jpg'}, {'end': 2436.353, 'src': 'embed', 'start': 2411.081, 'weight': 1, 'content': [{'end': 2417.464, 'text': "And you'll see that at the beginning, it took 20 years for AI to get up to the level of human ability.", 'start': 2411.081, 'duration': 6.383}, {'end': 2424.467, 'text': 'And by the time we reach 2020, AI is solving these tests pretty much as fast as we can create them.', 'start': 2418.004, 'duration': 6.463}, {'end': 2427.949, 'text': 'You can imagine what happens 2021, 2022, 2023.', 'start': 2425.187, 'duration': 2.762}, {'end': 2436.353, 'text': "Even for the experts, it's getting increasingly hard because progress is accelerating.", 'start': 2427.949, 'duration': 8.404}], 'summary': 'Ai progressed from 20 years to match human ability to solving tests at the same speed as they are created, with accelerating progress.', 'duration': 25.272, 'max_score': 2411.081, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ2411081.jpg'}, {'end': 2573.475, 'src': 'embed', 'start': 2544.568, 'weight': 2, 'content': [{'end': 2556.251, 'text': "We're talking about how a race dynamic between a handful of companies of these new Gollum class AIs are being pushed into the world as fast as possible.", 'start': 2544.568, 'duration': 11.683}, {'end': 2562.753, 'text': 'Right We have Microsoft that is pushing chat GPT into its products.', 'start': 2557.151, 'duration': 5.602}, {'end': 2563.913, 'text': "We'll get into this more later.", 'start': 2562.813, 'duration': 1.1}, {'end': 2570.915, 'text': "And again, until we know how these things are safe, we haven't even solved the misalignment problem with social media.", 'start': 2564.753, 'duration': 6.162}, {'end': 2573.475, 'text': 'So in this.', 'start': 2572.614, 'duration': 0.861}], 'summary': "Companies are rapidly deploying new gollum class ais, including microsoft's gpt chat, without addressing safety and misalignment issues.", 'duration': 28.907, 'max_score': 2544.568, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ2544568.jpg'}], 'start': 1863.3, 'title': "Ai's unpredictable and exponential development", 'summary': 'Delves into the unpredictable growth and rapid progress of ai, showcasing its sudden proficiency in various tasks and its potential to surpass human expertise. it also discusses the rapid deployment of golem class ais, highlighting their potential to cause exponential harms and posing serious risks, especially to younger users.', 'chapters': [{'end': 1961.281, 'start': 1863.3, 'title': "Ai's unpredictable development", 'summary': 'Discusses the unpredictable growth of ai capabilities, such as its sudden proficiency in arithmetic, language translation, and theory of mind, reaching a level comparable to a nine-year-old, and the discovery of these capabilities after years of deployment.', 'duration': 97.981, 'highlights': ["AI's rapid development, from struggling with arithmetic to suddenly gaining the ability to perform it, with no predictable timeline for this transformation.", 'The unpredictability of AI language abilities, as models trained for English question-answering unexpectedly start to answer questions in Persian upon increasing model size.', "The progression of AI's theory of mind, evolving from no capability in 2018 to reaching a level comparable to a nine-year-old by November last year, despite this development being discovered only last month."]}, {'end': 2522.199, 'start': 1961.821, 'title': "Ai's double exponential progress", 'summary': "Discusses the rapid and exponential progress of ai, including its ability to self-train and enhance its capabilities, surpassing human expertise in various tasks, and the challenges in understanding and controlling its emergent abilities, with examples such as ai's rapid advancements in solving tests and its implications on economic and national security.", 'duration': 560.378, 'highlights': ["AI's ability to self-train and enhance its capabilities, surpassing human expertise in various tasks AI is rapidly advancing in self-training and surpassing human expertise, as seen in its ability to generate research-grade chemistry knowledge and solve tests faster than humans.", "Challenges in understanding and controlling AI's emergent abilities The chapter highlights the challenges in understanding and controlling AI's emergent abilities, including the lack of knowledge about the content in AI models and the limitations in aligning AIs in a longer-term sense.", "Implications of AI's progress on economic and national security The rapid progress of AI poses implications on economic and national security, with the need to stay updated on AI advancements and the potential for rivals to exploit important trends."]}, {'end': 2953.042, 'start': 2522.199, 'title': 'Golem ai and the race for intimacy', 'summary': "Highlights the rapid deployment of golem class ais, such as chatgpt, into society, emphasizing their potential to cause exponential harms, with specific examples including alpha persuade and the deployment of such ais in social media platforms, revealing the alarming speed at which these ais are being integrated into people's lives, as evidenced by the deployment timeline of chatgpt and its integration into platforms like windows 11 and snapchat, posing serious risks, especially to younger users.", 'duration': 430.843, 'highlights': ['The rapid deployment of Golem class AIs, such as ChatGPT, into society, emphasizing their potential to cause exponential harms. The chapter underscores the potential dangers posed by the swift integration of Golem class AIs, like ChatGPT, into society, and the exponential harms that could result from their deployment.', 'The deployment of Alpha Persuade, an AI capable of outperforming humans in persuasion, showcasing the alarming advancement of AI technology. The introduction of Alpha Persuade, an AI designed to excel in persuasion, demonstrates the significant advancements in AI technology, posing potential risks in influencing human behavior.', 'The rapid integration of AIs, including ChatGPT, into social media platforms and its potential impact on users, especially younger individuals. The swift integration of AIs, like ChatGPT, into social media platforms, such as Snapchat, raises concerns about their potential impact on users, particularly younger individuals, as evidenced by the example of Snapchat embedding ChatGPT directly into its product, potentially exposing underage users to risks.']}], 'duration': 1089.742, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ1863300.jpg', 'highlights': ["AI's rapid development, from struggling with arithmetic to suddenly gaining the ability to perform it, with no predictable timeline for this transformation.", "AI's ability to self-train and enhance its capabilities, surpassing human expertise in various tasks AI is rapidly advancing in self-training and surpassing human expertise, as seen in its ability to generate research-grade chemistry knowledge and solve tests faster than humans.", 'The rapid deployment of Golem class AIs, such as ChatGPT, into society, emphasizing their potential to cause exponential harms.']}, {'end': 3248.785, 'segs': [{'end': 3001.166, 'src': 'embed', 'start': 2978.265, 'weight': 2, 'content': [{'end': 2987.871, 'text': 'Actually in the field there is a 30 to one gap in people building and doing gain of function, research on AIs and the people who work on safety.', 'start': 2978.265, 'duration': 9.606}, {'end': 2998.964, 'text': 'let alone the fact that, well, yeah, so then, at least now at least, the people who are doing safety research and people who are working in research,', 'start': 2989.235, 'duration': 9.729}, {'end': 3001.166, 'text': "they're not driven by the for-profit incentive right?", 'start': 2998.964, 'duration': 2.202}], 'summary': '30 to one gap in ai gain of function research vs. safety research.', 'duration': 22.901, 'max_score': 2978.265, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ2978265.jpg'}, {'end': 3070.177, 'src': 'embed', 'start': 3024.622, 'weight': 0, 'content': [{'end': 3030.767, 'text': "Now again, but at least the smartest people in AI safety believe that there's a way to do it safely.", 'start': 3024.622, 'duration': 6.145}, {'end': 3033.99, 'text': 'And again back to the start of this presentation,', 'start': 3031.408, 'duration': 2.582}, {'end': 3042.917, 'text': "50% of AI researchers believe there's a 10% or greater chance that humans go extinct from our inability to control AI.", 'start': 3033.99, 'duration': 8.927}, {'end': 3052.746, 'text': 'And we already said you would not get on that plane if that was the chance that the engineers who built the plane told you was going to happen.', 'start': 3043.598, 'duration': 9.148}, {'end': 3059.952, 'text': 'And currently, the companies are in a for-profit race to onboard humanity onto that plane from every angle.', 'start': 3053.669, 'duration': 6.283}, {'end': 3070.177, 'text': 'And the pace that Satya Nadella, the CEO of Microsoft, described that he and his colleagues are moving at at deploying AI is frantic.', 'start': 3061.873, 'duration': 8.304}], 'summary': '50% of ai researchers fear 10% chance of human extinction from uncontrolled ai, while companies rush to deploy ai at a frantic pace.', 'duration': 45.555, 'max_score': 3024.622, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ3024622.jpg'}, {'end': 3230.392, 'src': 'embed', 'start': 3197.579, 'weight': 1, 'content': [{'end': 3202.341, 'text': 'if you were back then, you would have thought that the entire world would have either ended or every country would have nukes.', 'start': 3197.579, 'duration': 4.762}, {'end': 3205.862, 'text': 'We were able to create a world where nukes only exist in nine countries.', 'start': 3203.181, 'duration': 2.681}, {'end': 3208.763, 'text': 'We signed nuclear test ban treaties.', 'start': 3206.682, 'duration': 2.081}, {'end': 3212.485, 'text': "We didn't deploy nukes to everywhere and just do them above ground all the time.", 'start': 3209.644, 'duration': 2.841}, {'end': 3216.746, 'text': 'I think of this public deployment of AI as above ground testing of AI.', 'start': 3212.505, 'duration': 4.241}, {'end': 3217.727, 'text': "We don't need to do that.", 'start': 3217.026, 'duration': 0.701}, {'end': 3222.604, 'text': 'We created institutions like the United Nations and Bretton Woods to create a positive some world,', 'start': 3218.7, 'duration': 3.904}, {'end': 3225.987, 'text': "so we wouldn't war with each other and try to have security.", 'start': 3222.604, 'duration': 3.383}, {'end': 3230.392, 'text': 'that would hopefully help us avoid nuclear war if we can get through the Ukraine situation.', 'start': 3225.987, 'duration': 4.405}], 'summary': 'Efforts led to nukes in only 9 countries, avoided public ai deployment, and aimed for global security.', 'duration': 32.813, 'max_score': 3197.579, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ3197579.jpg'}], 'start': 2953.082, 'title': 'Risks of unregulated ai', 'summary': 'Highlights the unbalanced focus on ai development, for-profit race to deploy ai, potential catastrophic consequences of uncontrolled ai, and the need for global regulation to avoid detrimental impact on society and the economy.', 'chapters': [{'end': 3248.785, 'start': 2953.082, 'title': 'Risks of unregulated ai', 'summary': 'Highlights the unbalanced focus on ai development and safety research, the for-profit race to deploy ai, the potential catastrophic consequences of uncontrolled ai, and the need for global regulation to avoid a detrimental impact on society and the economy.', 'duration': 295.703, 'highlights': ['The unbalanced ratio of 30 to one between AI development and safety research, indicating a lack of focus on AI safety.', "50% of AI researchers believe there's a 10% or greater chance of human extinction due to uncontrolled AI, emphasizing the severity of the potential consequences.", 'The frantic pace at which companies are deploying AI, as described by Satya Nadella, the CEO of Microsoft, highlighting the aggressive for-profit race to onboard humanity onto AI.', 'The comparison of the public deployment of AI to above ground testing of AI, stressing the need for global regulation and control to avoid catastrophic outcomes.', 'The historical reference to the creation of institutions like the United Nations and Bretton Woods to regulate and create a positive-sum world, drawing a parallel to the need for similar regulation for AI to avoid potential global crises.']}], 'duration': 295.703, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ2953082.jpg', 'highlights': ["50% of AI researchers believe there's a 10% or greater chance of human extinction due to uncontrolled AI, emphasizing the severity of the potential consequences.", 'The comparison of the public deployment of AI to above ground testing of AI, stressing the need for global regulation and control to avoid catastrophic outcomes.', 'The unbalanced ratio of 30 to one between AI development and safety research, indicating a lack of focus on AI safety.', 'The frantic pace at which companies are deploying AI, as described by Satya Nadella, the CEO of Microsoft, highlighting the aggressive for-profit race to onboard humanity onto AI.', 'The historical reference to the creation of institutions like the United Nations and Bretton Woods to regulate and create a positive-sum world, drawing a parallel to the need for similar regulation for AI to avoid potential global crises.']}, {'end': 3707.235, 'segs': [{'end': 3277.855, 'src': 'embed', 'start': 3248.825, 'weight': 0, 'content': [{'end': 3254.128, 'text': 'How many people here are aware of the film The Day After? OK, about half of you.', 'start': 3248.825, 'duration': 5.303}, {'end': 3258.651, 'text': 'It was the largest watch made for TV film in all of human history.', 'start': 3255.068, 'duration': 3.583}, {'end': 3266.555, 'text': "It was about the prospect of nuclear war, which, again, was a kind of abstract thing that people didn't really want to think about.", 'start': 3259.751, 'duration': 6.804}, {'end': 3268.396, 'text': "And let's repress it and not talk about it.", 'start': 3266.615, 'duration': 1.781}, {'end': 3269.037, 'text': "And it's really hard.", 'start': 3268.416, 'duration': 0.621}, {'end': 3277.855, 'text': 'But they basically said, We need to get the United States and Russia and its citizen populations to see what would happen in that situation.', 'start': 3269.537, 'duration': 8.318}], 'summary': 'The day after was the largest tv film, addressing the prospect of nuclear war, seen by about half of the audience.', 'duration': 29.03, 'max_score': 3248.825, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ3248825.jpg'}, {'end': 3548.674, 'src': 'embed', 'start': 3526.944, 'weight': 3, 'content': [{'end': 3537.491, 'text': 'and what we hear from everybody that would help the most is to selectively slow down the public deployment of these large language model AIs.', 'start': 3526.944, 'duration': 10.547}, {'end': 3542.93, 'text': 'This is not about stopping the research.', 'start': 3541.069, 'duration': 1.861}, {'end': 3545.612, 'text': 'This is not about not building AI.', 'start': 3543.491, 'duration': 2.121}, {'end': 3548.674, 'text': "It's about slowing down the public deployment.", 'start': 3545.792, 'duration': 2.882}], 'summary': 'To help the most, selectively slow down public deployment of large language model ais.', 'duration': 21.73, 'max_score': 3526.944, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ3526944.jpg'}, {'end': 3615.52, 'src': 'embed', 'start': 3592.185, 'weight': 4, 'content': [{'end': 3601.031, 'text': "If we slow down public deployment of AIs, aren't we just going to lose to China? And honestly, you know, we want to be very clear.", 'start': 3592.185, 'duration': 8.846}, {'end': 3603.692, 'text': 'All of our concerns, especially on social media as well.', 'start': 3602.111, 'duration': 1.581}, {'end': 3607.455, 'text': "We this is we want to make sure we don't lose to China.", 'start': 3603.852, 'duration': 3.603}, {'end': 3615.52, 'text': 'We would actually argue that the public deployment of AIs, just like social media that were unregulated, that incohered our society,', 'start': 3607.475, 'duration': 8.045}], 'summary': 'Delaying ai deployment risks losing to china, impacting society.', 'duration': 23.335, 'max_score': 3592.185, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ3592185.jpg'}, {'end': 3707.235, 'src': 'embed', 'start': 3686.77, 'weight': 5, 'content': [{'end': 3699.473, 'text': "So Facebook released their Gollum pre-trained foundation model 13 days ago and they had a sort of perfunctory form that you'd fill out to make sure it's only going to researchers.", 'start': 3686.77, 'duration': 12.703}, {'end': 3707.235, 'text': "but they didn't do a great job of it and within days it was leaked to the internet and in particular to 4chan.", 'start': 3699.473, 'duration': 7.762}], 'summary': "Facebook's gollum pre-trained model leaked to internet within 13 days.", 'duration': 20.465, 'max_score': 3686.77, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ3686770.jpg'}], 'start': 3248.825, 'title': "The impact of 'the day after' and nuclear ai arms race", 'summary': "Discusses the impact of the made-for-tv film 'the day after', watched by 100 million americans and aired to all russians in 1987, leading to a shared understanding of consequences of nuclear war. it also emphasizes the need for a democratic dialogue on the future of ai and the urgency to slow down public deployment to ensure safety, highlighting global competition, particularly with china's advancements.", 'chapters': [{'end': 3328.231, 'start': 3248.825, 'title': 'The day after: impact of a made-for-tv film', 'summary': "Discusses the impact of the made-for-tv film 'the day after', which was watched by 100 million americans and later aired to all russians in 1987, leading to a shared understanding of the consequences of full-scale nuclear war.", 'duration': 79.406, 'highlights': ["The made-for-TV film 'The Day After' was watched by 100 million Americans, and later aired to all Russians in 1987, leading to a shared understanding of the consequences of full-scale nuclear war.", "The film 'The Day After' was the largest watch made-for-TV film in all of human history.", "The film 'The Day After' aimed to get the United States and Russia and its citizen populations to see what would happen in the event of a full-scale nuclear war."]}, {'end': 3707.235, 'start': 3328.231, 'title': 'Nuclear future and ai arms race', 'summary': "Discusses the urgency of a democratic dialogue on the potential future of ai, emphasizing the need to slow down public deployment to ensure safety and avoid an ai arms race, while also highlighting the impact on global competition, particularly with china's advancements.", 'duration': 379.004, 'highlights': ['The need for a democratic dialogue on the future of AI and the urgency of slowing down public deployment for safety. The chapter emphasizes the importance of having a democratic dialogue about the future of AI and highlights the urgency of slowing down public deployment to ensure safety, paralleling it with the nuclear age and the creation of institutions like the UN Bretton Woods.', 'The impact of unregulated AI deployment on global competition and the potential to lose to China. The discussion raises concerns about unregulated or reckless deployment of AI, citing its potential to make democratic systems incoherent and lead to losing to China, while also highlighting how slowing down public AI releases could affect Chinese advancements.', "The influence of open source models on China's AI advancements and the leakage of Facebook's Gollum pre-trained foundation model. The chapter explains how open source models have influenced China's AI advancements, providing an example of Facebook's Gollum pre-trained foundation model being leaked to the internet and 4chan, impacting global AI competition and innovation."]}], 'duration': 458.41, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ3248825.jpg', 'highlights': ["The made-for-TV film 'The Day After' was watched by 100 million Americans, and later aired to all Russians in 1987, leading to a shared understanding of the consequences of full-scale nuclear war.", "The film 'The Day After' aimed to get the United States and Russia and its citizen populations to see what would happen in the event of a full-scale nuclear war.", "The film 'The Day After' was the largest watch made-for-TV film in all of human history.", 'The need for a democratic dialogue on the future of AI and the urgency of slowing down public deployment for safety. The chapter emphasizes the importance of having a democratic dialogue about the future of AI and highlights the urgency of slowing down public deployment to ensure safety, paralleling it with the nuclear age and the creation of institutions like the UN Bretton Woods.', 'The impact of unregulated AI deployment on global competition and the potential to lose to China. The discussion raises concerns about unregulated or reckless deployment of AI, citing its potential to make democratic systems incoherent and lead to losing to China, while also highlighting how slowing down public AI releases could affect Chinese advancements.', "The influence of open source models on China's AI advancements and the leakage of Facebook's Gollum pre-trained foundation model. The chapter explains how open source models have influenced China's AI advancements, providing an example of Facebook's Gollum pre-trained foundation model being leaked to the internet and 4chan, impacting global AI competition and innovation."]}, {'end': 4046.078, 'segs': [{'end': 3766.628, 'src': 'embed', 'start': 3729.102, 'weight': 0, 'content': [{'end': 3737.386, 'text': "And then, lastly, is that the recent US export controls have also been really good at slowing down China's progress on advanced AI,", 'start': 3729.102, 'duration': 8.284}, {'end': 3740.727, 'text': "and that's a different lever to sort of keep the asymmetry going.", 'start': 3737.386, 'duration': 3.341}, {'end': 3742.688, 'text': 'You can still do your research as fast as possible.', 'start': 3740.787, 'duration': 1.901}, {'end': 3748.111, 'text': 'You can just not do as much public deployment and still maintain your lead over China.', 'start': 3743.288, 'duration': 4.823}, {'end': 3757.815, 'text': "So the question that we have been asking literally everyone that we get on the phone with who's an AI safety person or AI risk person is simply this.", 'start': 3749.831, 'duration': 7.984}, {'end': 3761.041, 'text': 'What else that should be happening.', 'start': 3759.258, 'duration': 1.783}, {'end': 3764.325, 'text': "that's not happening, needs to happen, and how do we help close that gap?", 'start': 3761.041, 'duration': 3.284}, {'end': 3766.628, 'text': "And that's we don't know the answer to that question.", 'start': 3765.026, 'duration': 1.602}], 'summary': "Us export controls slowing china's progress on advanced ai, maintaining lead over china, need to close the gap in ai safety and risk", 'duration': 37.526, 'max_score': 3729.102, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ3729102.jpg'}, {'end': 4020.58, 'src': 'embed', 'start': 3993.134, 'weight': 4, 'content': [{'end': 3998.202, 'text': "And we're here to try to help coordinate or facilitate whatever other discussions need to happen that we can help make happen.", 'start': 3993.134, 'duration': 5.068}, {'end': 4004.791, 'text': 'But what we really wanted to do was just create a shared frame of reference for some of the problems, some of the dark side.', 'start': 3999.408, 'duration': 5.383}, {'end': 4010.654, 'text': "Just to repeat what Asa said, AI will continue to also create medical discoveries we wouldn't have had.", 'start': 4005.332, 'duration': 5.322}, {'end': 4015.737, 'text': "It's going to create new things that can eat microplastics and solve problems in our society.", 'start': 4010.855, 'duration': 4.882}, {'end': 4017.258, 'text': 'It will keep doing those things.', 'start': 4016.037, 'duration': 1.221}, {'end': 4020.58, 'text': 'And we are not wanting to take away from the fact that those things will happen.', 'start': 4017.798, 'duration': 2.782}], 'summary': 'Ai will continue to create medical discoveries and solutions for societal problems.', 'duration': 27.446, 'max_score': 3993.134, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ3993134.jpg'}], 'start': 3707.892, 'title': "China's ai progress and ai responsibilities", 'summary': "Explores china's catch-up in advanced ai through decentralization, the impact of us export controls, and the need for better ai safety measures. it also emphasizes the urgency of addressing ai responsibilities and risks, advocating for collective action and coordination to avoid past mistakes and harness ai's positive potential while mitigating its dangers.", 'chapters': [{'end': 3808.93, 'start': 3707.892, 'title': "China's progress on advanced ai", 'summary': "Discusses how decentralization has allowed china to catch up in advanced ai, the impact of us export controls on china's ai progress, and the need for better ai safety measures such as kyc and liability regulations.", 'duration': 101.038, 'highlights': ["Decentralization enabling China's catch up in advanced AI The decentralization of the internet is facilitating China's progress in advanced AI, posing a significant challenge to maintaining asymmetry in AI development.", "Impact of US export controls on China's AI progress Recent US export controls have effectively slowed down China's advancements in advanced AI, contributing to maintaining the lead over China in AI development.", 'Need for better AI safety measures There is a pressing need for improved AI safety measures such as KYC (Know Your Customer) and liability regulations to ensure responsible deployment and usage of AI models, addressing potential risks and consequences.']}, {'end': 4046.078, 'start': 3809.51, 'title': 'The future of ai: responsibilities and risks', 'summary': 'Discusses the urgency of addressing ai responsibilities and risks to avoid the mistakes made with social media, emphasizing the need for collective action and coordination, while acknowledging the positive potential of ai but warning about the dangerous concerns it poses.', 'duration': 236.568, 'highlights': ['The urgency of addressing AI responsibilities and risks to avoid the mistakes made with social media Emphasizes the need for collective action and coordination', 'Acknowledging the positive potential of AI while warning about the dangerous concerns it poses Stresses the need for a negotiated solution among the players', 'The importance of creating a shared frame of reference for the problems and dark side of AI Encourages starting a conversation and coordinating discussions']}], 'duration': 338.186, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/xoVJKj8lcNQ/pics/xoVJKj8lcNQ3707892.jpg', 'highlights': ["Decentralization enabling China's catch up in advanced AI", "Impact of US export controls on China's AI progress", 'Need for better AI safety measures', 'The urgency of addressing AI responsibilities and risks to avoid the mistakes made with social media', 'Acknowledging the positive potential of AI while warning about the dangerous concerns it poses', 'The importance of creating a shared frame of reference for the problems and dark side of AI']}], 'highlights': ['The Social Dilemma reached 100 million people in 190 countries in 30 languages, showcasing extensive global impact.', "50% of AI researchers believe there's a 10% or greater chance of human extinction due to uncontrolled AI, emphasizing the severity of the potential consequences.", 'The co-founders of the Center for Humane Technology have advised heads of state, global policymakers, members of Congress, and national security leaders, demonstrating their influence and engagement at the highest levels of decision-making.', 'The user engagement with the new technology has increased from 100 people to 10 million, generating over a billion images, indicating a significant increase in user engagement.', 'AI using fMRI to reconstruct images from brain activity, potentially leading to decoding of dreams and inner monologue.', 'Deepfake technology can mimic voices after just three seconds of audio, leading to potential scams and security threats.', "AI's rapid development, from struggling with arithmetic to suddenly gaining the ability to perform it, with no predictable timeline for this transformation.", "50% of AI researchers believe there's a 10% or greater chance of human extinction due to uncontrolled AI, emphasizing the severity of the potential consequences.", "The made-for-TV film 'The Day After' was watched by 100 million Americans, and later aired to all Russians in 1987, leading to a shared understanding of the consequences of full-scale nuclear war.", "Decentralization enabling China's catch up in advanced AI", 'The urgency of addressing AI responsibilities and risks to avoid the mistakes made with social media', 'The importance of creating a shared frame of reference for the problems and dark side of AI']}