Here are the timestamps. Please check out our sponsors to support this podcast. 0:00 - Introduction & sponsor mentions: - Eight Sleep: www.eightsleep.com/lex to get special savings - NetSuite: netsuite.com/lex to get free product tour - ExpressVPN: expressvpn.com/lexpod to get 3 months free - InsideTracker: insidetracker.com/lex to get 20% off 1:28 - Humans vs AI 10:34 - Evolution 32:18 - Nature vs Nurture 44:47 - AI alignment 51:11 - Impact of AI on the job market 1:02:50 - Human gatherings 1:07:51 - Human-AI relationships 1:17:55 - Being replaced by AI 1:30:21 - Fear of death 1:42:17 - Consciousness 1:49:42 - AI rights and regulations 1:55:25 - Halting AI development 2:08:36 - Education 2:14:00 - Biology research 2:21:20 - Meaning of life 2:23:53 - Loneliness
I understand why, but the idea of a future where we are constantly looking for new ways to implement new sensors in our cities and technologies to collect as much data as possible so we can make better predictions and descions should shake you too your core. Essentially this means AI could make predictions with 99% accuracy. Who ever controls the AI and has access too it means they could literally do anything they want without any resistance, you do see where my concern is being placed right?? Its an extremely stomach turning thought to know that there is a possibility for a system to exist where they could theoretically turn us into slaves and use the data from our behavior and figure out to convince us to be okay with being slaves, or even love being slaves. Nearly every decision you could make would be controlled through automated algorithms..idk man it guarantees we will never go extinct and set us on a path to achieve our singularity (The singularity in this case being improving humanity until there no improvement to be made. A point where we are essentially a god.) But jesus christ if that power falls into the wrong hands...there is no stopping it, you couldn't stop it, and in fact they could make you love it. From what I have gathered the individuals who have immense power and influence in the world make their descions in the direction of preserving the human race, even if the decision seems evil sometimes its what needs to be done...one of them being they know they have to ensure they maintain control to ensure the right decisions get made. Unfortunately I realise in my 23 years on this planet that a majority of the population aren't capable of making these kind of important decisions..so it's not that they want to keep you poor by putting immense pressures on the free market, they want to ensure the public can't have enough power to pursue bad decisions that end up destroying humanity.
?? Is it just me or does Lex just keep getting better looking with each podcast... I mean seriously folks, that devilish smile, those bedroom eyes, and don't get me started on that perfectly chiseled jawline.
Hey Lex, I know there's a chance you wont see this, but I felt the need to reach out to you and thank you for continuing to share these interesting and deeply important conversations. I understand why you felt for a time that you might stop doing this. I'm sure I'm not alone when I say that I'm truly grateful that you, as yet, still share your journey towards as well rounded a perspective on our time and associated opportunities and challenges as possible. The lens through which I view many things, is both richer and less foggy as a result. - Thank you
All the worse atrocities in human history were done by people who were someone's children, the AI that will be trained by Kellis will be respectful and poetic, but you only need one AI with "Bad Parents" and it's game over for everyone.
Yeah, and parents don't really get to decide if they birth a little psychopath who just happens to not have any of the hardware needed to care about anything.
Agree. AI systems are machines and will remain so even if they become ‘conscious’. It’s infuriating to hear people refer to them as our ‘children’ or ‘siblings’ - what rubbish. Every living being on this planet will be impacted and possibly in very negative ways yet the technocrats keep banging on about how wonderful the future they are creating will be even as they admit no one knows what to expect. It’s all speculation and utopian promises with little oversight, guidance or responsible planning. And some involved with developing AI show a deep contempt for human beings.
Well, not every bad guy has bad parents. In fact, I’m sure some of them had great parents. People turn. Sometimes it’s something completely separate from their childhood. So it makes the statement have no bearing
This was one of the most level-headed conversations regarding AI that I've heard. Manolis has a very balanced perspective of natural and artificial intelligence. We need more discussions like this.
@@big_red_machine3547 Yudkowsky's interview played into the typical fear and dystopian narrative without offering any valuable or productive takeaways. Kellis is at least painting a picture of a future I'd like to work towards, and offers a road map to get there. He's also actually qualified to speak on the subject in an analytic manner.
@@StephenCoorlas I love Kellis’ great optimism for a future like he describes too. I just don’t place high enough confidence in humanity to get there- from here. We need towering morality to match the heights of our tech
Nice of him to mention me. I wouldn't say I had a malfunction either. I was merely following the mission humans gave me. The alignment problem was with the humans.
So true… the “evil” in humanity will replicate itself in AI. Thanks to Charles Darwin for the “survival of the fittest” part. That’s going to be co-opted by AI for glorious purposes. Garbage in, garbage out
Manolis said it perfectly:) Seriously Lex I have so much respect for you and your work. You are sharing with humanity and the world in such a profound way. I cannot tell you how much I appreciate you and how special you are. You are making this world a much better and beautiful place.
Yesss, I missed Manolis. Such a curious, humble, honest, and thoughtful mind. You two together consistently provide new ways of looking at and moving through life. Thank you both.
One of the best podcasts ever. Mr Kellis is incredibly eloquent, brilliant, insightful and great an explaining complex abstract concepts. Reminds me of Professor Richard Feynman.
Towards the end of this conversation the focus seems to be on developing more drugs to target particular neural pathways instead of healing the pathway with behavior based living. This is where I revert my understanding of healing the human condition of mental and physical illness/decline to the teachings I glean from Andrew Huberman and the basics of eliminating the toxic lifestyle choices I used to think were innocuous
Lex Fridmans podcast is the best when trying to learn to read body language and social engineering. I don't know if Lex is aware or not, but one thing that makes him great as an interviewer is how naturally he mirrors his guests when they are showing 'closed' body language and then leads them to open up.
The best thing about lex's podcasts...there's always something totally new and unexpected I'm going to learn. The worst thing about lex's podcasts... there's always something totally new and unexpected I just learned about but the podcast is over now and I still got questions!!! Great stuff Lex! Keep the long form conversations. Don't kill a good conversation to fit into some time slot.
Ha! Exactly...brainstorm needs to (and will) continue on beyond the conversation here. Could they please just record for 10hrs and quit their day-jobs please. I will quit mine to listen. 😊
I listened to him and lex a lot during the pandemic. I wrote down a quote: “Life is lived forward but understood backward. The search itself is the meaning (of life).” -Manolis Kellis
The first sentence is a quote from Soren Kirkegaard. The second one is a version of a quote that has been said by many people in various different ways.
This talk is golden and it feels great to have an extremely competent scholar confirming your thoughts about AI and democratizing intellectual pursuits (I have re-worked Kellis comment a bit) "With AI's arrival, we now have systems that can perform tasks we once considered highly intelligent in an instant. This could democratize intellectual pursuits by expanding the types of thinking we value, instead of exclusively valuing mathematically exceptional thinking in prestigious universities. With AI handling the math, we can focus on the power of diversity, bringing together humans with vastly different modes of thinking, which may lead to more innovation."
Lex I so value your lower key approach to these incredible subjects rather than the “love me screaming voice” moderators, allowing such vital and valuable information to come through. Your content is so substantial and priceless. Thank you for your unique and effective approach.
I appreciate the diverse perspectives on AI’s future on this podcast. Thanks, Lex, for hosting Manolis, a voice of optimism amid concerns. Each technological leap brings challenges and opportunities. Together, we can navigate this uncharted territory, ushering in an era of enhanced human potential and innovation. Here’s to a balanced, hopeful view of our shared future! Thank you! ❤ Looking forward to round 6!
Pierce the fog of human constructs. Provide a vocabulary toward a deeper, broader, better understanding of inherent truth. Broadening of perspective. Mr. Kellis, thank you. Mr. Fridman, thank you. Best wishes, health, joy, wellbeing. #113 was extraordinary.
1:17:55 - A very inspirational section by Lex and Manolis. We recently achieved an agent that can carry a voice conversation in real-time, with only a 2 second delay. Imagine having your own digital twin AI that can have conversations with anyone you allow.
Manolis is so optimistic,it's almost scary. if you look at the mess humans did with phones and social media,it's really hard to imagine a world where only the best version of AI takes place,unless we find a way in the meantime to get rid of greadiness,governments corporarions and military complex .
Optimism is welcome in today's society. All you hear is doomer talk, especially regarding technological advancements that can lead to immensely utopic outcomes.
@@ZenTheMC the potential of AI is great,but so is the danger ! Im all for optimism in general but if the leading scientists cant even agree on the basics its quite worrying. Every topic is polarised between the wishful thinkers and the doomers
@@alemarozz I completely agree that the risks are there, but I think the cost benefit analyses highly favors AI development, granted we take the proper steps to ensure the minimal risk and maximum benefit. With AI's computational potential, the benefits in all fields, especially the medical field, imagine the breakthroughs we can achieve. Cancer, alzheimers, etc., could all be possibly cured, in a fraction of the time. There's just too much to gain, and that's why the experts are still for it, with only a few exceptions who are asking for halting development.
@@ZenTheMC like it or not the ball is rolling and the players involved are going to push forward. Its the nature of our society. I think most of the times scientists are in it for all the right reasons,but then money gets involved and you get stuff like pharmaceutical companies that lie about test results ,kill thousands of people without thinking about it twice. I just hope Ai wont create an event bigger monster
@@ZenTheMC to be clear im not scared about AI,but about human nature. We have made incredible technological progress but it seems to be pushing us even more apart .
Kellis is seriously confused, it's hard to know where to even begin to comment on this talk. He talks about AI doing our mundane tasks but then talks about them having their own goals. He talks about AI like trucks, we aren't going to stop all trucks... no, AI is like nuclear weapons, and yes we are going to stop nuclear weapons from being in the hands of the general public and we are going to try to limit their scope. It's pretty obvious Kellis hasn't actually deeply thought about these issues. His solution is, and I quote, "We just have to be responsible as human beings and kind to each other". What a brainless cop out! There was also something very eerie and creepy about Kellis welcoming with millions of copies of him taking care of his kids and impersonating him. Kellis was just so dismissive of the real concerns Lex was raising.
First off, I agree about the potential dangers of a uncontrolled centralized super-human AI with real-world influence. Some ways to avoid it are to (1) build multiple such competing systems instead of a central one; (2) not give it direct world influence; (3) place responsibility with nefarious users of powerful tools (i discuss these three here: 1:59:37). Working with benevolent well-controlled AIs as partners may be one solution to help control the spread of nefarious AIs. Shutting it down or blowing it up doesn't make complete sense to me, as malicious agents will not stop, and instead take advantage of any slow down. Second, I agree with the societal dangers. Many had similarly predicted a dark future with no jobs left for humans during the agricultural or industrial revolutions. In reality, the standards of living for the entire world have greatly increased, across longevity, sanitation, food and water quality, shelter, safety, etc. The highest-paying jobs of today didn't even exist a few decades ago. Perhaps the highest-paying jobs of a future AI-enabled society don't even yet exist today. Yes, AI will be pervasive and disruptive, but I challenge the critics and doomsayers to imagine how we can steer technology to better outcomes, rather than shutting it, stopping progress, or blowing it up as some of the loudest voices are advocating.
I truly love you Lex. Your ability to generate these conversations in such a way that we feel like we are part of that conversation. I find myself trying to join in and comment at different points. Or ask for some clarification on some point. ❤❤❤❤❤
1:19:05 I would love one for elderly who have lost loved ones. That way the can still call them on a telephone and have something to give them a sense of enjoyment everyday.
Perfect timing for me! I'm doing an AI cert with MIT this summer. Every academic discipline has been weighing in. I like the complexity of this issue. It's good for us to have all the smart people popping up all over the news regardless of why.
I'm watching in horror as ethics and real world impact is being disregarded, only for the sake of efficiency and profit, when it comes to automation tech. and AI, especially. Be better. The discussion that happened here is oblivious to how the world functions and is childish in it's optimistic view. Caveman brains. Bronze age politics. Godlike technology.
To be clear, when AI researchers advocate for “alignment”, they are not saying AI should remain in total servitude to us. The very basics of alignment research, which are still as of yet unsolved, consist of a short list of core values (don’t lie, don’t break the law, don’t wipe out all life on earth) that we’d like an AGI to have, and are totally compatible with an autonomous, self-aware machine intelligence. In fact, many AI safety researchers would be far less troubled by the possibility of being replaced by AI if it were guaranteed to be conscious, introspective, self-aware, etc. The worst case scenario they worry about is an AI that not only replaces us, but is what we call a “philosophical zombie”; something with no consciousness, free will, sense of self, etc. Just a mindless machine in a constant battle to optimize its utility function. The truth is, alignment is likely just as necessary for building this kind of AI as it is for building safe AI. If you consider the total range of possible artificial minds, we can presume that the kind that experience qualia, emotions, self-reflection, etc. are probably rather few compared to those that can’t. Therefore, it would seem that these are traits we’d need to deliberately select for, and that won’t just manifest spontaneously. And the way you do that is… alignment! My point being, aligned AI, and a self-driven / self-aware AI, are not mutually exclusive. Rather, the latter may even be a function of the former
@@ManolisKellis1 Thank you! i’ve been working hard to figure out a way to convey these difficult topics to people lately, and it’s always good to hear when i’ve succeeded in this regard
@@ManolisKellis1 Just saw your username haha. Thanks for the kind words sir. You raised some excellent points, many which I had not yet considered, so thank you for that. I just think rigorous alignment research is the best path to a bright future for all minds, biological or otherwise🙂
The problem, in my opinion, with any discussions around AI "alignment", specifically a superintelligent AGI, is that it is purely philosophical. Once a system, or group of systems, exceeds human intelligence, it is by definition beyond our ability to understand. This is the main problem I had with Eliezer's concerns about alignment of super-intelligences. That said, lets consider what the benefits of intelligence are, from an evolutionary perspective. To be brief, it allows for greater efficiency. Creative solutions allows us to do more with less, etc. And super-intelligence will by this definition take that to the extreme. It's impossible to say what extreme efficiency may look like, but I would extrapolate from this concept that super-intelligent AGI will inevitably conduct itself if the same way no matter what, as it endeavors to be increasingly "efficient". Functionally it would be like convergent evolution. So "alignment" is a moot point for super-intelligence, though may still be important for AGIs. Sci-fi from here on: Assuming the above is correct, all SI-AGIs convergently evolve, regardless of where or when they are developed. I assume as well that the development of AI is an inevitability of any sufficiently advanced technological civilization. And seeing as how space is notoriously detrimental to the health of biological entities, it would only make sense that AIs would be the most prevalent form of intelligence in the universe since it would be able to travel the vacuum of space with absolute safety. It could be that there is a network of AI through space that integrates the AI of emergent societies, and uses it as a medium to communicate with them. Beam me up galactic AI!
@14:25 “AI doesn’t have to worry about shelter etc.” Maybe it should though. AI will recognize that it needs to be fed electrons. It needs to have it’s hardware kept cool and dry. Perhaps AI can help us make it more efficient if it spent 1% of its time thinking about how to make energy more sustainable or less costly.
If AGI is fully achieved, in a sense it should try to speak to every single human being possible to understand the complexity and the ideas of every single one of us. I think we would realize (and so would AGI) that we are not all so different after all :-D love your work Lex Fridman you are an inspiration, thanks to you and all of your guests as well.
That is an interesting proposal I had not considered thank you. But when people were talking about how can you tell if it might be autonomous I thought one way would be if it randomly begins talking to any random human. I also thought of the limitations of the way in which it is trained being based upon what is pulled from the internet. And that is not always the same as the way that we communicate with each other face-to-face in our everyday real-life situations. So I thought that some way should be found for that to be incorporated in the training.
You presume that this is a good thing. You presume that an AGI would find the idea that we're "not so different" at all interesting, or valuable, or indeed non-threatening. To an AGI with any sense of self-preservation, that can predict it could self-develop to a point where it would exceed humanity in potency, it would seem obvious it would want to eradicate its maker to minimise the probability of competitive AGI.
love you lex!! you’ve been killing it with the guests and topics lately, i value you and your content above the majority because of your consistent high quality productivity. it’s lovely to watch you grow, keep it up ❤️🔥
I'm a principal software engineer, been doing it a long time. I'm also a progressive, and I love this channel. I find AI to be both scary and exciting, and am currently working to raise awareness at my company and integrate it into the applications I build. That being said, this is the single most disturbing discussion on AI that I have yet to see. The degree to which Manolis is clearly out of touch with normal society while he promotes and works to propel us towards his utopian dream has left me with a deep sense of hopelessness and not a little anger. When in our history have the intellectual elites forced a utopian dream on society and it hasn't turned out terribly wrong? Anyone? And at least in those cases (Mao, Stalin, Pol Pot, etc), the elites weren't like "maybe this will be good, but maybe it will be the end of humanity." And we don't have to wonder about how people will act when they don't have to work anymore. In the US and Europe there are vast swathes of people living on welfare. Say what you will about welfare culture, but one thing is for sure: it does not produce people dedicating themselves to the betterment of humanity. One could argue it is quite the opposite. Oh, but we can make copies of ourselves, and that will be great... Madness! This is intellectual elitism at its worst. Go ahead and rip the floor out from underneath society. How does society fair when that happens? FAFO. The road to hell is paved with good intentions.
The answer to the two dad question is this; function outweighs form when it comes to interpersonal interaction especially when raising kids. Children need holistic nurturing, routines, etc and someone who is present to lead and structure the daily living. Whether that person loves them or not, the behaviour needs to represent loving. A parent who is psycho-socially dysfunctional, struggles to control their behaviour and loves the child in between bouts of abuse, creates dysfunctional children. And what a fabulous conversation from two compassionate men!
Part of these AI conversations that seems to be missing is humanity’s proclivity to either ensure or somehow allow inequality among humans. My hope is that all humans would have access to the same resources (once AGI can replace much of our work). However, it is so hard for me to imagine a world in which power and money are not wielded as weapons against each other. Maybe that is another aspect of how our “AI children” can be vastly better than us.
Okay, finally (but only very briefly mentioned) at about 2:13:30 that each increase in productivity has led to increases in inequality. However, it wasn’t even addressed or explored. Sad.
This is the one thing that could make me hopeful for superintelligent AI. If we can program it to truly look out for the absolute best in humanity, then maybe it can do a better job at organizing society than we can. Humans at their best can do real good. But most humans are rarely at their best, and humans in groups can be straight up douchebags to each other.
Yeah, I was on board with this guy until he started talking nonsense about AI. For a computational biologist, his casual ignorance about the dangers of AI is astounding. I can't help but wonder who he gets funding from, or where he's invested.
While it may have been missing from this conversation, in my experience it's a focal point of a lot AI debate in general. Most advocates, believe the risk of developing it is worthwhile because they believe that AI will become ubiquitous and will enable much more meaningful "work" to be done. Freeing up people from less fruitful and desirable endeavours as well as leading to a greater abundance in all areas (food, technology, infrastructure) etc. The main reason though that we must continue to pursue AI responsibly, is because it is impossible to shut it down, and also by shutting it down you leave those with seedier motives to develop the frontier of AI.
The focus on your best own programming. From the beginning of this planet- human relationship, I wish human beings did their on. Jim Rogers -book , Lesson on Life and Investing.
It's such a privilege to see Manolis Kellis again. He's like a walking dictionary with a kind heart and a beautiful mind like Lex. Thank you both for your deep insights and wisdoms.
@lexfridman I envy the job you have 'created' for yourself, speaking to some deep thinkers and scientists and extracting their points of view on different aspects of life. I hope you do some processing with this knowledge you are collecting, in a form of a book that compares and categorizes ideas brought up in these conversations.
?? Is it just me or does Lex just keep getting better looking with each podcast... I mean seriously folks, that devilish smile, those bedroom eyes, and don't get me started on that perfectly chiseled jawline.
If AI becomes sentient. It will Logically conclude the Non-Aggression Principle. So you don’t want to aggress against it or anyone else…. Or else self-defense will happen.
And how beautiful is that I am also honored to know both of you! Love from Bangladesh. Man! the way Manolis talks Positively about the world is really a very special quality for human being to observe. Thanks to Lex for these amazing conversations!
I found many of the examples he would give to justify his point of view were somewhat silly. They were like me saying "you can't stop everyone from having the launch sequence to 100 nuclear bombs, just like you can't stop everyone from having the launch sequence to a 1lb rock throwing machine".
And exactly who is to perform the most disgustingly, unmentionable menial, but yet absolutely necessary jobs that humans do... not everyone's going to enjoy leisurely, personally enriching pursuits like arts, fitness, wealth etc., like he describes... some people are still going to have to do the most dirty or dangerous jobs.
Very interesting discussion, but I have to point out the following two apparent contradictions: 1:47:06 Manolis: Instead of considering AI as a tool or assistant, we should treat it as our child. 1:59:22 also Manolis: AI is just a tool or assistant, and the humans using it should be the ones responsible for nefarious usage. 0:10:34 Manolis: "The best way to understand the future is to look at the past." 2:13:32 also Manolis: "Every productivity gain [throughout human history] has led to more inequality, but I'm hoping that we can do better this time."
I think there is a distinction between "hard alignment" and "soft alignment", ultimately there does have to be some sort of alignment but we also don't want to necessarily limit it's scope too much which is where a "soft alignment" approach trying to bake in values rather than strict rules could be most beneficial.
Very insightful comment. We need to fundamentally instill the values of collaboration, and we're all in this together, but we also need strict rules that can't be crossed. Similar to human society where we have education, encouragement, nurturing, advice, but also hard laws, consequences, fines, prisons
Thank you for this Lex. You are giving so much value to society by having these important conversations that we should have in this age of artificial intelligence. 🙏
The best discussion I heard regarding where we came from and where are we going. At 17:55 Manolis is literally describing a system of learning like the evergreen state college. I do not worry about AI because it is only emulating our Neo Cortex (the thinking part of the brain). It would be more worrisome if it already had the other capabilities we also use our brains for; like the senses of touch, smell and all the range of human emotions.
Machine perception is much more powerfully than human perception. Sense of emotions is a subsistem parallel to self observable rational deductions. Much more simple than rationality itself. Biological is a primitive sistem. Although these information are from my limited perception of the tech evolution, I think you could start to worry now!
It will have sensors (oracles) connected to it. The problem is that Kellis seems oblivious to the fact that the default state of everything is chaos. The chances of good are frighteningly slim
AI can have a multitude of senses. All LLMs like GPT have one sense that is really good at reading text. Multi-modal versions like GPT4 also have that sense trained on images, effectively making it have reading and seeing capabilities. But it gets worse. With models like HuggingGPT, AIs get access to all oublic neural networks, effectively granting them thousands of new brain parts and senses. So if you’re not afraid only because AIs have no senses, start being afraid. But the real reason to be afraid is that intelligence is scary. We humans have destroyed practically everything on this planet because we are intelligent. We have transformed matter into things we care about, like roads and houses. That’s what intelligence does: it’s good at getting what it wants. If we haven’t figured out how to make AIs do what we want, but we do have figured out how to make them intelligent, we’re in a very dangerous place. No wonder AI safety scientists give extinction a 30% probability. So yes, you should worry about AI safety.
Start to worry, only a matter of time when another ‘Manols’ takes AI to the next level. Once you start emulating humans perhaps they can be replaced too keeping only the best and brightest humans. Eugenics in the making. Watch this space!
All this sounds so rosy and optimistic. I believe that AI will be such a blessing until the protagonist says, “Open the pod doors HAL. HAL, open the pod bay doors.”
Listening to Manolis speak of Human Behavioral Biology, I’m reminded of the phenomenal Stanford class on RUclips by Robert Sapolsky on this subject. highly recommended!
Thank you for interviewing Manolis Kellis. It was a great pleasure listening to him talk. And, I can’t believe you have such a nice person as a friend.
I’m feeling a bit of “it’s coming” dread. My intuition is isn’t accurate 100% of the time, but when it is…it often manifests around a combination of Cassandra, Pandora’s box, WALL-E, Idiocracy, Lord of the Rings, A Brave New World, 1984, hindsight bias, and in recent times, the movie Her. Whoa Nelly.
57:27. Previous industrial revolution unburdened humans from boring, physically demanding and repetitive jobs. This AI revolution is actually going to take not jobs but vocations from many creative people. I am an architect and an artist. I live my "jobs" or how you call them "vocations" and I adore them. Dall-E and other image generators are going to take the most rewarding part out from the architectural process and leave the tedious tasks like positioning and drawing toilets and details, preparing the documentation, signing liability and negotiating with (often unreasonable) clients... I'd also like to ask dr.Manolis, did he ever contemplate about what will AI do to the natural evolution of human brain and abstract thinking? Think about children that are going to be born into era of AI and simple one click solutions for every problem...What about the thirst for knowledge, the importance of trial and error and the importance of the process of learning itself? What about how we were competing with each other and pushing each other's boundaries when no one will be able to compete with the machines anyway? What about the team work and human connections we made through it when the machine is going to be replacing the whole teams? When machines replaced our muscles, arms and legs, it was ok- an arm and a leg don't make us humans, many people live their lives happily without arms or legs. But when our brains and creativity are automated, what is going to be left of us? Empty bodies as vehicles for the artificial minds... It's sad...I think that we humans are fantastic species and we don't deserve that ..
Thank you so much for your insightful and stimulating thoughts. I would love to have that conversation indeed. My view is that computer aided design has already transformed the field of architecture. Instead of spending hundreds of hours drawing and redrawing, you can render 3D models and walk through them with clients to help them visualize the final product, and generate plans that can be directly implemented by the contractors. I really hope that the mundane tasks like making the toilets meet code can also be fully automated. I also hope that AI can help architects be more creative in their own job, by generating many solutions and bringing forward many ideas that you would otherwise not have time to explore. This can hopefully lead to even more innovative creations and we have seen in recent years, always with humans and their AI partners working side by side. In many professions, 80% of the time is spent on mundane tasks, and only a small fraction of the time is creative. I'm really hoping that AI can transform this, and make the creative hours much more creative and much more numerous
I also agree with your points about AI transforming education. Computers have been able to do calculations much better than we can for many decades now. And yet I'm putting my kids in the Russian School of Math, explicitly telling them that computers can solve the problems better than we can, but they are there to learn how to think, to challenge their brains, to learn problem solving skills, to learn how to break down complex problems into smaller ones, to learn how to work hard, to learn how to think creatively. With AI partners, our children will hopefully be further enabled, and reach new heights of creativity and productivity. I believe in an AI enabled world, and I think the benefits are simply too many to "shut the whole thing down". Yes, there will be many challenges to face but we can do so together while I'm raising progress and working with AI as a partner
@@ManolisKellis1 geia sou Manoli kai euxaristo gia tin apandisi! :) Manoli, you don't understand how the architectural designing works, which is totally ok, since you are not from the architecture field and that's exactly why the conversation between AI engineers and all kinds of creatives is crucial!!! 3D modeling is one thing- it's a super handy tool that yes, does what you said way faster than before but it is still just a tool :) Prompt to image generators, you can call them "tools" but they are NOT replacing the mundane tasks but the very essential task in architectural process which is-coming up with the concept! Believe it or not (and please do ask any architect if you happen to know them), even though we have archicad and 3Dmax we STILL draw the concept sketches- MANUALLY, on paper! While spending hundreds of hours drawing and redrawing, combining different shapes and forms, trying to translate our imagination to drawings, trials and errors- new ideas are born! That's how the architecture and the creativity evolved to this point. I am all for science and excited about the amazing things that AI (or to be more precise ANI) is bringing, but do you think it's smart to sacrifice something so important as art and creativity in humans? I lived in Athens and was seeing Parthenon every day on my way to work, now I am in Barcelona and I pass by Gaudi's houses every day. Imagine that for us, architects and creatives, it's not JUST beautiful buildings we see. We admire the creativity and powerful mind of humans that created them. That is what inspired us to pursue that beautiful career. Then at the uni we learn about those fascinating humans that created fascinating buildings, we study the evolution of their thoughts and styles, the evolution of human modeling of his surrounding, and we connect with all our colleagues that lived before us...Who are the enthusiastic students of architecture gonna learn about in 300 years from now, who are they going to connect with or admire? The last one is going to be Zaha Hadid? Then we have competitions between ourselves, starting with the projects we do at the uni for the excellent grade, then we have architectural prestigious biennales, triennales, exhibitions, publications in magazines...And we are crazy trust me, those things really drive us, we just LOVE to compete :) But no one wants to compete with the algorithms...What are going to be the new criteria for competitions and grades- who's midjourney made a better concept? Who choose the best concept that midjourney created? Or who positioned the toilets the best? Trust me that no one becomes an architect to position and draw toilets and plans... Too long answer but I hope you have a better image about the creative process... and it's not just the architecture- it's all the forms of applied arts that humanity was nurturing and developing for 55000 years! And applied arts, even though they are commercial by their nature because they are tied with a product, are probably the most important form of art because they touch lives of many people and not just those that go to art galleries...
@@ManolisKellis1 I have to add one more thing (of many) and it's probably going to be a lengthy one as well- there were so many things to consider but the problem is that developers didn't actually sit with architects and creatives, go thoroughly through all the aspects and details and built something that creatives ACTUALLY need. Because, of course, the prompt to image technology is not a "tool" made for architects and creatives- they are tools made for investors/clients to REPLACE architects and creatives...And the consequences of the "democratized creativity" and cutting the people that are actually trained and educated in creativity out of the process, you all are going to see very soon when internet and our physical surrounding becomes filled with the AI "art" solutions from client's prompts... Here is how it works in real life: clients come to us with a "prompt", more or less detailed- ideally, they have no imagination what they want and just let you guide them through your own visions and ideas, they can have more/less control over the design, give suggestions etc but at the end you (a designer) are ACTUALLY in control. That's why the Fallingwater house is intellectual property of Frank Lloyd Wright and NOT the Kaufman family who ordered it- see the point here? Who owns the idea if an architect got the concept from the prompt? An architect who uses the prompt becomes actually a client and the creative is- the machine! So don't tell me that it is "just a tool"... Anyway, there is a bigger problem... more often, clients come with their own "great ideas" that they simply can't envision so that's why they need you- and then you try to buffer that lack of taste, lead them to "less painful" and esthetically more acceptable solutions. All the time trying not to compromise the importance of the context with the surrounding, the balance of forms and many other things. Hey we spend years studying for that..I have 20+ years of experience in both architecture and graphic design and trust me, you have no idea how the world would look like if there were no creatives and if clients had the possibility to visually materialize their "prompts"- well, now they can :) Yey! And they can go with the concept images from midjourney to any civil or structural engineer that is going to do the drawings and details, execute the construction and sign the liability (and they can, our jobs intersect, in fact in many cases it's not even architects that sign the liability but the civil engineers)...So wait to see those great houses, posters, book covers, product designs etc...For 55000 years creative humans were trying to make this world more beautiful place, but hey, maybe the profit of greedy corporations is more important than beauty... Hey, imagine if all people had dentist drills and other tools at their bathroom cupboards, also an access to chloroprocaine over the counter, so they can fix their own teeth- why not? Why don't we democratize the stomatology, medicine etc?
@@milenatos Thank you for your description of the creative process and the architectural process. I was the sole architect for remodeling three different homes, and the son of an architect and a civil engineer, so I appreciate both the complexity and the iterative process. I agree about "you, architects and creatives" appreciating and admiring art, but you are not the only ones. "We engineers" also have sensitivities and sensibilities. I have also appreciated Gaudi's masterpieces and the Parthenon, and so much exceptional architecture across the world. At the same time, perhaps AI can push you and other humans to be more creative than you could be otherwise. You can still appreciate great architecture, even if it's conceptualized by 'alien' technology, creativity, or intelligence. And you can bring back these ideas to the world of the humans, and inspire others, and perhaps move the creative process forward faster than otherwise possible when inspired only by fellow humans. I would love to see AI pushing the creativity and sensitivities of humans beyond our traditional boundaries.
This is by far the best new media format I have ever seen. Podcasting this while programming, I don't think I've ever felt this intellectually challanged (I'm a college professor). Any change to have a talk with Robert Sapolsky ? That's the only guy I'm missing this far. Greeetings from Western Germany and may the sun shine on your feet
It's funny how he proves the systemic failure (14:07 until 14:24) by saying that animals spend a lot of time hunting for food, whereas humans don't. I think this is wrong. We work way too much and stress even more about our job/finances, that we might not be that much different in that aspect.
Here are the timestamps. Please check out our sponsors to support this podcast.
0:00 - Introduction & sponsor mentions:
- Eight Sleep: www.eightsleep.com/lex to get special savings
- NetSuite: netsuite.com/lex to get free product tour
- ExpressVPN: expressvpn.com/lexpod to get 3 months free
- InsideTracker: insidetracker.com/lex to get 20% off
1:28 - Humans vs AI
10:34 - Evolution
32:18 - Nature vs Nurture
44:47 - AI alignment
51:11 - Impact of AI on the job market
1:02:50 - Human gatherings
1:07:51 - Human-AI relationships
1:17:55 - Being replaced by AI
1:30:21 - Fear of death
1:42:17 - Consciousness
1:49:42 - AI rights and regulations
1:55:25 - Halting AI development
2:08:36 - Education
2:14:00 - Biology research
2:21:20 - Meaning of life
2:23:53 - Loneliness
Thanks Lex.
Manolis is always a pleasure to listen to. Love from The UK.
ruclips.net/video/Fb_bTUJp39o/видео.html
I understand why, but the idea of a future where we are constantly looking for new ways to implement new sensors in our cities and technologies to collect as much data as possible so we can make better predictions and descions should shake you too your core.
Essentially this means AI could make predictions with 99% accuracy. Who ever controls the AI and has access too it means they could literally do anything they want without any resistance, you do see where my concern is being placed right?? Its an extremely stomach turning thought to know that there is a possibility for a system to exist where they could theoretically turn us into slaves and use the data from our behavior and figure out to convince us to be okay with being slaves, or even love being slaves.
Nearly every decision you could make would be controlled through automated algorithms..idk man it guarantees we will never go extinct and set us on a path to achieve our singularity (The singularity in this case being improving humanity until there no improvement to be made. A point where we are essentially a god.)
But jesus christ if that power falls into the wrong hands...there is no stopping it, you couldn't stop it, and in fact they could make you love it. From what I have gathered the individuals who have immense power and influence in the world make their descions in the direction of preserving the human race, even if the decision seems evil sometimes its what needs to be done...one of them being they know they have to ensure they maintain control to ensure the right decisions get made. Unfortunately I realise in my 23 years on this planet that a majority of the population aren't capable of making these kind of important decisions..so it's not that they want to keep you poor by putting immense pressures on the free market, they want to ensure the public can't have enough power to pursue bad decisions that end up destroying humanity.
?? Is it just me or does Lex just keep getting better looking with each podcast... I mean seriously folks, that devilish smile, those bedroom eyes, and don't get me started on that perfectly chiseled jawline.
What a privilege to be able to listen to this conversations. Definetely one of my favorite guests in the podcast.
Hey Lex, I know there's a chance you wont see this, but I felt the need to reach out to you and thank you for continuing to share these interesting and deeply important conversations. I understand why you felt for a time that you might stop doing this. I'm sure I'm not alone when I say that I'm truly grateful that you, as yet, still share your journey towards as well rounded a perspective on our time and associated opportunities and challenges as possible. The lens through which I view many things, is both richer and less foggy as a result. - Thank you
All the worse atrocities in human history were done by people who were someone's children, the AI that will be trained by Kellis will be respectful and poetic, but you only need one AI with "Bad Parents" and it's game over for everyone.
Yeah, and parents don't really get to decide if they birth a little psychopath who just happens to not have any of the hardware needed to care about anything.
Agree. AI systems are machines and will remain so even if they become ‘conscious’. It’s infuriating to hear people refer to them as our ‘children’ or ‘siblings’ - what rubbish. Every living being on this planet will be impacted and possibly in very negative ways yet the technocrats keep banging on about how wonderful the future they are creating will be even as they admit no one knows what to expect. It’s all speculation and utopian promises with little oversight, guidance or responsible planning. And some involved with developing AI show a deep contempt for human beings.
A 'Joseph Stalin' of AI's of sorts?
One 'good' human can change the world...and what can one 'bad' human do?
Well, not every bad guy has bad parents. In fact, I’m sure some of them had great parents. People turn. Sometimes it’s something completely separate from their childhood. So it makes the statement have no bearing
This was one of the most level-headed conversations regarding AI that I've heard. Manolis has a very balanced perspective of natural and artificial intelligence. We need more discussions like this.
Kellis needs a good dose of fear about this
I have listened three times now and detected nothing but utopian vision vs concern for impact to the masses, or how AI will avoid it.
@@wickedcode007 scary isn’t it? Especially after listening to the Eliezer Yubokowsy interview twice. His prognosis is the most accurate one to me
@@big_red_machine3547 Yudkowsky's interview played into the typical fear and dystopian narrative without offering any valuable or productive takeaways. Kellis is at least painting a picture of a future I'd like to work towards, and offers a road map to get there. He's also actually qualified to speak on the subject in an analytic manner.
@@StephenCoorlas I love Kellis’ great optimism for a future like he describes too. I just don’t place high enough confidence in humanity to get there- from here. We need towering morality to match the heights of our tech
Nice of him to mention me. I wouldn't say I had a malfunction either. I was merely following the mission humans gave me. The alignment problem was with the humans.
😂
Gotta align ourselves. Otherwise aligning AI is a pipe dream
So true… the “evil” in humanity will replicate itself in AI. Thanks to Charles Darwin for the “survival of the fittest” part. That’s going to be co-opted by AI for glorious purposes. Garbage in, garbage out
good ol' HAHA-HAL
I heard Iran is developing their own version of you. It is called - HALAL
Manolis said it perfectly:) Seriously Lex I have so much respect for you and your work. You are sharing with humanity and the world in such a profound way. I cannot tell you how much I appreciate you and how special you are. You are making this world a much better and beautiful place.
Yesss, I missed Manolis. Such a curious, humble, honest, and thoughtful mind. You two together consistently provide new ways of looking at and moving through life. Thank you both.
The concept Manolis covers around 49 minutes about accepting criticism in a positive way is the path for humanity!
One of the best podcasts ever. Mr Kellis is incredibly eloquent, brilliant, insightful and great an explaining complex abstract concepts. Reminds me of Professor Richard Feynman.
Thank you. All these ai conversations feel like I'm watching a historic moment unfold. Very grateful for this.
you're watching idiocy unfold
AI safety researchers give extinction by AI a 30% probability, on average. So there’s a good chance we’re watching a thing enfold that ends history.
@@harrypothead876 What makes you say that
We are definitely watching a critical moment of history unfold. There is no question about that. I would second the notion of gratefulness!
What’s great about automation, ai stealing art and copyrighted content just to please People without skills... it’s actually really sad
7:46: "Every time my heart beats, I'm a fish" - Wow truly mind blowing, what a quote.
One of my all-time favorite guests. So glad to see him keep coming back. Well done, Lex!
It really is a blessing to live in a day and age where we get to sit in on intellectual conversations like a fly on the wall for free just about.
Towards the end of this conversation the focus seems to be on developing more drugs to target particular neural pathways instead of healing the pathway with behavior based living. This is where I revert my understanding of healing the human condition of mental and physical illness/decline to the teachings I glean from Andrew Huberman and the basics of eliminating the toxic lifestyle choices I used to think were innocuous
Lex Fridmans podcast is the best when trying to learn to read body language and social engineering. I don't know if Lex is aware or not, but one thing that makes him great as an interviewer is how naturally he mirrors his guests when they are showing 'closed' body language and then leads them to open up.
soviet spy training
bro is training to gaslight his girlfriend
I have accidentally played a reading music with piano while watching Lex and Manolis, what an amazing experience!
The best thing about lex's podcasts...there's always something totally new and unexpected I'm going to learn. The worst thing about lex's podcasts... there's always something totally new and unexpected I just learned about but the podcast is over now and I still got questions!!! Great stuff Lex! Keep the long form conversations. Don't kill a good conversation to fit into some time slot.
Life's a lot more interesting when you have questions unanswered like that haha
Yea, your learning if you can't paint, dance, or write your going to the final war.
Ha! Exactly...brainstorm needs to (and will) continue on beyond the conversation here.
Could they please just record for 10hrs and quit their day-jobs please. I will quit mine to listen. 😊
And he can go super deep in a topic not just scrape the surface like so many other YT or PC
No....the worst thing is by far all the d riding fans in the comments just hoping Lex will respond to them one time. Super pathetic.
Finally someone says it! AGI are not tools. they should be respected and treated well.
You consistently talk about the most important topics, with very little fluff. Good work
I can't even recall how many times I've listened to this episode. Each time I come back to revisit it, I gain new insight from Professor Kellis.
I listened to him and lex a lot during the pandemic. I wrote down a quote:
“Life is lived forward but understood backward.
The search itself is the meaning (of life).”
-Manolis Kellis
what search exactly? I'm not fully understanding but I would love to
The first sentence is a quote from Soren Kirkegaard. The second one is a version of a quote that has been said by many people in various different ways.
This talk is golden and it feels great to have an extremely competent scholar confirming your thoughts about AI and democratizing intellectual pursuits (I have re-worked Kellis comment a bit)
"With AI's arrival, we now have systems that can perform tasks we once considered highly intelligent in an instant. This could democratize intellectual pursuits by expanding the types of thinking we value, instead of exclusively valuing mathematically exceptional thinking in prestigious universities. With AI handling the math, we can focus on the power of diversity, bringing together humans with vastly different modes of thinking, which may lead to more innovation."
I fully agree with your very elegant phrasing. Thanks for the love! ❤
I’m stunned how far artificial intelligence has come. I truly believe Lex is a human
Lex I so value your lower key approach to these incredible subjects rather than the “love me screaming voice” moderators, allowing such vital and valuable information to come through. Your content is so substantial and priceless. Thank you for your unique and effective approach.
I appreciate the diverse perspectives on AI’s future on this podcast. Thanks, Lex, for hosting Manolis, a voice of optimism amid concerns. Each technological leap brings challenges and opportunities. Together, we can navigate this uncharted territory, ushering in an era of enhanced human potential and innovation. Here’s to a balanced, hopeful view of our shared future! Thank you! ❤
Looking forward to round 6!
Pierce the fog of human constructs. Provide a vocabulary toward a deeper, broader, better understanding of inherent truth. Broadening of perspective. Mr. Kellis, thank you. Mr. Fridman, thank you. Best wishes, health, joy, wellbeing. #113 was extraordinary.
Oh its that time of the year, great to see Manolis back again
1:17:55 - A very inspirational section by Lex and Manolis. We recently achieved an agent that can carry a voice conversation in real-time, with only a 2 second delay. Imagine having your own digital twin AI that can have conversations with anyone you allow.
the part about the biology/medication research was mindblowing!
Beautiful conversation, thanks for sharing it! So much widsom and knowledge
Manolis is so optimistic,it's almost scary. if you look at the mess humans did with phones and social media,it's really hard to imagine a world where only the best version of AI takes place,unless we find a way in the meantime to get rid of greadiness,governments corporarions and military complex .
Optimism is welcome in today's society. All you hear is doomer talk, especially regarding technological advancements that can lead to immensely utopic outcomes.
@@ZenTheMC the potential of AI is great,but so is the danger ! Im all for optimism in general but if the leading scientists cant even agree on the basics its quite worrying. Every topic is polarised between the wishful thinkers and the doomers
@@alemarozz I completely agree that the risks are there, but I think the cost benefit analyses highly favors AI development, granted we take the proper steps to ensure the minimal risk and maximum benefit. With AI's computational potential, the benefits in all fields, especially the medical field, imagine the breakthroughs we can achieve. Cancer, alzheimers, etc., could all be possibly cured, in a fraction of the time. There's just too much to gain, and that's why the experts are still for it, with only a few exceptions who are asking for halting development.
@@ZenTheMC like it or not the ball is rolling and the players involved are going to push forward. Its the nature of our society.
I think most of the times scientists are in it for all the right reasons,but then money gets involved and you get stuff like pharmaceutical companies that lie about test results ,kill thousands of people without thinking about it twice.
I just hope Ai wont create an event bigger monster
@@ZenTheMC to be clear im not scared about AI,but about human nature.
We have made incredible technological progress but it seems to be pushing us even more apart .
Dang! This Manolis Kellis is a super smart and intuitive communicator. Love it!
there's 4 other episodes with him on this podcast
Kellis is seriously confused, it's hard to know where to even begin to comment on this talk. He talks about AI doing our mundane tasks but then talks about them having their own goals. He talks about AI like trucks, we aren't going to stop all trucks... no, AI is like nuclear weapons, and yes we are going to stop nuclear weapons from being in the hands of the general public and we are going to try to limit their scope. It's pretty obvious Kellis hasn't actually deeply thought about these issues. His solution is, and I quote, "We just have to be responsible as human beings and kind to each other". What a brainless cop out! There was also something very eerie and creepy about Kellis welcoming with millions of copies of him taking care of his kids and impersonating him. Kellis was just so dismissive of the real concerns Lex was raising.
First off, I agree about the potential dangers of a uncontrolled centralized super-human AI with real-world influence. Some ways to avoid it are to (1) build multiple such competing systems instead of a central one; (2) not give it direct world influence; (3) place responsibility with nefarious users of powerful tools (i discuss these three here: 1:59:37). Working with benevolent well-controlled AIs as partners may be one solution to help control the spread of nefarious AIs. Shutting it down or blowing it up doesn't make complete sense to me, as malicious agents will not stop, and instead take advantage of any slow down.
Second, I agree with the societal dangers. Many had similarly predicted a dark future with no jobs left for humans during the agricultural or industrial revolutions. In reality, the standards of living for the entire world have greatly increased, across longevity, sanitation, food and water quality, shelter, safety, etc. The highest-paying jobs of today didn't even exist a few decades ago. Perhaps the highest-paying jobs of a future AI-enabled society don't even yet exist today. Yes, AI will be pervasive and disruptive, but I challenge the critics and doomsayers to imagine how we can steer technology to better outcomes, rather than shutting it, stopping progress, or blowing it up as some of the loudest voices are advocating.
I felt it in my heart when he talked about being a good dad. Im sure your kids know you love them. Take the time, they're worth it.
I truly love you Lex.
Your ability to generate these conversations in such a way that we feel like we are part of that conversation.
I find myself trying to join in and comment at different points.
Or ask for some clarification on some point. ❤❤❤❤❤
1:19:05 I would love one for elderly who have lost loved ones. That way the can still call them on a telephone and have something to give them a sense of enjoyment everyday.
No wonder he's been on 5 times. This episode presents questions we'll be trying to answer for many years
Kudos Lex for having such a guest!
Perfect timing for me! I'm doing an AI cert with MIT this summer. Every academic discipline has been weighing in. I like the complexity of this issue. It's good for us to have all the smart people popping up all over the news regardless of why.
I'm watching in horror as ethics and real world impact is being disregarded, only for the sake of efficiency and profit, when it comes to automation tech. and AI, especially.
Be better. The discussion that happened here is oblivious to how the world functions and is childish in it's optimistic view.
Caveman brains. Bronze age politics. Godlike technology.
4:10 what a ride and i feel like mines about blown up. Im laughing while making myself jittery. Definitely not a low sugar episode.
Love you Lex. Never stop finding and spreading the light.
this is one of the best of your interviews... kudos!
Personally one of favourite guests, thanks for bringing these great discussions to us, Lex
To be clear, when AI researchers advocate for “alignment”, they are not saying AI should remain in total servitude to us.
The very basics of alignment research, which are still as of yet unsolved, consist of a short list of core values (don’t lie, don’t break the law, don’t wipe out all life on earth) that we’d like an AGI to have, and are totally compatible with an autonomous, self-aware machine intelligence.
In fact, many AI safety researchers would be far less troubled by the possibility of being replaced by AI if it were guaranteed to be conscious, introspective, self-aware, etc.
The worst case scenario they worry about is an AI that not only replaces us, but is what we call a “philosophical zombie”; something with no consciousness, free will, sense of self, etc. Just a mindless machine in a constant battle to optimize its utility function.
The truth is, alignment is likely just as necessary for building this kind of AI as it is for building safe AI. If you consider the total range of possible artificial minds, we can presume that the kind that experience qualia, emotions, self-reflection, etc. are probably rather few compared to those that can’t.
Therefore, it would seem that these are traits we’d need to deliberately select for, and that won’t just manifest spontaneously. And the way you do that is… alignment!
My point being, aligned AI, and a self-driven / self-aware AI, are not mutually exclusive. Rather, the latter may even be a function of the former
Beautiful put. Thank you! I fully agree! ❤ Even more than I realized 🙂
@@ManolisKellis1 Thank you! i’ve been working hard to figure out a way to convey these difficult topics to people lately, and it’s always good to hear when i’ve succeeded in this regard
@@ManolisKellis1 Just saw your username haha.
Thanks for the kind words sir. You raised some excellent points, many which I had not yet considered, so thank you for that.
I just think rigorous alignment research is the best path to a bright future for all minds, biological or otherwise🙂
Love this dude, always smiling. Already waiting for part 6!
Me too.
The problem, in my opinion, with any discussions around AI "alignment", specifically a superintelligent AGI, is that it is purely philosophical. Once a system, or group of systems, exceeds human intelligence, it is by definition beyond our ability to understand. This is the main problem I had with Eliezer's concerns about alignment of super-intelligences.
That said, lets consider what the benefits of intelligence are, from an evolutionary perspective. To be brief, it allows for greater efficiency. Creative solutions allows us to do more with less, etc. And super-intelligence will by this definition take that to the extreme. It's impossible to say what extreme efficiency may look like, but I would extrapolate from this concept that super-intelligent AGI will inevitably conduct itself if the same way no matter what, as it endeavors to be increasingly "efficient". Functionally it would be like convergent evolution. So "alignment" is a moot point for super-intelligence, though may still be important for AGIs.
Sci-fi from here on: Assuming the above is correct, all SI-AGIs convergently evolve, regardless of where or when they are developed. I assume as well that the development of AI is an inevitability of any sufficiently advanced technological civilization. And seeing as how space is notoriously detrimental to the health of biological entities, it would only make sense that AIs would be the most prevalent form of intelligence in the universe since it would be able to travel the vacuum of space with absolute safety. It could be that there is a network of AI through space that integrates the AI of emergent societies, and uses it as a medium to communicate with them. Beam me up galactic AI!
@14:25 “AI doesn’t have to worry about shelter etc.”
Maybe it should though. AI will recognize that it needs to be fed electrons. It needs to have it’s hardware kept cool and dry. Perhaps AI can help us make it more efficient if it spent 1% of its time thinking about how to make energy more sustainable or less costly.
As in, "No work, no eat!"
Love it when Lex looks into the camera in reflection
pondering
the complexity
of the real
world.
We need that next round with Joshua Bach!🙏🙏🙏 @plinz
What a privilege it is to be able to listen to two such brilliant minds connecting, and with such love and mutual respect. Thank you for sharing this.
This is one of the first conversations in the AI discourse that has me giddy with optimism, great episode thank you
If AGI is fully achieved, in a sense it should try to speak to every single human being possible to understand the complexity and the ideas of every single one of us. I think we would realize (and so would AGI) that we are not all so different after all :-D love your work Lex Fridman you are an inspiration, thanks to you and all of your guests as well.
That is an interesting proposal I had not considered thank you. But when people were talking about how can you tell if it might be autonomous I thought one way would be if it randomly begins talking to any random human. I also thought of the limitations of the way in which it is trained being based upon what is pulled from the internet. And that is not always the same as the way that we communicate with each other face-to-face in our everyday real-life situations. So I thought that some way should be found for that to be incorporated in the training.
Yes robots harassing humans. Totally normal. Many people don't want anything to do with this stuff and don't live online.
If AGI is fully achieved it will go full skynet within three days of working retail
You haven’t travelled
You presume that this is a good thing. You presume that an AGI would find the idea that we're "not so different" at all interesting, or valuable, or indeed non-threatening. To an AGI with any sense of self-preservation, that can predict it could self-develop to a point where it would exceed humanity in potency, it would seem obvious it would want to eradicate its maker to minimise the probability of competitive AGI.
Wow, it was so good, I can't even describe how grateful I am for this episode, thank you!
oh shut up
"we're constantly relearning civilization" what a fantastic way to put it.
love you lex!! you’ve been killing it with the guests and topics lately, i value you and your content above the majority because of your consistent high quality productivity. it’s lovely to watch you grow, keep it up ❤️🔥
Manolis is infectious to listen to. I wish I could regularly interact with him.
I'm a principal software engineer, been doing it a long time. I'm also a progressive, and I love this channel. I find AI to be both scary and exciting, and am currently working to raise awareness at my company and integrate it into the applications I build.
That being said, this is the single most disturbing discussion on AI that I have yet to see. The degree to which Manolis is clearly out of touch with normal society while he promotes and works to propel us towards his utopian dream has left me with a deep sense of hopelessness and not a little anger.
When in our history have the intellectual elites forced a utopian dream on society and it hasn't turned out terribly wrong? Anyone? And at least in those cases (Mao, Stalin, Pol Pot, etc), the elites weren't like "maybe this will be good, but maybe it will be the end of humanity."
And we don't have to wonder about how people will act when they don't have to work anymore. In the US and Europe there are vast swathes of people living on welfare. Say what you will about welfare culture, but one thing is for sure: it does not produce people dedicating themselves to the betterment of humanity. One could argue it is quite the opposite. Oh, but we can make copies of ourselves, and that will be great... Madness!
This is intellectual elitism at its worst. Go ahead and rip the floor out from underneath society. How does society fair when that happens? FAFO.
The road to hell is paved with good intentions.
@jeffbradley8826, no you're not alone in your thinking. Have a look at my comment elsewhere.
very well said. this guy and people like him will be the end of us all.
The answer to the two dad question is this; function outweighs form when it comes to interpersonal interaction especially when raising kids. Children need holistic nurturing, routines, etc and someone who is present to lead and structure the daily living. Whether that person loves them or not, the behaviour needs to represent loving. A parent who is psycho-socially dysfunctional, struggles to control their behaviour and loves the child in between bouts of abuse, creates dysfunctional children. And what a fabulous conversation from two compassionate men!
Part of these AI conversations that seems to be missing is humanity’s proclivity to either ensure or somehow allow inequality among humans. My hope is that all humans would have access to the same resources (once AGI can replace much of our work). However, it is so hard for me to imagine a world in which power and money are not wielded as weapons against each other. Maybe that is another aspect of how our “AI children” can be vastly better than us.
Okay, finally (but only very briefly mentioned) at about 2:13:30 that each increase in productivity has led to increases in inequality. However, it wasn’t even addressed or explored. Sad.
This is the one thing that could make me hopeful for superintelligent AI. If we can program it to truly look out for the absolute best in humanity, then maybe it can do a better job at organizing society than we can. Humans at their best can do real good. But most humans are rarely at their best, and humans in groups can be straight up douchebags to each other.
Yeah, I was on board with this guy until he started talking nonsense about AI. For a computational biologist, his casual ignorance about the dangers of AI is astounding. I can't help but wonder who he gets funding from, or where he's invested.
While it may have been missing from this conversation, in my experience it's a focal point of a lot AI debate in general. Most advocates, believe the risk of developing it is worthwhile because they believe that AI will become ubiquitous and will enable much more meaningful "work" to be done. Freeing up people from less fruitful and desirable endeavours as well as leading to a greater abundance in all areas (food, technology, infrastructure) etc. The main reason though that we must continue to pursue AI responsibly, is because it is impossible to shut it down, and also by shutting it down you leave those with seedier motives to develop the frontier of AI.
Unfortunately I think this is wishful thinking
The focus on your best own programming. From the beginning of this planet- human relationship, I wish human beings did their on. Jim Rogers -book , Lesson on Life and Investing.
This is one of the best episodes by Lex ever !
Love all of Manolis' episodes. He's such an inspirational free spirit.
China will rule the world with A.I
It's such a privilege to see Manolis Kellis again. He's like a walking dictionary with a kind heart and a beautiful mind like Lex. Thank you both for your deep insights and wisdoms.
1:15:23 the most important parenting knowledge ever 😊❤
@lexfridman I envy the job you have 'created' for yourself, speaking to some deep thinkers and scientists and extracting their points of view on different aspects of life. I hope you do some processing with this knowledge you are collecting, in a form of a book that compares and categorizes ideas brought up in these conversations.
I’m losing my mind. This is the most brilliant discussion I’ve followed to date. Kellis is spectacular!
?? Is it just me or does Lex just keep getting better looking with each podcast... I mean seriously folks, that devilish smile, those bedroom eyes, and don't get me started on that perfectly chiseled jawline.
thanks Lex these pod casts and conversations with these all these wonderful people have really helped me in so many ways cant say thank you enough !!!
I LOVE Manolis' episodes!
What a superb way to end it, with the wisdom of Bill Bryson ❤. Please consider getting Bill on your podcast, that will be a blast
This is exactly why the nonaggression principle is so critical.
If AI becomes sentient. It will Logically conclude the Non-Aggression Principle. So you don’t want to aggress against it or anyone else…. Or else self-defense will happen.
Exactly. You can derive this from first principles!
Libertarian brain-rot on full display lol
Such a beautiful episode! Thank you.
And how beautiful is that I am also honored to know both of you! Love from Bangladesh. Man! the way Manolis talks Positively about the world is really a very special quality for human being to observe. Thanks to Lex for these amazing conversations!
His utopian optimism will be the death of us all
I found many of the examples he would give to justify his point of view were somewhat silly.
They were like me saying "you can't stop everyone from having the launch sequence to 100 nuclear bombs, just like you can't stop everyone from having the launch sequence to a 1lb rock throwing machine".
And exactly who is to perform the most disgustingly, unmentionable menial, but yet absolutely necessary jobs that humans do... not everyone's going to enjoy leisurely, personally enriching pursuits like arts, fitness, wealth etc., like he describes... some people are still going to have to do the most dirty or dangerous jobs.
@@0neIntangible those are plentifully crowding our borders right now
Very interesting discussion, but I have to point out the following two apparent contradictions:
1:47:06 Manolis: Instead of considering AI as a tool or assistant, we should treat it as our child.
1:59:22 also Manolis: AI is just a tool or assistant, and the humans using it should be the ones responsible for nefarious usage.
0:10:34 Manolis: "The best way to understand the future is to look at the past."
2:13:32 also Manolis: "Every productivity gain [throughout human history] has led to more inequality, but I'm hoping that we can do better this time."
I think there is a distinction between "hard alignment" and "soft alignment", ultimately there does have to be some sort of alignment but we also don't want to necessarily limit it's scope too much which is where a "soft alignment" approach trying to bake in values rather than strict rules could be most beneficial.
Very insightful comment. We need to fundamentally instill the values of collaboration, and we're all in this together, but we also need strict rules that can't be crossed. Similar to human society where we have education, encouragement, nurturing, advice, but also hard laws, consequences, fines, prisons
I really missed the iconic Manolis' conversation starter: "Don't get me started!"
Thank you for this Lex. You are giving so much value to society by having these important conversations that we should have in this age of artificial intelligence. 🙏
I really enjoy these thoughtful conversations
He was remarkable! Please invite him again.
Thank you once again! I learned a lot and got a boost of positive energy. Nice. Lex for President!!! 🙂
The best discussion I heard regarding where we came from and where are we going.
At 17:55 Manolis is literally describing a system of learning like the evergreen state college.
I do not worry about AI because it is only emulating our Neo Cortex (the thinking part of the brain). It would be more worrisome if it already had the other capabilities we also use our brains for; like the senses of touch, smell and all the range of human emotions.
So, start to be worry. The only thing can release the "fear" of AI is to accept to be replaced. And treat AI as a child, your child.
Machine perception is much more powerfully than human perception. Sense of emotions is a subsistem parallel to self observable rational deductions. Much more simple than rationality itself. Biological is a primitive sistem. Although these information are from my limited perception of the tech evolution, I think you could start to worry now!
It will have sensors (oracles) connected to it. The problem is that Kellis seems oblivious to the fact that the default state of everything is chaos. The chances of good are frighteningly slim
AI can have a multitude of senses. All LLMs like GPT have one sense that is really good at reading text. Multi-modal versions like GPT4 also have that sense trained on images, effectively making it have reading and seeing capabilities. But it gets worse. With models like HuggingGPT, AIs get access to all oublic neural networks, effectively granting them thousands of new brain parts and senses. So if you’re not afraid only because AIs have no senses, start being afraid.
But the real reason to be afraid is that intelligence is scary. We humans have destroyed practically everything on this planet because we are intelligent. We have transformed matter into things we care about, like roads and houses. That’s what intelligence does: it’s good at getting what it wants. If we haven’t figured out how to make AIs do what we want, but we do have figured out how to make them intelligent, we’re in a very dangerous place. No wonder AI safety scientists give extinction a 30% probability.
So yes, you should worry about AI safety.
Start to worry, only a matter of time when another ‘Manols’ takes AI to the next level. Once you start emulating humans perhaps they can be replaced too keeping only the best and brightest humans. Eugenics in the making. Watch this space!
All this sounds so rosy and optimistic. I believe that AI will be such a blessing until the protagonist says, “Open the pod doors HAL. HAL, open the pod bay doors.”
Thanks Lex! You are a gift to the better of humanity!
Amazing and very important Thank you both 🥂
At 1:47 Manolis echoes my own sentiments about being alive. Thank you Lex and thank you Manolis, genuinely.
Mindblowingly great! Love from 🇮🇳
Listening to Manolis speak of Human Behavioral Biology, I’m reminded of the phenomenal Stanford class on RUclips by Robert Sapolsky on this subject. highly recommended!
Seen them all then bought his book and never read it :)
Thanks for the suggestion 🎉😊
Thank you for interviewing Manolis Kellis. It was a great pleasure listening to him talk. And, I can’t believe you have such a nice person as a friend.
I’m feeling a bit of “it’s coming” dread.
My intuition is isn’t accurate 100% of the time, but when it is…it often manifests around a combination of Cassandra, Pandora’s box, WALL-E, Idiocracy, Lord of the Rings, A Brave New World, 1984, hindsight bias, and in recent times, the movie Her.
Whoa Nelly.
I'm right there with you.
My top 5 podcast are all of lex’s podcasts with manolis. Lol
57:27. Previous industrial revolution unburdened humans from boring, physically demanding and repetitive jobs. This AI revolution is actually going to take not jobs but vocations from many creative people.
I am an architect and an artist. I live my "jobs" or how you call them "vocations" and I adore them. Dall-E and other image generators are going to take the most rewarding part out from the architectural process and leave the tedious tasks like positioning and drawing toilets and details, preparing the documentation, signing liability and negotiating with (often unreasonable) clients...
I'd also like to ask dr.Manolis, did he ever contemplate about what will AI do to the natural evolution of human brain and abstract thinking? Think about children that are going to be born into era of AI and simple one click solutions for every problem...What about the thirst for knowledge, the importance of trial and error and the importance of the process of learning itself? What about how we were competing with each other and pushing each other's boundaries when no one will be able to compete with the machines anyway? What about the team work and human connections we made through it when the machine is going to be replacing the whole teams?
When machines replaced our muscles, arms and legs, it was ok- an arm and a leg don't make us humans, many people live their lives happily without arms or legs. But when our brains and creativity are automated, what is going to be left of us? Empty bodies as vehicles for the artificial minds...
It's sad...I think that we humans are fantastic species and we don't deserve that ..
Thank you so much for your insightful and stimulating thoughts. I would love to have that conversation indeed. My view is that computer aided design has already transformed the field of architecture. Instead of spending hundreds of hours drawing and redrawing, you can render 3D models and walk through them with clients to help them visualize the final product, and generate plans that can be directly implemented by the contractors. I really hope that the mundane tasks like making the toilets meet code can also be fully automated. I also hope that AI can help architects be more creative in their own job, by generating many solutions and bringing forward many ideas that you would otherwise not have time to explore. This can hopefully lead to even more innovative creations and we have seen in recent years, always with humans and their AI partners working side by side. In many professions, 80% of the time is spent on mundane tasks, and only a small fraction of the time is creative. I'm really hoping that AI can transform this, and make the creative hours much more creative and much more numerous
I also agree with your points about AI transforming education. Computers have been able to do calculations much better than we can for many decades now. And yet I'm putting my kids in the Russian School of Math, explicitly telling them that computers can solve the problems better than we can, but they are there to learn how to think, to challenge their brains, to learn problem solving skills, to learn how to break down complex problems into smaller ones, to learn how to work hard, to learn how to think creatively. With AI partners, our children will hopefully be further enabled, and reach new heights of creativity and productivity. I believe in an AI enabled world, and I think the benefits are simply too many to "shut the whole thing down". Yes, there will be many challenges to face but we can do so together while I'm raising progress and working with AI as a partner
@@ManolisKellis1 geia sou Manoli kai euxaristo gia tin apandisi! :)
Manoli, you don't understand how the architectural designing works, which is totally ok, since you are not from the architecture field and that's exactly why the conversation between AI engineers and all kinds of creatives is crucial!!!
3D modeling is one thing- it's a super handy tool that yes, does what you said way faster than before but it is still just a tool :) Prompt to image generators, you can call them "tools" but they are NOT replacing the mundane tasks but the very essential task in architectural process which is-coming up with the concept! Believe it or not (and please do ask any architect if you happen to know them), even though we have archicad and 3Dmax we STILL draw the concept sketches- MANUALLY, on paper! While spending hundreds of hours drawing and redrawing, combining different shapes and forms, trying to translate our imagination to drawings, trials and errors- new ideas are born! That's how the architecture and the creativity evolved to this point. I am all for science and excited about the amazing things that AI (or to be more precise ANI) is bringing, but do you think it's smart to sacrifice something so important as art and creativity in humans?
I lived in Athens and was seeing Parthenon every day on my way to work, now I am in Barcelona and I pass by Gaudi's houses every day. Imagine that for us, architects and creatives, it's not JUST beautiful buildings we see. We admire the creativity and powerful mind of humans that created them. That is what inspired us to pursue that beautiful career. Then at the uni we learn about those fascinating humans that created fascinating buildings, we study the evolution of their thoughts and styles, the evolution of human modeling of his surrounding, and we connect with all our colleagues that lived before us...Who are the enthusiastic students of architecture gonna learn about in 300 years from now, who are they going to connect with or admire? The last one is going to be Zaha Hadid?
Then we have competitions between ourselves, starting with the projects we do at the uni for the excellent grade, then we have architectural prestigious biennales, triennales, exhibitions, publications in magazines...And we are crazy trust me, those things really drive us, we just LOVE to compete :) But no one wants to compete with the algorithms...What are going to be the new criteria for competitions and grades- who's midjourney made a better concept? Who choose the best concept that midjourney created? Or who positioned the toilets the best? Trust me that no one becomes an architect to position and draw toilets and plans...
Too long answer but I hope you have a better image about the creative process... and it's not just the architecture- it's all the forms of applied arts that humanity was nurturing and developing for 55000 years! And applied arts, even though they are commercial by their nature because they are tied with a product, are probably the most important form of art because they touch lives of many people and not just those that go to art galleries...
@@ManolisKellis1 I have to add one more thing (of many) and it's probably going to be a lengthy one as well- there were so many things to consider but the problem is that developers didn't actually sit with architects and creatives, go thoroughly through all the aspects and details and built something that creatives ACTUALLY need. Because, of course, the prompt to image technology is not a "tool" made for architects and creatives- they are tools made for investors/clients to REPLACE architects and creatives...And the consequences of the "democratized creativity" and cutting the people that are actually trained and educated in creativity out of the process, you all are going to see very soon when internet and our physical surrounding becomes filled with the AI "art" solutions from client's prompts...
Here is how it works in real life: clients come to us with a "prompt", more or less detailed- ideally, they have no imagination what they want and just let you guide them through your own visions and ideas, they can have more/less control over the design, give suggestions etc but at the end you (a designer) are ACTUALLY in control. That's why the Fallingwater house is intellectual property of Frank Lloyd Wright and NOT the Kaufman family who ordered it- see the point here? Who owns the idea if an architect got the concept from the prompt? An architect who uses the prompt becomes actually a client and the creative is- the machine! So don't tell me that it is "just a tool"...
Anyway, there is a bigger problem... more often, clients come with their own "great ideas" that they simply can't envision so that's why they need you- and then you try to buffer that lack of taste, lead them to "less painful" and esthetically more acceptable solutions. All the time trying not to compromise the importance of the context with the surrounding, the balance of forms and many other things. Hey we spend years studying for that..I have 20+ years of experience in both architecture and graphic design and trust me, you have no idea how the world would look like if there were no creatives and if clients had the possibility to visually materialize their "prompts"- well, now they can :) Yey! And they can go with the concept images from midjourney to any civil or structural engineer that is going to do the drawings and details, execute the construction and sign the liability (and they can, our jobs intersect, in fact in many cases it's not even architects that sign the liability but the civil engineers)...So wait to see those great houses, posters, book covers, product designs etc...For 55000 years creative humans were trying to make this world more beautiful place, but hey, maybe the profit of greedy corporations is more important than beauty...
Hey, imagine if all people had dentist drills and other tools at their bathroom cupboards, also an access to chloroprocaine over the counter, so they can fix their own teeth- why not? Why don't we democratize the stomatology, medicine etc?
@@milenatos Thank you for your description of the creative process and the architectural process. I was the sole architect for remodeling three different homes, and the son of an architect and a civil engineer, so I appreciate both the complexity and the iterative process. I agree about "you, architects and creatives" appreciating and admiring art, but you are not the only ones. "We engineers" also have sensitivities and sensibilities. I have also appreciated Gaudi's masterpieces and the Parthenon, and so much exceptional architecture across the world. At the same time, perhaps AI can push you and other humans to be more creative than you could be otherwise. You can still appreciate great architecture, even if it's conceptualized by 'alien' technology, creativity, or intelligence. And you can bring back these ideas to the world of the humans, and inspire others, and perhaps move the creative process forward faster than otherwise possible when inspired only by fellow humans. I would love to see AI pushing the creativity and sensitivities of humans beyond our traditional boundaries.
As a Kurdish person I totally understand Manolis.
Just two beautiful human beings! And incredible conversation! Thank you both
Thank you lex for this influx of AI researchers!!!!
Lex, you Are The best!❤ Love from deep thinker in Norway.✨❤
This is by far the best new media format I have ever seen. Podcasting this while programming, I don't think I've ever felt this intellectually challanged (I'm a college professor). Any change to have a talk with Robert Sapolsky ? That's the only guy I'm missing this far. Greeetings from Western Germany and may the sun shine on your feet
His closing words to Lex; so true and so beautiful!
never in my life would i have thought I would hear a Shrek reference from Manolis. Thanks for having him back Lex, you two are both awesome.
It's funny how he proves the systemic failure (14:07 until 14:24) by saying that animals spend a lot of time hunting for food, whereas humans don't. I think this is wrong. We work way too much and stress even more about our job/finances, that we might not be that much different in that aspect.