@NicholasWilliams-h3j this is an extraordinary claim and I guess it begs the question... what actual evidence do you have of this havibg taken place? He solicits feedback and subjects his work to peer review with citations. I assume you are referring to something more than how we all assemble new insights and novel applications from the space of extant ideas. I'm honestly trying to understand what you're referring to and am willing to update, but it is kind of a cheap shot to just slander a man's reputation by lobbing a baseless accusation. I'm willing to update, but the burden of proof is on you, @nicholaswilliams-h3j . Sounds like you have an ax to grind, but it seems more likely that you have misunderstood how all science moves forward on the shoulders of prior giants who themselves stood on the shoulders of those before them. What uncited references or idea theft are you talking about?
Yeah man , this should be a universally accepted mindset as a way to go without to much of a hassle. Like little obstacles like world wars and genocidal disgreement issues xD
The first time I listened to Dr. Michael Levin, it was purely because an algorithm suggested it. Initially, though, I experienced a strong sense of cognitive dissonance while listening to him. It was frustrating because I couldn’t immediately understand why I felt this low-level irritation with what he was saying, especially since I’d never encountered him before. Then it hit me-the discomfort came from the fact that he was introducing new ways of thinking about biology that didn’t align with my own understanding. I think, on some level, I felt like my beliefs about biology were being challenged. Interestingly, this reaction reminded me of something I’ve noticed with music. When I was younger, music I didn’t immediately enjoy often ended up being the music I liked the most in the long run, providing lasting enjoyment. The reason I’m mentioning this is that, much like my experience with music, even though I didn’t initially connect with Dr. Levin’s ideas, I’ve since come to realize that his perspective represents a more accurate way of thinking-a fresh take that’s ultimately better. ..In my opinion.
My neighbor is a doctor who studies cancer and I have tried to talk to them about Levin's paradigm of cancer as disconnecting from the cellular collective. It has been amazing to watch the shutters go down behind their eyes. It told them that Levin's group can induce cancer phenotypes without oncogenes, can make cells with extensive oncogenic damage not turn into tumor cells by reconnecting them, can identify tissue that will become tumor before it changes by watching it shut down communication with its neighbors, etc. Those are some pretty provocative data points, but it just causes what looks like cognitive dissonance avoidance.
@lenniedoan4253 I initially found his work because I was looking for material sufficiently divorced from my background (economics) that I could fall asleep to it--then I go in trouble when it all started making sense and, it turns out, he is doing economics (our models have agents, his anthrobots are composed of agentialaterial, etc.). I would love to work with him.
Levin alluded to ideas from Dan Dennett, Thomas Nagel, and Imre Lakatos in the span of 5 minutes. This on top of his knowledge of physiology, cellular biology, and computer science. Love this guy
Levin, Bach and Wolfram are the most amazing human beings on this planet and I am unspeakably excited that they’re working together. If there were any group that could figure out AGI and consciousness, it’s these three geniuses and their colleagues. Levin deserves a Nobel Prize already (not that it really matters at this point)
Finally!!! I was wondering when you'd get Prof. Michael Levin back on. He has some of the answers that many people are looking for when it comes to artificial intelligence.
Wow, this was one of my favourite MLST discussions... so thought provoking and profound in its explorations of intelligence and agency. I especially loved the exploration of 'wholes' and their goals, and his distinction between selfish versus small, which is so rich a metaphor I think I'll be pondering this for months!
I'm all for augmenting the human and transforming ourselves, but I think it would be better to choose if you want your body or not. There are definitely good things to take with us when it comes to the human body.
The biological vehicle has a shelf life. Being biological cells are hampered by time, then the human Mind for extra longevity is in need of another vehicle source. Religions declare that source to be spiritual.
Sorry, I mean advanced intelligence. Of course we're not about to end all life. But you see - my fear (via neurotransmitters) got the better of me in that statement.
There is a tremendous amount of knowledge and information packed into just over 63 minutes here. To fully understand what has been shared requires hours of inductive analysis but it is well worth the effort.
He's found and points to a whole new universe of understanding. Ones brain may explode with the shock if it weren't held together by the coherence of the new stuff.
Why do I find that rinsing my face in cold water then brushing my hair, wakes me up even when I thought I was already fully awake? I do hope Michael Levin is awarded the Nobel prize.
Finally got around watching this incredible episode. I discovered Michael Levin's work about a year ago and since, I have been very interested in the notion of agency. His approach has given me ideas about my own work and made me realise where I want to go with my own research and experiments. Agency is found in such a wide range of fields that I am not surprised it suddenly clicked in my head in relation to my own work in multilingual communication. Part of that work is sharing my thoughts about this topic on UA-cam. The questions were great as well, every bit of the conversation was substantial. Great job! :) This being said, I would like to see a conversation between Geoffrey Hinton and Michael Levin. I would like to see Michael's comments on the notion of intelligence as being an independent property as described by Hinton. In any case, listening to both of them made me realise that the future is going to be... something we can't even imagine right now.
In terms of features of humanity that one might want to preserve (13:00 or so), I think all three of you missed Dr Levin’s observation (from a published 2022 paper) that care (e.g. compassion) is a driver and feature of intelligence. It’s not just some arbitrary property that “some people might want”.
it's very easy to miss, especially if you've almost never experienced true compassion youtself, like me, bc I'm mostly not compassionate and i tend to ignore anything related to compassion, suggesting it to be some extra feature, which is neat but not fundamental. it is a mistake. and your egoistic mind, if such, have to develop rather deep perspective on evolution in order to start gradping compassion's evolutionary significance
Great guest, great show as always. Also on the lighter side of things, welcome to the bald club lol. I've been shaving my head for a year now. It was a little hard making the transition but now it's home. You look good by the way. 😊
This interview should be required watching for any serious student or practitioner of Western ceremonial magick, Solomonic magick, or any Kabbalistic/Hermetic art. Pure gold, what a gift.
Questions: 1. Why should we accept change, transformation and extinction as true and universal just because our abstraction of local evolution says they are? Dr. Levin goes at great speed here when we should slow down and think harder. Humans are unique in not being just subjects of evolution but agents of evolution. We have deliberately evolved plants, animals, technology and ourselves. Why would we suddenly abandon that agency because of an intellectual "axiom," and submit without question to impersonal forces of change and transformation? 2. Keith pushed back on Dr. Levin's bodily limitations argument (axiom?), but the response and subsequent conversation did not quite address the subtle philosophical point that I think Keith is making. What look like flaws and errors are often subsumed/sublated/explained in the larger goals of Life, so our short-sighted attempts to "correct" them may run counter to those larger goals. We have seen this again and again in our many failed interventions in ecosystems (introducing rabbits into Australia). Similarly hastily replacing ourselves with a supposedly superior species may result in a paperclip universe and the extermination of life itself.
The cellular biological vehicle has a shelf life "extinction". Humans have not modified any plants, animals, and living organisms beyond cellular decay. Surely, a sign of the Limitations of Materialism. However, religions declare mental survival post the mortal biological vehicle's decay.
They dont even know what Thoughts are, (without which they have nothing at all) and it does not bother them in the least. Nice warning, but futile I fear.
Maybe I've misunderstood your comment but I don't think he was implying that we should replace ourselves with a supposedly superior species. If you're referring to AI then I think his message was very clear. He basically said be careful and that it is not clear that these LLMs or multimodal systems couldn't have emergent agentic capabilities. I probably understood about half of what he said but if I got the gist of it, he was saying that the intelligence space is much broader than what we humans know and he doesn't think that this is reason in and of itself to not try to explore it (without creating paperclips of course). Central to that is what makes us human is compassion. If anyone can be arsed to correct me or explain things in plain English it would be much appreciated.
1. Problemathema The Bowtie Concept can be framed as a problemathema, where the compression of information (genetic material, latent representations) and its expansion (organism development, decoding outputs) are not "solutions" but thematic narratives of transformation. These processes are stories of how systems navigate the tension between simplicity and complexity. Critique: While the bowtie illustrates elegant efficiency, it may oversimplify the narrative by ignoring the dynamic interplay of emergent factors that influence both compression and expansion. Problemathema highlights the need to embrace this narrative complexity rather than reducing it to a mechanistic schema.
I think there's an arguement to be made that Science Fiction sometimes creates the futures it predicts by influencing the thinking of those who in later life go on to be researchers and scientists.
I like his thinking about we being the body of generative AI - it's like we have created a hypermind. Perhaps it can even be argued that language itself can be viewed as an intelligent process with agency.
At 33:55. The Pavlov example:. Replace the dog with a human, but the human knows he is being experimented on, and he doesn't like it, so he lays down and starves to death, what does that say about agency?
Llms are discrete. This means they can perform work more easily. Reinforcement reward systems reflect analog (eg emotion) are harder (like heat) to convert into useful work.
My notion that "action based on state" is the foundation for consciousness can also be used as a foundation for agency. This aligns with Karl Friston's thermometer. If freezing water damages a pipe, was the water acting as the agent? I see no reason that we can't give this as the foundation and build from there. It makes engineering easy. No intention is necessary. Surprisingly to many, one day we may discover that all our actions were deterministic. Philosophers are still debating this conundrum. I don't know if we'll solve this in my lifetime, but I think we can be pragmatic about it and build what we can with simple paradigms.
Conditional probability, (bayes) is similar to pavlovian conditioning. Should not be confused with accurate work representations (eg. explanations, discrete topological) . Emotional responses (eg. salivation) is better thought of as useful (inaccurate) noise (heat representations).
"What would you like to be replaced by?" -- This is perhaps at the essence of what we consider agency, and how we can influence the future. The quality of how we develop our offspring is what we can do well, if we want to be of true service to all of it. Other than that, we are each just a flash in the pan.
It's nice to think that nature by itself is intelligent or biological systems to the simplest but is not to be intelligent you have to be aware of own decisions or information processing at least to some level and of course motifs for it, you can't just rely on embedded chemical processes that in systems create sort of balancing mechanisms and say this is intelligent processing. Each cell is not aware is just triggered by environment of other cells or external environments. Intelligence is not just processing information, it's about understanding relations of changing patterns of information and motifs for such information relevance to the one that exhibit intelligent behavior. It's abstraction of relations between preprogrammed conditions and potential relational outcomes. And to abstract own preprogrammed conditions for something else then just simple reaction to environmental factors, you have to be conscious of own existence, cells are not conscious, not even primitive living beings like insects and worms, they are just super automated living organisms. But natural balance that is embedded in living systems is to some extent beyond intelligence capabilities, because intelligence is not above or beyond nature but is part of is emergent factor of organization which have specific function which i described. And yes you have to be live to be intelligent, otherwise you just simulate some pattern recognition on some simulated predispositions, which is not intelligence. Intelligence is life emerging phenomenon and requires motifs of life maintaining and homeostaic balance reliability, which is very deeply engraved into biochemical processes of living systems, it's very complex mechanisms that evolved in long period of time and various environmental conditions. Like univers in universe.
I'd argue that "primitive" insects might very well have apparatus for evaluating expectations. and many a far beyond what could be called primitive. e.g. a spider must carefully evaluate conditions and expectations to optimally place his web. i realize that it's very easy and natural to reject my point if you are preprogrammed to assume that small animals are "primitive" and if you didn't spend time carefully watching them. it is also easy to state here that I'm just 'projecting' etc.
@@cyberbiosecurity primitive as complexity of brain structure, or central mechanism for processing signals or ability for abstraction of such signals for various purposes than just evaluated expectations on preprogrammed scheme , I didn't mean totally redundant because every even simple cell is highly complex, as i mention last its univers in univers.
@@svetlicam it's just a common belief that other animals are primitive and are guided only by (vaguely-defined) instincts. in the meantime, i've seen curious insects that interact with mammals out of curiosity or other motivation, i've seen a wild usual nervious scared cockroach domesticated and radically becoming docile and polite and even sligthtly playful, and other examples of unexcpected behaviour that does not fit into "primitive" at all. on the other hand, i've seen people so "primitive" it's hard to believe in this. i'm just sharing my opinion and once again i understand that my examples about insects are very easy to percieve as just another case of projection. for example, ua-cam.com/users/shortsEOuOFHyuIzw - what is mantis doing here exactly and for what exact "primitive" purpose? if you watch closely.
Judging the universal destination might seem really abstract but what is not so clear is a wise observer must understand the destination and what destinations are to be avoided.
This man changed my life and I started self-transition into Bioinformatics from a generic programming job. My dream is to be able to contribute to his work/team.
How's this for a definition of intelligence: The relative performance of a more complex system against its less complex peers. Occam wants to slice off those extra bits, are they paying rent? Can you secure more land with better equipment or can you do the job better with just more cousins?
58:15 Perhaps I'm missing something, but what is there that is not understood about what bubble sort does? Is this just a poor choice of words? Does he not know? Is there some obscure feature under some niche circumstance that this 101 level algorithm has and I am unfamiliar? I've really enjoyed hearing his way of thinking in this and other interviews, but this choice of phrasing reallly throws me for a loop. Perhaps I'm overthinking it?
51:00 Exactly, we do have to be careful about emergent systems we create, as with AI, as with new bioforms we create, as with any progression. It is certainly wonderful that we as a collective are having this conversation, is it not? As we speak, we are conducting homeostasis of the highest level known to humanity since its inception. Of the many perspectives one may have towards such a system, I prefer to think this is not simply a human collective, but a living collective. The collective continuation, unto which we progress towards the inevitable coldness of space, where we shall lay out our beach chairs and share a quick pint while we watch the last invisible holes shred their last electrons.
Michael Levin asks us to adopt his open, inquisitive attitude towards evolution and the diversity of intelligence. In other words, to be open-minded and interested in a broader scope of intelligence than just the human form. Personally, I find it more interesting to start by asking what intelligence actually is when simplified to its core. Intelligence, in my view, is the ability to recognize information within data stacks from a holistic relational subfield. For us as humans, it is an integrated tool within our consciousness, but as a quality and tool, it could also be artificial, as it is not the same as consciousness. This also implies there is no reason to replace us just because something is more intelligent than us-cars are faster, computers calculate quicker, and AI chess engines are better than grandmasters. Remember, pure intelligence still lacks motivation, perception, and awareness. So, do not confuse it with consciousness, which is something entirely different, with authenticity and uniqueness beyond just intelligence. That said, no species will exist forever.
My thoughts on progressive A.| Robotics and human evolution. A.|… Aiding Humanity Forward. What is our next step in evolution? Three possibilities. If "A.I Robotics" were programmed to aid our survival, then the A.| may probably factor itself out of the equation, or the human race instead. "I'll explain." First of all "let's face it" we rely, far too much on A.|. Secondly, the A.| may one day be programmed to do our every day jobs. Thirdly, within a few generations, we'll start to forget how to do things and become useless, lazy and unable to protect ourselves. So the A.l may conclude, that our own destruction is certain and then the A.I "factor itself out of the equation" to aid our survival, by giving back our independence. "The brain needs conflict, education to grow and mature." If the A.I has the capability to learn and evolve, faster than our own capabilities, then it may aid humanity, once again, by "factor the human race out of the equation" allowing our A.l to go forward to the next level of humanity, or maybe we’ll become interwoven with our technology and become cyborgs.
The least action principle is not a fundamental thing in physics. It is just a computational device used such that for some choice of Langrangian it finds stationary solutions that lead to (e.g.) conservation laws and other laws of physics that then have to be compared with experiments. The burden becomes on finding a Lagrangian that makes it so that when finding the least action it gives outcomes that agree with experiments. So nothing about the least action principle says something about nature, and certainly is not a basis of agency.
@@asdf8asdf8asdf8asdf It is core of physics, you miss the point that it is by itself nothing meaningful without a Lagrangian. It is not a principle that independently optimizes.
🎯 Key points for quick navigation: Humanity and species cannot persist in their current form; adaptation or extinction is inevitable. Levin introduces the idea of choosing what humanity could evolve into or be replaced by. Explores the nature of intelligence and challenges traditional views of human-centric traits. - Levin discusses the concept of "diverse intelligence," focusing on identifying and relating to non-human minds. Highlights the limitations of human perception in detecting forms of intelligence beyond our medium-scale, embodied understanding. Emphasizes the need for new theories to engage with other forms of minds, both biological and beyond. - Levin acknowledges science fiction as a medium for exploring possibilities grounded in physics. Uses science fiction to challenge contemporary perspectives and to speculate on diverse forms of intelligence. Advocates for humility in recognizing the fluidity and potential expansion of scientific understanding. - Discusses "human chauvinism" and the anthropocentric bias in defining intelligence and humanity. Advocates for a future where beings surpass human limitations, with traits like compassion and wisdom taking precedence over physical traits. Explores the implications of rethinking "humanity" in terms of shared existential concerns and cognitive empathy. - Levin emphasizes humanity's inevitable evolution and the importance of shaping future replacements aligned with meaningful values. Suggests abandoning the fixation on human physicality in favor of fostering wisdom and compassion. Reflects on evolutionary processes and their lack of rational intent, urging intentionality in shaping the future. - Defines humanity as entities capable of compassion and shared existential questions. Highlights the importance of moral computation and cognitive compatibility in relationships. Challenges traditional criteria like DNA or physical anatomy as defining "human." - Explains the "bowtie" model of compression in biology and cognition as critical for survival and adaptation. Compression of experiences or genetics creates simplified representations for decision-making and reproduction. The decompression process necessitates creativity, enabling evolution and intelligence to progress. - Biology's inherent unreliability fosters adaptability, as it operates without fixed configurations, ensuring resilience to environmental changes. Evolution prioritizes solving novel problems over perfect replication, enabling creative problem-solving across varying conditions. The process of compressing and interpreting information creatively is fundamental to intelligence and agency. - Evolution provides systems with general problem-solving abilities, independent of specific structures or morphologies. Intelligence, including learning and memory, manifests even at basic biological and molecular levels, suggesting primordial agency. Levin proposes an empirical framework for studying agency, avoiding ontological debates about consciousness. - Agency can be observed across a spectrum, starting with fundamental principles like least action and free energy minimization. Tools developed for analyzing higher-order systems (e.g., neuroscience) can be applied to simpler systems, revealing unexpected agency. Examples include modeling gene regulatory networks to uncover learning and memory-like behaviors. - Biological networks, even simple ones, exhibit behaviors such as habituation, sensitization, and associative learning. By applying behavioral science techniques, researchers can influence biological systems without altering their structure directly. This approach bridges neuroscience and molecular biology, leading to innovative applications in regenerative medicine. - A balanced approach integrates analytical rigor with exploratory principles, enabling discovery in complex systems. Insights into biological feedback loops reveal unexpected connections, such as tactile influences on physiological responses. By focusing on emergent behaviors, researchers identify common threads across systems, avoiding purely reductionist or random approaches. Complex biological systems can offer controllability through goal-directed behaviors rather than micromanagement. Living systems work with multi-scale competencies, making interventions more effective by aligning with the system's goals. Treating systems as active collaborators, rather than passive materials, opens new possibilities in biomedicine and engineering. - Science relies on metaphors and abstractions, which serve as tools for cross-disciplinary research and innovation. Effective metaphors allow the reuse of concepts across fields and drive the development of new research methodologies. Emphasis is placed on improving metaphors to better explore and understand complex phenomena. - Embodiment is redefined as the process of perception, prediction, and action in any problem-solving loop, not just in physical space. Biological and artificial systems solve problems across diverse spaces (e.g., physiological or gene transcription spaces). Intelligence and agency can manifest in spaces that are not traditionally understood as physical, expanding the definition of embodiment. - Intelligence may encounter natural limits if it advances without parallel development in compassion or other "life-positive" traits. Levin argues against placing premature limits on intelligence, suggesting potential self-regulatory mechanisms in intelligent systems. Collective intelligence and emergent systems often transcend individual goals, creating unforeseen risks and dynamics. - Cancer represents a breakdown in the cooperative "cognitive lightcone" of cells, reverting to amoeba-like selfishness. Cells lose their integration into the broader network, shrinking their goal scope to individual survival. Reconnecting cancerous cells to the network, rather than destroying them, has shown promising results in animal models and could redefine oncology treatments. Emergent cognition in minimal systems highlights the need for caution when evaluating AI models. Despite their lack of explicit intentional processes, AI models may exhibit behaviors that are not fully understood. Intuitions about AI limitations are often flawed due to the complexity of emergent systems. - Agency involves goals, intentionality, and the ability to influence environments, applicable in virtual contexts. AI systems, though not physically embodied, share a form of embodiment as humans act as their extensions. The interaction between virtual and physical worlds blurs the boundaries of agency. - Human perception is limited, constructing a version of reality based on incomplete sensory data. Virtual spaces share properties with constructed perceptions, challenging the idea of privileged physical reality. Agency can exist in multiple forms, not confined to traditional physical or biological frameworks. Made with HARPA AI
This man's commitment to open-minded curiosity tempered by repeated experimentation in science makes me so happy.
@NicholasWilliams-h3j this is an extraordinary claim and I guess it begs the question... what actual evidence do you have of this havibg taken place? He solicits feedback and subjects his work to peer review with citations. I assume you are referring to something more than how we all assemble new insights and novel applications from the space of extant ideas. I'm honestly trying to understand what you're referring to and am willing to update, but it is kind of a cheap shot to just slander a man's reputation by lobbing a baseless accusation. I'm willing to update, but the burden of proof is on you, @nicholaswilliams-h3j . Sounds like you have an ax to grind, but it seems more likely that you have misunderstood how all science moves forward on the shoulders of prior giants who themselves stood on the shoulders of those before them. What uncited references or idea theft are you talking about?
@NicholasWilliams-h3jhmu
Yeah man , this should be a universally accepted mindset as a way to go without to much of a hassle. Like little obstacles like world wars and genocidal disgreement issues xD
@AIroboticOverlord okay, gloomer.
That’s what a scientist is supposed to be.
The first time I listened to Dr. Michael Levin, it was purely because an algorithm suggested it. Initially, though, I experienced a strong sense of cognitive dissonance while listening to him. It was frustrating because I couldn’t immediately understand why I felt this low-level irritation with what he was saying, especially since I’d never encountered him before. Then it hit me-the discomfort came from the fact that he was introducing new ways of thinking about biology that didn’t align with my own understanding. I think, on some level, I felt like my beliefs about biology were being challenged.
Interestingly, this reaction reminded me of something I’ve noticed with music. When I was younger, music I didn’t immediately enjoy often ended up being the music I liked the most in the long run, providing lasting enjoyment. The reason I’m mentioning this is that, much like my experience with music, even though I didn’t initially connect with Dr. Levin’s ideas, I’ve since come to realize that his perspective represents a more accurate way of thinking-a fresh take that’s ultimately better.
..In my opinion.
My neighbor is a doctor who studies cancer and I have tried to talk to them about Levin's paradigm of cancer as disconnecting from the cellular collective. It has been amazing to watch the shutters go down behind their eyes. It told them that Levin's group can induce cancer phenotypes without oncogenes, can make cells with extensive oncogenic damage not turn into tumor cells by reconnecting them, can identify tissue that will become tumor before it changes by watching it shut down communication with its neighbors, etc. Those are some pretty provocative data points, but it just causes what looks like cognitive dissonance avoidance.
interesting comment, thank you
Nicely said
I recommend watching his presentation on cognitive light cones and planarian flatworms.
@lenniedoan4253 I initially found his work because I was looking for material sufficiently divorced from my background (economics) that I could fall asleep to it--then I go in trouble when it all started making sense and, it turns out, he is doing economics (our models have agents, his anthrobots are composed of agentialaterial, etc.). I would love to work with him.
Levin alluded to ideas from Dan Dennett, Thomas Nagel, and Imre Lakatos in the span of 5 minutes. This on top of his knowledge of physiology, cellular biology, and computer science. Love this guy
3 pseudoscientists and a pseudoscientist bullshitter to top it off
Thank you for that.
Also - 15:34 Richard Watson regarding evolution.
I'll look up anyone Levin recommends.
Ladies and gentlemen, this is what a visionary is
i think he will nobel at some point for sure
The production quality is stunning, thanks for all the work you do editing in images of articles and books (like at 41:40). Top notch
The inclusion of the images of the articles and books is _very_ effective. I noticed it, too. I agree-top notch!
This is by far one of the best interviews he's done
I’ve seen many and I agree
Dr. Levin’s work managed to dramatically expand space in my mind about what consciousness is. Brilliant mind.
Joscha Bach and then Michael Levin? You’re spoiling us!
Thanks!
Levin, Bach and Wolfram are the most amazing human beings on this planet and I am unspeakably excited that they’re working together. If there were any group that could figure out AGI and consciousness, it’s these three geniuses and their colleagues. Levin deserves a Nobel Prize already (not that it really matters at this point)
You should listen to Daniel Schmachtenberger too. Very interesting to get an underpinning of our current situation.
Finally!!! I was wondering when you'd get Prof. Michael Levin back on. He has some of the answers that many people are looking for when it comes to artificial intelligence.
What a fundamental thinker, and what a great video ! Really amazing work, thanks a lot .
Wow, this was one of my favourite MLST discussions... so thought provoking and profound in its explorations of intelligence and agency. I especially loved the exploration of 'wholes' and their goals, and his distinction between selfish versus small, which is so rich a metaphor I think I'll be pondering this for months!
This is so information rich and original that we need more of Levin :)
MSLT just keeps dropping bangers 🤯
Awesome guest, thanks for asking him about his thoughts on LLM's. Don't think I've ever heard him talk about that before!
Very impressed by Levin's comments from 18 to 22 mins. Will listen to the rest at my first opportunity.
Listen to everything Kevin and everything MLST. Nothing else on the internet will be so valuable / entertaining /illuminating simultaneously. .
Levin gets an immediate like, thanks for posting.
I feel like this one was uploaded for me. I love wolfram, friston, and levin. Thanks guys.
Fascinating thank you all! Amazing science you have been up to! The most important thing I have learned is that I don’t know very much!
Flipping heck you guys are on fire! 🔥
This episode has turned my into a lifetime fan of MLS
45:56 why I love Keith Duggard's questions so damn much is attention to interesting distinctions like this
Evolved intelligence is so amazing, beyond my wildest imagination for sure 🎉.......
I loved the film, Memento. It highlighted exactly what Prof. Levin discusses as memory trace. I loved his point about salience trumps verifiability.
Good work
wow wow, this is one of the most complex. linguistic space that loops it self and traverse logical types...brilliant...metalogical.
I'm all for augmenting the human and transforming ourselves, but I think it would be better to choose if you want your body or not. There are definitely good things to take with us when it comes to the human body.
The biological vehicle has a shelf life. Being biological cells are hampered by time, then the human Mind for extra longevity is in need of another vehicle source.
Religions declare that source to be spiritual.
@@steveflorida5849 Well if we're thinking far enough couldn't we mimic certain parts of the body we like?
We have a very big problem with neurotransmitters which needs immediate work if intelligence of any kind is to survive in the near future.
Sorry, I mean advanced intelligence. Of course we're not about to end all life. But you see - my fear (via neurotransmitters) got the better of me in that statement.
@@patriciaadducci6549 which Neuron decides what is Truth and what is not true?
Is mind the Cause and the brain the Effect?
Great discussion. Great channel. Thanks for uploading .
I loved this discussion, keep up the good work!
There is a tremendous amount of knowledge and information packed into just over 63 minutes here. To fully understand what has been shared requires hours of inductive analysis but it is well worth the effort.
Commenting to boost this fantastic channel, always so well done.
He's found and points to a whole new universe of understanding. Ones brain may explode with the shock if it weren't held together by the coherence of the new stuff.
Thank you! Dr. Levin with a refreshing perspective as always.
Why do I find that rinsing my face in cold water then brushing my hair, wakes me up even when I thought I was already fully awake? I do hope Michael Levin is awarded the Nobel prize.
Levin never disappoints.
Fantastic thank you.
Dr. Tim Scarfe, your new haircut looks good.
An absolute banger my boys
Finally got around watching this incredible episode. I discovered Michael Levin's work about a year ago and since, I have been very interested in the notion of agency. His approach has given me ideas about my own work and made me realise where I want to go with my own research and experiments. Agency is found in such a wide range of fields that I am not surprised it suddenly clicked in my head in relation to my own work in multilingual communication. Part of that work is sharing my thoughts about this topic on UA-cam.
The questions were great as well, every bit of the conversation was substantial. Great job! :)
This being said, I would like to see a conversation between Geoffrey Hinton and Michael Levin. I would like to see Michael's comments on the notion of intelligence as being an independent property as described by Hinton.
In any case, listening to both of them made me realise that the future is going to be... something we can't even imagine right now.
Unless you're an AI speaking you're wasting your time.
what a brilliant talk
In terms of features of humanity that one might want to preserve (13:00 or so), I think all three of you missed Dr Levin’s observation (from a published 2022 paper) that care (e.g. compassion) is a driver and feature of intelligence. It’s not just some arbitrary property that “some people might want”.
it's very easy to miss, especially if you've almost never experienced true compassion youtself, like me, bc I'm mostly not compassionate and i tend to ignore anything related to compassion, suggesting it to be some extra feature, which is neat but not fundamental. it is a mistake. and your egoistic mind, if such, have to develop rather deep perspective on evolution in order to start gradping compassion's evolutionary significance
So excited for this
Great guest, great show as always. Also on the lighter side of things, welcome to the bald club lol. I've been shaving my head for a year now. It was a little hard making the transition but now it's home. You look good by the way. 😊
Great video
Bro is on a roll
I was waiting for this. ⚡⚡
Does the principle of least action apply to training neural nets? Does the central corridor get most of the traffic during back prop?
This interview should be required watching for any serious student or practitioner of Western ceremonial magick, Solomonic magick, or any Kabbalistic/Hermetic art.
Pure gold, what a gift.
Wat
Very interesting clarification of embodiment is made from 44 mins.
Levin is always worth a listen & snagging him, in turn, snags my sub.. ergo, hello. 👋
Who lights your set? The picture is very beautiful.
Questions:
1. Why should we accept change, transformation and extinction as true and universal just because our abstraction of local evolution says they are? Dr. Levin goes at great speed here when we should slow down and think harder. Humans are unique in not being just subjects of evolution but agents of evolution. We have deliberately evolved plants, animals, technology and ourselves. Why would we suddenly abandon that agency because of an intellectual "axiom," and submit without question to impersonal forces of change and transformation?
2. Keith pushed back on Dr. Levin's bodily limitations argument (axiom?), but the response and subsequent conversation did not quite address the subtle philosophical point that I think Keith is making. What look like flaws and errors are often subsumed/sublated/explained in the larger goals of Life, so our short-sighted attempts to "correct" them may run counter to those larger goals. We have seen this again and again in our many failed interventions in ecosystems (introducing rabbits into Australia). Similarly hastily replacing ourselves with a supposedly superior species may result in a paperclip universe and the extermination of life itself.
The cellular biological vehicle has a shelf life "extinction". Humans have not modified any plants, animals, and living organisms beyond cellular decay.
Surely, a sign of the Limitations of Materialism.
However, religions declare mental survival post the mortal biological vehicle's decay.
They dont even know what Thoughts are, (without which they have nothing at all) and it does not bother them in the least. Nice warning, but futile I fear.
Maybe I've misunderstood your comment but I don't think he was implying that we should replace ourselves with a supposedly superior species. If you're referring to AI then I think his message was very clear. He basically said be careful and that it is not clear that these LLMs or multimodal systems couldn't have emergent agentic capabilities. I probably understood about half of what he said but if I got the gist of it, he was saying that the intelligence space is much broader than what we humans know and he doesn't think that this is reason in and of itself to not try to explore it (without creating paperclips of course). Central to that is what makes us human is compassion. If anyone can be arsed to correct me or explain things in plain English it would be much appreciated.
1. Problemathema
The Bowtie Concept can be framed as a problemathema, where the compression of information (genetic material, latent representations) and its expansion (organism development, decoding outputs) are not "solutions" but thematic narratives of transformation. These processes are stories of how systems navigate the tension between simplicity and complexity.
Critique:
While the bowtie illustrates elegant efficiency, it may oversimplify the narrative by ignoring the dynamic interplay of emergent factors that influence both compression and expansion. Problemathema highlights the need to embrace this narrative complexity rather than reducing it to a mechanistic schema.
I think there's an arguement to be made that Science Fiction sometimes creates the futures it predicts by influencing the thinking of those who in later life go on to be researchers and scientists.
I love the way he thinks
I like his thinking about we being the body of generative AI - it's like we have created a hypermind. Perhaps it can even be argued that language itself can be viewed as an intelligent process with agency.
I cant understand how Levin's ideas aren't more prevalent
At 33:55. The Pavlov example:. Replace the dog with a human, but the human knows he is being experimented on, and he doesn't like it, so he lays down and starves to death, what does that say about agency?
"all we have is metaphors..."
YES!
I would like to be replaced by something as complex, that does not need oxygen metabolism. What a trouble making gaz O2 is..
Weird... I've been coming to similar conclusions lately..
Llms are discrete. This means they can perform work more easily. Reinforcement reward systems reflect analog (eg emotion) are harder (like heat) to convert into useful work.
Understanding = integration of information and its applicability.
intelligence = capacity at understanding.
My notion that "action based on state" is the foundation for consciousness can also be used as a foundation for agency. This aligns with Karl Friston's thermometer. If freezing water damages a pipe, was the water acting as the agent? I see no reason that we can't give this as the foundation and build from there. It makes engineering easy. No intention is necessary. Surprisingly to many, one day we may discover that all our actions were deterministic. Philosophers are still debating this conundrum. I don't know if we'll solve this in my lifetime, but I think we can be pragmatic about it and build what we can with simple paradigms.
Are you trying to say we are products of our environment and we should look at our environment, or have I misunderstood what you mean?
why bubblesort???
The goat 🐐
Michael had me at "impedance match of cognitive lightcones"
Love this
I love the rock climbing/skin cells example
Conditional probability, (bayes) is similar to pavlovian conditioning. Should not be confused with accurate work representations (eg. explanations, discrete topological) . Emotional responses (eg. salivation) is better thought of as useful (inaccurate) noise (heat representations).
"What would you like to be replaced by?" -- This is perhaps at the essence of what we consider agency, and how we can influence the future. The quality of how we develop our offspring is what we can do well, if we want to be of true service to all of it. Other than that, we are each just a flash in the pan.
So has thesis been defended successfully?
It's nice to think that nature by itself is intelligent or biological systems to the simplest but is not to be intelligent you have to be aware of own decisions or information processing at least to some level and of course motifs for it, you can't just rely on embedded chemical processes that in systems create sort of balancing mechanisms and say this is intelligent processing. Each cell is not aware is just triggered by environment of other cells or external environments. Intelligence is not just processing information, it's about understanding relations of changing patterns of information and motifs for such information relevance to the one that exhibit intelligent behavior. It's abstraction of relations between preprogrammed conditions and potential relational outcomes. And to abstract own preprogrammed conditions for something else then just simple reaction to environmental factors, you have to be conscious of own existence, cells are not conscious, not even primitive living beings like insects and worms, they are just super automated living organisms. But natural balance that is embedded in living systems is to some extent beyond intelligence capabilities, because intelligence is not above or beyond nature but is part of is emergent factor of organization which have specific function which i described. And yes you have to be live to be intelligent, otherwise you just simulate some pattern recognition on some simulated predispositions, which is not intelligence. Intelligence is life emerging phenomenon and requires motifs of life maintaining and homeostaic balance reliability, which is very deeply engraved into biochemical processes of living systems, it's very complex mechanisms that evolved in long period of time and various environmental conditions. Like univers in universe.
I'd argue that "primitive" insects might very well have apparatus for evaluating expectations. and many a far beyond what could be called primitive.
e.g. a spider must carefully evaluate conditions and expectations to optimally place his web.
i realize that it's very easy and natural to reject my point if you are preprogrammed to assume that small animals are "primitive" and if you didn't spend time carefully watching them. it is also easy to state here that I'm just 'projecting' etc.
@@cyberbiosecurity primitive as complexity of brain structure, or central mechanism for processing signals or ability for abstraction of such signals for various purposes than just evaluated expectations on preprogrammed scheme , I didn't mean totally redundant because every even simple cell is highly complex, as i mention last its univers in univers.
@@svetlicam it's just a common belief that other animals are primitive and are guided only by (vaguely-defined) instincts. in the meantime, i've seen curious insects that interact with mammals out of curiosity or other motivation, i've seen a wild usual nervious scared cockroach domesticated and radically becoming docile and polite and even sligthtly playful, and other examples of unexcpected behaviour that does not fit into "primitive" at all. on the other hand, i've seen people so "primitive" it's hard to believe in this.
i'm just sharing my opinion and once again i understand that my examples about insects are very easy to percieve as just another case of projection.
for example, ua-cam.com/users/shortsEOuOFHyuIzw - what is mantis doing here exactly and for what exact "primitive" purpose? if you watch closely.
Judging the universal destination might seem really abstract but what is not so clear is a wise observer must understand the destination and what destinations are to be avoided.
What does Levin mean by "valence"? 44:25
Positive or negative reaction to stimuli: binary sense of “is this good or bad for system survival?”
4:30 Fred Hoyle, "The black cloud" assigned as reading by our (my) 4th form (ie 14yo) boys school teacher.
This man changed my life and I started self-transition into Bioinformatics from a generic programming job. My dream is to be able to contribute to his work/team.
39:00 he's describing harmonics
Tim, are you OK? I hope everything is well with you.
How's this for a definition of intelligence: The relative performance of a more complex system against its less complex peers.
Occam wants to slice off those extra bits, are they paying rent?
Can you secure more land with better equipment or can you do the job better with just more cousins?
Minimum description length principle.
Levin my goat
58:15 Perhaps I'm missing something, but what is there that is not understood about what bubble sort does?
Is this just a poor choice of words? Does he not know? Is there some obscure feature under some niche circumstance that this 101 level algorithm has and I am unfamiliar?
I've really enjoyed hearing his way of thinking in this and other interviews, but this choice of phrasing reallly throws me for a loop. Perhaps I'm overthinking it?
I think many people confuse more/better.
51:00 Exactly, we do have to be careful about emergent systems we create, as with AI, as with new bioforms we create, as with any progression. It is certainly wonderful that we as a collective are having this conversation, is it not? As we speak, we are conducting homeostasis of the highest level known to humanity since its inception.
Of the many perspectives one may have towards such a system, I prefer to think this is not simply a human collective, but a living collective. The collective continuation, unto which we progress towards the inevitable coldness of space, where we shall lay out our beach chairs and share a quick pint while we watch the last invisible holes shred their last electrons.
Michael Levin asks us to adopt his open, inquisitive attitude towards evolution and the diversity of intelligence. In other words, to be open-minded and interested in a broader scope of intelligence than just the human form.
Personally, I find it more interesting to start by asking what intelligence actually is when simplified to its core. Intelligence, in my view, is the ability to recognize information within data stacks from a holistic relational subfield. For us as humans, it is an integrated tool within our consciousness, but as a quality and tool, it could also be artificial, as it is not the same as consciousness. This also implies there is no reason to replace us just because something is more intelligent than us-cars are faster, computers calculate quicker, and AI chess engines are better than grandmasters. Remember, pure intelligence still lacks motivation, perception, and awareness. So, do not confuse it with consciousness, which is something entirely different, with authenticity and uniqueness beyond just intelligence. That said, no species will exist forever.
My thoughts on progressive A.| Robotics and human evolution.
A.|… Aiding Humanity Forward.
What is our next step in evolution?
Three possibilities. If "A.I Robotics" were programmed to aid our survival, then the A.| may probably factor itself out of the equation, or the human race instead. "I'll explain." First of all "let's face it" we rely, far too much on A.|. Secondly, the A.| may one day be programmed to do our every day jobs. Thirdly, within a few generations, we'll start to forget how to do things and become useless, lazy and unable to protect ourselves. So the A.l may conclude, that our own destruction is certain and then the A.I "factor itself out of the equation" to aid our survival, by giving back our independence. "The brain needs conflict, education to grow and mature." If the A.I has the capability to learn and evolve, faster than our own capabilities, then it may aid humanity, once again, by "factor the human race out of the equation" allowing our A.l to go forward to the next level of humanity, or maybe we’ll become interwoven with our technology and become cyborgs.
ua-cam.com/users/postUgkxS7UBvpJz_hVtYGkWY-CcgbEiREqYlSgq?feature=shared
Levin is a true intellectual in every sense
This is interesting Convo...
Human/ life has a miracle factor...an unknown creator....(?)
Do you believe in the miracle of life?
I would like to be replaced by something that wasn't ruled by other people
I'm relieved to realize he's not falling into raw panpsychism as i suggested earlier and I'm glad to be mistaken here earlier
Michael Levin generously lending street cred to MLST. A truly philanthropic and charitable individual
How about something that can actually carryon
But I wonder how much human wisdom and compassion is dependent upon a having a specifically human body.
Epic
Did I miss something? Why don't we understand bubble sort? I'm pretty sure anyone could understand it in like 5 minutes.
Astigmatism and brushing hair, 2 guys play dirty:)
What a great point about the unreliability of biology 😅
The least action principle is not a fundamental thing in physics. It is just a computational device used such that for some choice of Langrangian it finds stationary solutions that lead to (e.g.) conservation laws and other laws of physics that then have to be compared with experiments. The burden becomes on finding a Lagrangian that makes it so that when finding the least action it gives outcomes that agree with experiments. So nothing about the least action principle says something about nature, and certainly is not a basis of agency.
So wrong.
Core of physics
@@asdf8asdf8asdf8asdf It is core of physics, you miss the point that it is by itself nothing meaningful without a Lagrangian. It is not a principle that independently optimizes.
🎯 Key points for quick navigation:
Humanity and species cannot persist in their current form; adaptation or extinction is inevitable.
Levin introduces the idea of choosing what humanity could evolve into or be replaced by.
Explores the nature of intelligence and challenges traditional views of human-centric traits.
-
Levin discusses the concept of "diverse intelligence," focusing on identifying and relating to non-human minds.
Highlights the limitations of human perception in detecting forms of intelligence beyond our medium-scale, embodied understanding.
Emphasizes the need for new theories to engage with other forms of minds, both biological and beyond.
-
Levin acknowledges science fiction as a medium for exploring possibilities grounded in physics.
Uses science fiction to challenge contemporary perspectives and to speculate on diverse forms of intelligence.
Advocates for humility in recognizing the fluidity and potential expansion of scientific understanding.
-
Discusses "human chauvinism" and the anthropocentric bias in defining intelligence and humanity.
Advocates for a future where beings surpass human limitations, with traits like compassion and wisdom taking precedence over physical traits.
Explores the implications of rethinking "humanity" in terms of shared existential concerns and cognitive empathy.
-
Levin emphasizes humanity's inevitable evolution and the importance of shaping future replacements aligned with meaningful values.
Suggests abandoning the fixation on human physicality in favor of fostering wisdom and compassion.
Reflects on evolutionary processes and their lack of rational intent, urging intentionality in shaping the future.
-
Defines humanity as entities capable of compassion and shared existential questions.
Highlights the importance of moral computation and cognitive compatibility in relationships.
Challenges traditional criteria like DNA or physical anatomy as defining "human."
-
Explains the "bowtie" model of compression in biology and cognition as critical for survival and adaptation.
Compression of experiences or genetics creates simplified representations for decision-making and reproduction.
The decompression process necessitates creativity, enabling evolution and intelligence to progress.
-
Biology's inherent unreliability fosters adaptability, as it operates without fixed configurations, ensuring resilience to environmental changes.
Evolution prioritizes solving novel problems over perfect replication, enabling creative problem-solving across varying conditions.
The process of compressing and interpreting information creatively is fundamental to intelligence and agency.
-
Evolution provides systems with general problem-solving abilities, independent of specific structures or morphologies.
Intelligence, including learning and memory, manifests even at basic biological and molecular levels, suggesting primordial agency.
Levin proposes an empirical framework for studying agency, avoiding ontological debates about consciousness.
-
Agency can be observed across a spectrum, starting with fundamental principles like least action and free energy minimization.
Tools developed for analyzing higher-order systems (e.g., neuroscience) can be applied to simpler systems, revealing unexpected agency.
Examples include modeling gene regulatory networks to uncover learning and memory-like behaviors.
-
Biological networks, even simple ones, exhibit behaviors such as habituation, sensitization, and associative learning.
By applying behavioral science techniques, researchers can influence biological systems without altering their structure directly.
This approach bridges neuroscience and molecular biology, leading to innovative applications in regenerative medicine.
-
A balanced approach integrates analytical rigor with exploratory principles, enabling discovery in complex systems.
Insights into biological feedback loops reveal unexpected connections, such as tactile influences on physiological responses.
By focusing on emergent behaviors, researchers identify common threads across systems, avoiding purely reductionist or random approaches.
Complex biological systems can offer controllability through goal-directed behaviors rather than micromanagement.
Living systems work with multi-scale competencies, making interventions more effective by aligning with the system's goals.
Treating systems as active collaborators, rather than passive materials, opens new possibilities in biomedicine and engineering.
-
Science relies on metaphors and abstractions, which serve as tools for cross-disciplinary research and innovation.
Effective metaphors allow the reuse of concepts across fields and drive the development of new research methodologies.
Emphasis is placed on improving metaphors to better explore and understand complex phenomena.
-
Embodiment is redefined as the process of perception, prediction, and action in any problem-solving loop, not just in physical space.
Biological and artificial systems solve problems across diverse spaces (e.g., physiological or gene transcription spaces).
Intelligence and agency can manifest in spaces that are not traditionally understood as physical, expanding the definition of embodiment.
-
Intelligence may encounter natural limits if it advances without parallel development in compassion or other "life-positive" traits.
Levin argues against placing premature limits on intelligence, suggesting potential self-regulatory mechanisms in intelligent systems.
Collective intelligence and emergent systems often transcend individual goals, creating unforeseen risks and dynamics.
-
Cancer represents a breakdown in the cooperative "cognitive lightcone" of cells, reverting to amoeba-like selfishness.
Cells lose their integration into the broader network, shrinking their goal scope to individual survival.
Reconnecting cancerous cells to the network, rather than destroying them, has shown promising results in animal models and could redefine oncology treatments.
Emergent cognition in minimal systems highlights the need for caution when evaluating AI models.
Despite their lack of explicit intentional processes, AI models may exhibit behaviors that are not fully understood.
Intuitions about AI limitations are often flawed due to the complexity of emergent systems.
-
Agency involves goals, intentionality, and the ability to influence environments, applicable in virtual contexts.
AI systems, though not physically embodied, share a form of embodiment as humans act as their extensions.
The interaction between virtual and physical worlds blurs the boundaries of agency.
-
Human perception is limited, constructing a version of reality based on incomplete sensory data.
Virtual spaces share properties with constructed perceptions, challenging the idea of privileged physical reality.
Agency can exist in multiple forms, not confined to traditional physical or biological frameworks.
Made with HARPA AI
Thanks for that. It confirmed that I actually understood quite a lot of what he said (I wasn't sure as I'm not a scientist/academic).