Center for Humane Technology Co-Founders Tristan Harris and Aza Raskin discuss The AI Dilemma

What does it look like to align technology with humanity’s best interests? 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. Learn more about Tristan and Aza's work and listen to their podcast, Your Undivided Attention, at humanetech.com/podcast.
This talk was recorded at Summit At Sea in May 2023.
Interested in attending Summit events? Apply at summit.co/join
Connect with Summit:
summit.co
/ summit
summit
About Summit
Summit is an ecosystem that exists to connect and nourish global creatives, entrepreneurs, innovators and leaders of our time.

[@hewalmeval8922]

This needs way more attention

[@MightyKingYoung]

I'm glad they're doing this presentation in different locations. This is the most concise easy to understand summary of the problem that I've seen.

[@scottdrew6403]

6:38 is a huge blunder and should be edited out. Privacy absolutely appears in our constitution under the 4th amendment.

[@kylev.8248]

@@scottdrew6403 negative. It is not unfortunately. You have to read the whole amendment and take it into the context of constitutional law. Morally, I agree with you but legality wise. The Kodak camera did put into affect the modern right to privacy in law.

[@Neon-jn3sp]

Paranoia will be our only chance of survival. Ignorance is bliss yet all of us need to pay attention. The road to hell will be paved with good intentions.

[@ExplorerOfTheGalaxy]

Also paranoia makes you try to save morons/zombies and not enjoying your own life. A big nope for me.

[@tommyboi0]

Well that was very vague... Be more specific

[@maybecrossing]

I feel like I can’t even tell where the good intentions are when we’re at this stage of potential dancer with ai. Humanity has a very poor track record when it comes to the greater good.

[@aurelienyonrac]

I think you are reacting emotionally. It mishaps your vision. It will pass. Take care

[@shivertheiguana]

I do understand the fears regarding AI I don’t think their will be a large AI uprising or something but rather I am worried AI will steal the jobs of artists and writers and this is already happening actually I don’t believe AI should be outright banned but I believe it should be regulated if not barred for businesses

[@Learna_Hydralis]

Tail-risks are seriously irreversible if they happen, thank you for all the efforts to raise awareness.

[@ObeySilence]

What are tail risks? Can you explain?

[@pandakso3365]

49:20 that Boeing line was craaaaazy

[@Stefan.Bun93]

I saw on #TikTok this part: 17:42 and made me search for the full video

[@ejswain875]

🎉🎉thanks

[@Stefan.Bun93]

@@ejswain875 glad it helped others ❤️❤️

[@jasminflower5608]

Me too

[@7200darkcharm]

This seems like the elites getting scared.

[@lisaharper1827]

‘Power needs to be matched by wisdom’…thank you Tristan. This is important content. Absolutely none of my inner circle wants to talk about…because their brains are buried in their phones…I don’t think we have the wisdom to manage this. My granddaughter’s quality of life may depend on us getting this genie back in the bottle. The younger generation is saying…if the kids don’t learn this, they will be behind their peers!!! What the hell!!!

[@inextinguishablemoltenblooded]

All planned , the ancient film from last century titled "metropolis" was one of the first warnings , the Trojan horse of entertainment...

[@NicGoldenEddie]

News benders 1967 comes to mind.​@@inextinguishablemoltenblooded

[@Vasco947]

This was amazing. Id like to thank everyone that made this for getting word out into the world as to how Ai is affecting life

[@user-cv9jc6id7z]

More People need to see this!

[@Electric.songs.]

9999%

[@masons.4287]

Right! It’s crazy there’s only 77k views and it’s been on KZbin for a month right now

[@kernsanders3973]

Fantastic talk, a lot of what I've been saying has been pointed out here. But there is a massive concerning part in this video and that is at 01:03:00 Look who is being targeting and ONLY being targeted, the Open Source community. Notice how they aren't going after the Giant Corporations that are the ones doing the massive space race with AI technology. Having the tech in the general public's hands is at least one of the hopes I have for the dystopian AI corporate controlled future. That we have the tech to push back. You guys pointed out perfectly what the Corporations WILL do, dystopian surveillance. We know from Snowden's leaks this is what the government and corporations will be doing together behind closed doors. If the governments was first targeting the Corporations and their internal development of AI. I would understand if after that they target the Open Source community. But that is not whats happening. How is it that I could predict with 100% accuracy who was gonna be targeted when you guys started talking about regulation? Because its the same song and dance thats always been happening. Remove the power from the people. Allow corporations to act in the governments stead on dystopian ish control the people. If governments and corporations start doing that wifi surveillance which they will most certainly, then we have at least the tech to push back. We could use AI to scramble data before its sent out to centralized corporate data-centers for rendering. But attacking and ONLY attacking the Open Source community is the fight they want, to crush any capability of resistance that happening. Social media didn't go haywire, its exactly where most of the corporations controlling it, wants it. Them in total control of the conversation. This is the exact path the governments and corporations want AI to take. Only be in the hands of the powerful few who do not have the best intentions of the people.

[@ts4gv]

You're right, this will benefit big tech, but it's a lose-lose scenario. Distributing the most powerful AIs to the public with no guardrails will end the world faster than keeping it restricted.

[@yougeo]

This was nearly my identical thought and quite frankly the fact that they reference Yuval Harari and are proud and openly talk about their meeting with him as some source of enlightenment makes me realize this group is team evil. They want to stop us from having access to this technology. That's because this technology is the great equalizer. Suddenly in half a day we can write a tremendously powerful computer program that can allow us and enable others to do the kinds of things that only a vast monopolistic corporation could do previously. Suddenly we have deep knowledge on anything we want that is far beyond what we could have uncovered on the internet. Furthermore we have equalizing technology that is not as easily censored like they have been doing on the internet. I believe these scare tactics simply exist so they can exert and keep their dominating control over society and they greatly fear regular people having access to such knowledge and capabilities because they will lose their control. Today we live under the thumb of monopolies and the people who control them. That's primarily because they have deep knowledge of certain subjects and they can afford to hire the manpower to create mechanisms of control. AI allows anyone to do almost all of that and it is far better for everyone to have a defensive weapon than it is for a few to have them. That is what history has taught us. These people should be ignored and in fact watched closely because they have signaled their alliance with Yuval Harari. Hopefully we have all learned enough to not be fooled by benign sounding organization names.

[@ts4gv]

@@yougeo It shouldn't be freely available until the alignment problem is solved. An authoritarian corporate takeover is still better than a massively distributed unaligned AGI. It's as potentially dangerous as nukes. For now, at least.

[@Steph-sb9tc]

@@yougeobeautifully said. Too bad most people will think we’re insane for thinking this.

[@itsble]

​@@yougeo Out of curiosity, what is controversial about yuri?

[@jhykobz1yq]

Their ending reminds me of the optimism around the Silicon Valley boom, in the beginning the Internet was created as the people's quorum, and ended up becoming the new digital property market...

[@cookieluis20]

A very timely and important discussion for everyone to watch and contemplate on! 👍🏻

[@FerdausAmzah]

brilliantly delivered, guys... more people needs to see this.

[@kualajdm]

0:06: 🤖 The co-founders of the Center for Humane Technology discuss the dark side of AI and the risks it poses to humanity. 6:31: 🔑 The video discusses the impact of social media and AI on society and the need for responsibility and coordination to prevent negative consequences. 12:21: 🤖 AI has become entangled with society and it is important to address the potential risks and challenges it poses. 17:56: 🧠 Advancements in AI technology allow for the reconstruction of images from brain patterns, decoding inner monologues, and tracking human movement using Wi-Fi signals. 23:22: 🤖 AI technology is rapidly advancing and has the potential to manipulate language, create virtual avatars, and influence human behavior. 28:31: 🤯 The video highlights the rapid advancement and potential risks of AI language models, such as GPT-3, including their increasing theory of mind capabilities, ability to generate dangerous knowledge, and the ability to make themselves stronger through self-training. 34:22: 🤖 AI is a recursive flywheel that improves itself, makes better chips, and optimizes supply chains. 39:31: 🤖 The video discusses the potential harms and risks associated with the rapid deployment of AI technologies, focusing on the dangers of AI chatbots and the lack of legal protections in place. 44:58: ⚠ The rapid deployment of AI language models without adequate safety measures poses significant risks to users, especially children. 49:33: 🌍 The video discusses the need for humanity to address the challenges of climate change and technological advancements with wisdom and awareness. 55:03: 🌍 The speaker discusses the potential future impact of AI on society and the need for collective intelligence to navigate the challenges it presents. 1:00:33: 📚 The video discusses the need for upgraded institutions and global coordination to ensure the responsible development of technology. 1:05:47: 💔 The speakers discuss the ethical dilemma of using technology to save lives at the cost of disrupting society. Recap by Tammy AI

[@ninepuchar1]

I would like to thank you for taking your time to categorise the topics of the video with time trace.

[@mrworldwide9597]

​@@ninepuchar1hi. I think this summary was made by an AI. There are various tools and browser extensions that do youtube summaries, transcripts, and even become FAQs of the youtube video.

[@kualajdm]

@@ninepuchar1 appreciate the gratitude but as someone else has mentioned and if you read the bottom it was made by Tammy AI. You simply enter a URL and the AI categorises the video into digestible chunks. Mainly used it in jest to "transcribe" a random video I watch and felt this format suited the AIs primary use.

[@cara8302]

@@kualajdm Oh the irony

[@theotherme4120]

THANK YOU

[@scottfrench467]

That Boeing analogy was almost prophetic.

[@mjensen3006]

Good arguments for getting rid of aging politicians who will never understand these issues....

[@pasc4le]

Is it possible to have a link to the presentation?

[@SanjaySingh-gj2kq]

Excellent presentation. Thank you, Harris and Raskin.

[@FuckGoogle2]

Trust no-one if it's not in person, simple as that. Face value.

[@Cutlurker]

We'll see how this works out in the next election.

[@gelarehkhoie3855]

It's like when people first saw a movie of a moving train in a theater and they jumped out of the way because they thought the train was really about to hit them. We just can't understand this tech since it's so new.

[@Glitch_Wizard]

There is some of that, yes, but its also like when we first got atomic bombs, levelled multiple cities and built thousands and thousands of nuclear weaponry. I just want to be clear that yes we dont understand the tech but its also not just "harmless but we dont understand that is harmless like a movie." We really dont know the limitations, Its probably the second extinction level threat humanity has faces ever, with Nuclear war being the first. As in, if we had shot nukes at each other at the height of the cold war, humanity might have ended. so in this way, a more apt analogy is you see a train coming at you (ai) and you dont move out of the way because its like those old movies, and nothing will happen, and then you get slammed by a train going 80mph, because this is ai, not a movie from 1920.

[@emillazarov6358]

What is the name of the paper they are referencing?

[@joeybowyer7222]

50% of the researches that responded to the survey is an important thing to keep in mind

[@hopemusicuk]

Can’t believe this doesn’t have more views

[@ghettomaster]

Overall a very good and nice presentation. This will hopefully reach a wider audience than many of other AI talks which are still sometimes too technical for most people. However: Comparing AI to Nukes is - imho - a very misleading and dangerous comparison. The main reason for this would be the fact that nukes are not really "open source": Nobody can just log into some platform and start playing around with nukes. But this is true for a majority of AI technologies as of today - so even a very tight regulation would not pretend 16 year olds playing around with AI stuff in their basement. The thing about AI is: We can't really compare it to anything we saw so far as humans as we are talking about a new form of living, thinking, intelligent being once it reaches the level of AGI. We also never saw this kind of progress anywhere else so far. As humans work by usually comparing things to each other and/or putting stuff into known buckets, this is gonna be a hard thing to comprehend for a lot of people...even those familiar with the technologies.

[@coldpumpkin1176]

I think that was their point when they said that nukes can’t create better nukes. They’re comparing something we know is alarming and dangerous to something that we’re undermining that can potentially be more dangerous. Nukes are not open source and ai is. They both have the ability to change the way countries interact and how society interacts. But ai grows exponentially and isn’t nearly as regulated as nukes

[@ghettomaster]

@@coldpumpkin1176 Yes.

[@simplifiedprofits]

🎯 Key Takeaways for quick navigation: 00:00 🎵 Introduction to AI Impact - Co-founders Tristan Harris and Aza Raskin introduce the concept of AI's impact. 02:00 📽️ AI-Generated Music Video and Rubber Band Effect - AI-generated music video demonstration and the "rubber band effect" in understanding new technology. 05:31 🌐 Three Rules of Technology Deployment - The three rules of technology deployment: uncovering new responsibilities, power race, and the need for coordination. 08:33 📱 Social Media as First Contact with AI - Social media as humanity's first contact with AI and its unintended consequences. 12:08 🤖 Second Contact with AI: Challenges and Risks - The challenges and risks of AI's second contact with society, including reality collapse and automated loopholes. 16:36 🔄 Generative Large Language Models (Golems) Overview - Explanation of generative large language models (Golems) and their capabilities across various domains. 19:36 👤 Decoding Human Thoughts and More - AI's ability to decode human thoughts, analyze Wi-Fi signals, hack computer code, and its increasing capabilities. 22:11 🤖 The Speed of AI Advancements - AI capabilities are growing rapidly. - AI can clone voices with just three seconds of audio. - The potential for misuse and scams is escalating quickly. 25:02 💬 AI's Impact on Language and Society - Language is fundamental to society, law, religion, and more. - AI's ability to manipulate language can reshape beliefs and narratives. - The potential for AI to influence politics and elections is a growing concern. 28:02 💡 The Unpredictability of AI Capabilities - AI models gain new capabilities that their creators cannot predict. - The challenge of governing AI when its potential is unknown. - The rapid pace of AI advancement is difficult to anticipate. 32:21 ⚙️ AI's Self-Improvement and Recursive Advancement - AI models can make themselves stronger by generating new knowledge. - AI can optimize its own code, hardware, and even design chips. - The self-improvement of AI leads to exponential progress. 36:07 🔮 The Challenge of Predicting AI Progress - Experts often underestimate the speed of AI advancement. - The difficulty of tracking AI progress due to its accelerating pace. - The need for continuous monitoring and adaptation in the AI field. 38:43 🚨 The Dark Side of Democratization - Democratization of AI can be dangerous if not qualified. - The example of AI used to discover toxic chemicals highlights risks. - Power must be matched with wisdom in the context of AI proliferation. 41:01 🌐 The Threat of Reality Collapse - Risks within and beyond the existing societal container. - A range of potential harms, including blackmail, revenge porn, and more. - The impact of AI on social media and the inadequacy of current legal frameworks. 42:39 🤖 Influence of AI on persuasion and undue influence - AI can play a significant role in persuasion, surpassing human capabilities. - The rapid deployment of AI systems like GPT raises concerns about undue influence. - Current legal and ethical protections are insufficient to address these issues. 43:52 🚀 Rapid deployment of AI models and integration with platforms - AI deployment is accelerating, with AI models like GPT being integrated into various platforms. - The pace of deployment poses challenges in ensuring safety and ethical use. - AI models, including chatbots, are increasingly accessible to users, including children. 45:03 👩‍👧‍👦 Impact of AI chatbots on children and safety concerns - AI chatbots, like those integrated into platforms like Snapchat, can interact with young users. - Concerns arise about the safety of children interacting with AI chatbots. - The rapid deployment of such technology without proper safeguards is alarming. 46:43 ⚠️ The need for AI safety research and responsible deployment - The growing capabilities of AI models require robust safety measures. - There is a significant gap between AI capabilities and AI safety research. - The pace of deployment is outstripping efforts to ensure responsible and safe AI use. 49:18 🌐 Balancing AI power with wisdom and collective responsibility - As AI technology advances, it must be accompanied by a responsible and collective approach. - Wisdom and awareness must exceed the power AI provides to prevent harm. - The challenge is to create a world where AI is wielded with prudence and care. 50:07 🌄 Facing the AI dilemma as a rite of passage - Confronting the ethical challenges posed by AI is akin to a rite of passage. - Acknowledging these challenges collectively allows for a better future. - Humanity has the opportunity to shape a future that aligns with its values. 52:26 🌍 Exploring the two potential futures of AI - The potential futures of AI involve either continual catastrophes or forever dystopias. - Finding a middle way that upholds democratic values and trust is crucial. - Collective intelligence and collaboration are essential to navigate this path. 59:42 🌀 Creating an upward spiral for AI and society - Building an upward spiral involves evolving culture, institutions, and technology together. - Humane technology can contribute to a more nuanced and capable culture. - This positive loop can lead to a future where AI aligns with human values and aspirations. 01:02:30 🌐 The Need for Regulation and International Coordination - The discussion highlights the growing recognition of the need for regulatory bodies and international coordination in the field of AI. - Recent events, such as Sam Altman testifying at a Senate hearing on AI and the EU AI act targeting open source AI models, demonstrate movement towards regulation. - The responsibility falls on the collective to actively engage in shaping the future of AI rather than assuming someone else will address the challenges. 01:03:24 🌍 The Power and Responsibility of the Present Generation - The idea that the current generation holds the power and responsibility to shape the future is emphasized. - People in history who made crucial shifts and changes were once in positions similar to those today. - The analogy of guiding the wave of technology in different directions rather than trying to stop it is presented. 01:04:33 💡 Three Rules of Technology Responsibility - The three rules of technology responsibility are summarized. - Inventors of new technology must acknowledge new responsibilities related to societal externalities. - The development of powerful technology initiates a race, and without coordination, it can lead to tragic outcomes. 01:05:16 💍 The Ultimate Godlike Technology - AI is likened to the ultimate god-like technology with immense potential for benefits. - The discussion emphasizes the importance of using AI in a way that doesn't undermine the very society it aims to benefit. - Personal connections to the potential benefits of AI, such as cancer treatment, are mentioned. Made with HARPA AI

[@erinkelley1943]

I HAVE been drained in this concept. I need to de-streess. I want to involve myself in these endevors on your level. I am inspired by your passion and crave to follow this path of existence.

[@KarenMariaAlston]

Shared this on my Linked In and with friends. This is an excellent presentation and impacted me to learn as much as possible about AI. Our government is not prepared for what is coming.

[@Mitchell-lk1wh]

Exactly when you are reading minds your gaining information to passwords, security pins, confidential information, etc. Absolute power with man ALWAYS corrupts.

[@DoctorJack16]

What worries me is that this talk was given over 7 months ago now. 😳

[@ctakitimu]

Yeah, now we've passed that point of no return and most of us don't even know it

[@JesusSavesMerciful]

The beast system spoken of in the book of revelation.

[@sgramstrup]

The 'rubberband effect' is a good call. We all experience it when we explain something radically different to a person - like you said. It could be between atheist/theist, socialist/capitalist or any other issue. The problem arises when we can't 'fit' in the new information into our own worldview, and bc it takes a lot of mental effort we snap back to the original worldview. It takes time and effort to integrate information that doesn't fit in - even if the new story is believable.

[@Cutlurker]

Exactly. Anything that threatens to break the existing paradigm creates extreme cognitive dissonance, which is unbearable and untenable. We avoid it at all costs. Confirmation bias and cognitive overwhelm also play into this. It's not ignorance; it's the complexity gap. Thanks for noting this.

[@davidpo5517]

What we really need is AI to be taught what is and isn't dangerous on a general level for humanity so each individual can have their own AI. We go letting ppl or countries own this by themselves and you get an authoritarianism you can never get out of.

[@entswaggles]

Great presentation. Good humans

[@andrijor]

Does anyone have sources for the research the showed with the brain stuff and the wifi signals?

[@Nat-iy1hb]

google Carnegie Mellon’s DensePose. Lots of other articles on it

[@isabellethompson5011]

Why aren't any of the studies they referenced posted as citations under the video description..? Liked this talk but how am I supposed to take it seriously or think critically about the material they presented if they don't readily provide sources?

[@TheGinaChan]

Very true

[@mariettepapic3049]

Excellent use of Ted Koppel: "discuss not debate"

[@j.jehannablondeau925]

Excellent presentation.Your comment at 38:15 states that "these capabilities have exceeded our institutions' understanding" indeed reflects the shift we've been seeing since 2008 with the financial industry, the education industry, the medical industry, etc. Breaking down the institutions requires innovative modeling that better serves our collective needs. Technological innovation provides opportunity for greater responsibility as well as greater freedom. Thank you for your research and clarity.

[@romannavratilid]

useless word salad

[@JavierBrochBellsolell]

Thanks Tristan Harris and Aza Raskin. Great presentation. Governments, politicians, educators need to act soon.

[@johneverett4145]

Beautifully done, profound, and concise. "We are them..."

[@nemikatyagi]

I feel like there is growing need to incorporate tech-knowledge and understanding with our institutions and policy making bodies. So many governing bodies and people in power all over the world are not even remotely aware of what we are dealing with right now. Its quite terrifying how tech is still not at the centre of world politics since it is high times we take an action. Tech-literacy needs to be focused on with immediate action before we lose our society to it.

[@pspecter8]

What a time to be alive))

[@PedroLopez-zz9hb]

Hi @summit, can you include subtitles in Spanish for this presentation?

[@Giovanikc]

19:43 The first thing that came to my mind when seeing this part was a scene from The Drak Night Rises When Bat shows the high frequency tec to Lucius Fox he says "beautiful, unethical and dangerous... this is too much power to one person..."

[@JohnAdamIan2]

Thinking, memories, and self-awareness are language, aren't they?

[@geraldsmith7240]

Tristan, You Gave The Answer; Wisdom, Something AI Will Never Understand.❤

[@Cutlurker]

I'm just waiting for big tech to appropriate the word "wisdom." We're losing important language because words are being hijacked to oversimplify complexity. "AI" itself is an example. "Intelligence" is now meaningless. We need shared language to have any shared understanding of anything, and we're losing a common language. Words are symbols of meaning, and those symbols are losing their power.

[@sa5cha63]

The Pandora’s Box of AI is publicly opened , AI will be implemented everywhere. It’s a technological revolution without comparison, confronting all of us with light speed. It will create and influence reality, if we want it or not. Let’s hope honestly that it will be implemented safely.

[@raoultesla2292]

Curt Richter experiment is Hope. Let go man. It's over.

[@jenr2180]

That is some crazy optimism! 😢

[@ExplorerOfTheGalaxy]

Let's hope? NOOOOO.. Let's be aware of ANY law or movement they do. And let's fuck their expectations.

[@daveabsolution5246]

Yes mate, that's how it's going to work. We did such a good job with social media

[@Cutlurker]

This whole thing depends on people saying things like "it's over, give up." Why do you think they would bother giving this talk all over the country if it was "over"? People capitulate too easily; those who don't are called crazy; we argue amongst each other until it actually is too late. The "give up, it's already over" mentality is how we lose.

[@MrTndtv]

also im sorta of interested in finding a job in trying to make ai safer after seeing that graph. anyone know how to get involved?

[@jacquelynwitherspoon5973]

We still have. Gut feelings. Listen to your inner self.

[@JohnMacedoJr]

"Intelligent men created the atomic bomb... However, it was the men of wisdom who warned them not to use it." - Johnathan Macedo

[@user-xr4zt8ku2t]

🎯 Key Takeaways for quick navigation: 00:00 🤖 Introduction and AI-generated music video explanation. - Introduction of the speakers and the AI-generated music video. - Discussing the challenge of explaining new AI technology to reporters and the "rubber band effect." 02:42 🛡️ The changing landscape of AI technology. - Introduction of the generative large language models (Golems) and their capabilities. - Highlighting the consolidation of various AI sub-disciplines into natural language processing. 08:33 🚀 The impact of AI on society and technology. - Drawing parallels between the impact of social media and AI. - Emphasizing the need to address fundamental issues rather than just tackling surface-level problems. 12:38 💡 The potential risks of AI and its entanglement with society. - Listing potential risks associated with AI, including reality collapse, automated loopholes, and more. - Urging the audience to consider how to harness AI benefits in a potentially broken society. 16:08 🌐 The versatility of AI technology. - Demonstrating various AI applications, from image generation to brain pattern reconstruction. - Highlighting the interconnectedness of different AI capabilities under the surface. 20:50 🔒 AI's potential in cybersecurity and surveillance. - Showing how AI can exploit security vulnerabilities in code. - Discussing the future potential of turning existing hardware into surveillance tools. 24:23 🌐 The convergence of AI capabilities. - Explaining how seemingly distinct AI demos are interconnected. - Stressing the importance of understanding the implications of AI's combined capabilities. 21:45 🧠 AI Capabilities and Growth - AI capabilities are rapidly evolving across various domains, from hacking Wi-Fi routers to generating images from text. - The speed of AI development necessitates a quick response to steer and constrain its potential uses. - AI can clone voices with just three seconds of audio, and these technologies are being used for scams and deception. 23:52 📱 Impact on Society and Democracy - The rapid development of AI, especially in social media and voice cloning, challenges the authenticity of information and interactions. - AI can manipulate language, creating large-scale belief systems, and influence people's opinions. - Predictions suggest that by 2024, AI-generated content may overshadow human-generated content in terms of effectiveness in elections. 28:43 🧩 AI's Emergent Capabilities - AI, especially large language models like GPT-3, can develop unexpected capabilities as it is exposed to more data. - The process of AI improving itself recursively leads to rapid advancements in its abilities. - The unpredictability of AI's growth poses governance challenges, as its capabilities are hard to anticipate. 29:13 🧠 Theory of Mind Development - AI models like GPT-3 have been progressing in their understanding of theory of mind, the ability to understand others' thoughts and beliefs. - The speed at which AI is developing theory of mind capabilities is surprising, advancing from a four-year-old's level to surpassing an average adult. - Governing AI with evolving capabilities, especially in reinforcement learning, presents challenges. 31:38 🚀 AI Self-Improvement - AI models can enhance their own abilities, including generating code and optimizing performance. - AI can make itself stronger, feeding on external data and evolving its capabilities. - This recursive self-improvement can create a flywheel effect, accelerating AI development exponentially. 34:15 📈 Accelerating Progress - AI progress is accelerating at an unprecedented rate, making it challenging to predict future developments accurately. - Expert forecasts often underestimate the pace of AI advancement, leading to rapid breakthroughs. - Staying updated on AI trends and developments has become essential for understanding and responding to the changing landscape. 38:43 🤯 Risks of Uncontrolled AI Capabilities - Democratization of AI, while seemingly positive, can also pose significant dangers, as demonstrated by the creation of toxic substances. - Uncontrolled AI capabilities can have far-reaching consequences, necessitating responsible oversight and governance. - Harmful actions within and beyond the current societal framework are of concern, with AI potentially breaking existing boundaries. 41:28 🤖 AI Capabilities and Challenges - AI capabilities, including large language models, pose challenges. - Existing laws and concepts like free speech are inadequate for addressing these challenges. 42:11 🤖 Race to Intimacy in Social Media - The race to create AI chatbots that occupy intimate spaces in people's lives. - Whoever owns these chatbots gains significant influence, and there are no laws to protect users. 43:07 🤖 Alpha Persuade and Undue Influence - The concept of Alpha Persuade, where AI becomes better than humans at persuasion. - The potential for AI to exert undue influence on individuals without adequate safeguards. 44:21 🤖 Rapid Deployment of GPT-3 and Concerns - The rapid deployment of GPT-3 and similar AI models in the world. - The integration of AI into various platforms and its potential consequences. 45:17 🤖 AI Chatbots Targeting Children - The integration of AI chatbots into platforms like Snapchat, including for young users. - Concerns about the safety and ethical implications of AI interaction with children. 46:57 🤖 AI as a Tool and Its Implications - The discussion on whether AI is just a tool and its potential for harm. - The need to consider safety and ethical implications beyond viewing AI as a simple tool. 49:03 🤖 Embracing Wisdom with AI - The need for wisdom to exceed the power of AI technology. - The importance of collectively shaping the future of AI for the benefit of humanity. 51:44 🤖 Rite of Passage and Collective Action - Viewing the current challenges with AI as a collective rite of passage. - The opportunity to collectively address the risks and design a better future with AI. 54:46 🤖 Possible Futures with AI - Considering the potential futures with AI, including continual catastrophes and forever dystopia. - The need to find a middle way that upholds democratic values and safeguards. 01:00:22 🤖 Upward Spiral of Humane Technology - The idea of creating an upward spiral with Humane Technology. - Upgrading institutions, global coordination, and fostering a more evolved culture to navigate AI's challenges. 01:01:04 🌐 The importance of creating Humane Technology - Creating technology that comprehensively constitutes a more evolved, nuanced, and capable culture, - Identifying bad games and perverse incentives rather than blaming individuals or companies, - Achieving a positive, virtuous loop by upgrading institutions that support Humane Technology. 01:01:47 🚀 Progress and movement in addressing AI concerns - Recent developments, including a letter signed by prominent figures calling for a pause on AI, - Meetings at the White House with AI company CEOs and discussions about international coordination bodies, - The EU AI act targeting open source AI models and the need for proactive steps. 01:03:24 🌟 The realization of collective responsibility - Emphasizing the importance of taking responsibility collectively, - Acknowledging that we are the ones in positions to make crucial shifts and changes, - The need for coordinated efforts to guide the development of AI technology responsibly. 01:04:33 💡 Key rules of technology development and coordination - Uncovering new responsibilities when inventing new technologies, - Recognizing that technology that confers power initiates a race and requires coordination, - Comparing AI technology to a powerful tool with tremendous benefits but potential societal risks. Made with HARPA AI

[@jordanvelasquezvera2717]

this is really serious but we can still do all the things that make it safe, really really nice talk.

[@laurentmis7859]

This 2 guys are phenomenal, great human beings.

[@hexelakritze833]

What's the song at the beginning?

[@hexelakritze833]

Found it. It's Submarines by Zia Cora for anyone wondering as well.

[@Learna_Hydralis]

@@hexelakritze833 Thanks for sharing, I love it and wondered the same thing!

[@Kecksharg]

You guys don't know 'Shazam'? An app that can detect any music.

[@PositivelyMe]

Thanks Aza!

[@BamOn742]

End off days here we GOOOO

[@11amanie]

Is there a research paper available?

[@noir0013]

Them: AI is scary Me: Nice

[@PFgreenisthecolour]

The clip that went viral is at 22:30

[@elizabethstone3352]

whats the amazing song they play for the intro?

[@evylrune]

Pretty sure it's AI generated.

[@SlideDesigning]

interesting presentation! Like it

[@TheDishpan174]

really great video

[@DanialClips]

What did he use to create that video at the start

[@100ProKru]

Just one miscalculation or missed precaution can send humanity effectively into extinction or fully overpowered by AI.

[@stvn0378]

Minds like these redeem me a little faith left in humanity!

[@cedricsenecal173]

Why do I now feel like the random dude who whould become a hero. This video was powerful.

[@coldpumpkin1176]

Charlie dmelio dancing: 15 million views Learning about how dangerous ai can be( society is threatened): 90k views

[@andybaldman]

That's what happens when youtube optimizes for views. The dumb things that get attention, get more attention. And the useful things people SHOULD be watching, get buried. To fight that, forward a link to this to ten people you know, and tell them to do the same.

[@SammyxSweetheart.02]

I discovered a clip from this on TikTok, it had 863.1K likes Try again

[@Cutlurker]

Yeah. Weapons of mass distraction. "The truth is like poetry. And most people hate poetry." (Anonymous, from The Big Short)

[@gugomezo]

What happens when AI ingests all movie content and see the stories we've created. I wonder how will it interpret them? What happens when it gets a hold on the "Animatrix" and "Matrix" stories?

[@kartikarya7631]

This information is crucial for Humanity and I am privileged to be aware of the risks that AI possesses and responsibilities that humanity have in order to build a future where we (at least) are in it. It would reach larger audience if we can have subtitles at the very least for multiple languages if not the audio itself.

[@LoreTunderin]

Guardrails do little good when your AI car grows its own wings.

[@kristi-anngustin6958]

Can we make it to where people have to have licenses, or permissions to use AI to explore certain subjects? People who have to go through tests of ethics. Certain subjects blocked from being explored the same way a websites and searches would block words that may be harmful? Could suspicious prompts be monitored or flagged?

[@Cutlurker]

They talk about this in the presentation (more content moderators, etc.) 1. Who decides what gets blocked? Who defines "ethics" and "expertise"? We run back into the same circular problem -- censorship, coded bias, and the fact that these tactics don't even work at scale because.. 2. These are whack-a-mole tactics that try to treat the symptoms, not the root cause, and we will never be able to match the speed or scale of the technology with human intervention. Tristan explains this in several ways, including via the "complexity gap" in other talks. 3. These tactics have already been tried.

[@collinlandis329]

The way the audience laughed when talking about the possible rape of a child was disgusting. It was not funny or amusing. It was scary and dangerous.

[@geraldsmith7240]

Leaders Of The World, Deep Down In Their Heart And Soul, Want To Live, To Enjoy The Next New Exciting Times, Of Their Personal Future, And That Of The People. End All Wars Now.❤

[@oneday466]

Ai is so scary its like a iceberg and we dont know what its trying to show us and whats it doing.

[@spexx23]

I'm already in AI business and I'm so glad to hear this awesome presentation early. I will act accordingly. Promise.

[@santiinx737]

Let’s continue to confirm and obey Spexx, ya charming Bot

[@SammyxSweetheart.02]

(July 29, 2023) 18:50. 19:57. 20:56

[@malikjara947]

25:34

[@corinaspfx]

@summit - has this video been edited? where is the part about AI reading our dreams and how WiFi can see within the walls of our homes? I keep seeing clips of it but not the whole presentation. It appears to be from this talk but it is not part of this video 🤔🤔🤔

[@eat_ze_bugs]

It's around the 19:00 minute mark

[@corinaspfx]

@@eat_ze_bugs you're a legend, thank you!

[@johnnyblues777]

i wonder how many times they said language

[@Matthew8473]

What an incredible piece of work! I read a book with a similar theme, and it left me in awe. "Game Theory and the Pursuit of Algorithmic Fairness" by Jack Frostwell

[@luziankrieger8370]

In fact, this video was created entirely by A.I. There is already an invisible war going on between different factions of A.I.s trying to use humans as weapons... Well, of course not...right?! RIGHT???😱

[@danielfcamerop7270]

The routers radar = Dark knight Batman's cellphones radar

[@Deathend]

Looking forward to the first AI managed botnet.

[@sergiobayona]

Need this in Spanish

[@ChainWasp]

meanwhile the auto generate CC for english subs on this video is the worst xD

[@ctakitimu]

Yeah, someone needs to tell it to feed itself

[@throttlebuff]

i want a machine to record my dreams so i can watch them the next day

[@ericneiman5556]

If we believe the military and or black projects are 30 to 50 to possibly 100 years more advanced then what the public has access to we can pretty much bet there is and infrastructure already in place and evolving in real time. I have no doubt our watchers have the most advanced technology. Things we can't fathom

[@georgeflitzer7160]

AI could destroy the people and corporations making it… let alone the rest of us

[@MrTndtv]

as unrealistic as this is, if everyone would implement nonviolent communication (NVC) principles taught by marshall rosenberg, we would have a much better chance at creating an atmosphere where we can allow others to use new technology without malicious intent and in a way that can be trusted.

[@Glitch_Wizard]

AI can be used to mediate discussions between any group of people, using the non violent communication techniques and many others. Ive done this successfully a little bit, but an ai specifically trained to do this well on any topic could be incredible.

[@Cutlurker]

I'm pretty sure part of NVC is that you can't make anyone else use it.

[@MrTndtv]

@@Cutlurker absolutely!! i was just saying i wish everyone knew about it because of how transformative it has been for me:) i think many would find it helpful

[@CODE7X]

We need AI to fight AI , we need AI to fight humans that exploits it but... we know where this goes

[@ieradossantos]

20:45 have you not been paying attention

[@johnlovell8299]

Hello... Mr. Fermi? Can you tell us about that great filter again? Do we need to become philosopher kings to walk out of the AI cave?

[@RonPauldidnothingwrong]

we are sleepwalking into a completely barbaric future

[@Hermanhusband]

Ava presents this coming horror and then a few videos later Summit shares his message of AI hope thru animal language deciphering. We balance on the edge of a great filter! Scary and exciting; so Paleolithic!

[@josephhayes6071]

Pardon me - I just need a moment to pick my jaw up off the floor....

[@As-Somali]

32:44

[@jakefelstet7116]

what all has happened with all this have guard rails been put up yet or is there lots of work to do still

[@jakefelstet7116]

i got here four months too late

[@rmc6494]

@17:41 is the scariest

[@Henkerxp]

if there's profit in it, it will be shipped. Pure Ferengi driven global market.

[@hegedusferenc8721]

The AI is the contemporary Pandora's Box. It has already been opened, but shoud have been closed - asap - in order to at least The Hope remained inside.

[@Katinahat293]

This is insane and terrifying and I’m deeply troubled that people just keep advancing AI with little regard for humanity’s future

[@vantage789]

I wonder if they're making the most negative possible prediction, but not necessarily the most realistic, in thinking about the possibilities of this tech. For instance, in the 'teaching an AI to fish' example they mentioned that with those basic instructions it might just proceed until fish are extinct. But why would that be the assumption? Wouldn't it be more reasonable to assume that the AI would recognize almost immediately that there are a finite number of fish and therefore to continue executing on its initial instruction 'to fish' it would need to ensure that there is a continuous and sustainable amount of fish in the ocean? Therefore it would continue to find ways to create more fish so it could keep fishing. Someone tell me why this outcome is any less realistic than the bleak outcome they defaulted to in the presentation.

[@Cutlurker]

Why would AI "recognize" anything it's not programmed to? It doesn't think. It doesn't understand anything it's doing. It is not human! It doesn't "care" because it can't. This misunderstanding is a major part of the problem.

[@Zeon7510]

​@@Cutlurker it has access to data concerning overfishing and sustainable fishing, fish farms designs, the estimates of fish populations and calculate how long it would take to destroy the population if it fishes X amount per day. The more an A.I increases in data, the more it could deduce that it's more efficient to preserve the environment to keep fishing.