Вставка
- Опубліковано 20 тра 2024
- Support us! / mlst
MLST Discord: / discord
Michael Levin is a Distinguished Professor in the Biology department at Tufts University, and the holder of the Vannevar Bush endowed Chair. He is the Director of the Allen Discovery Center at Tufts and the Tufts Center for Regenerative and Developmental Biology. His research focuses on understanding the biophysical mechanisms of pattern regulation and harnessing endogenous bioelectric dynamics for rational control of growth and form.
The capacity to generate a complex, behaving organism from the single cell of a fertilized egg is one of the most amazing aspects of biology. Levin' lab integrates approaches from developmental biology, computer science, and cognitive science to investigate the emergence of form and function. Using biophysical and computational modeling approaches, they seek to understand the collective intelligence of cells, as they navigate physiological, transcriptional, morphognetic, and behavioral spaces. They develop conceptual frameworks for basal cognition and diverse intelligence, including synthetic organisms and AI.
Also joining us this evening is Irina Rish. Irina is a Full Professor at the Université de Montréal's Computer Science and Operations Research department, a core member of Mila - Quebec AI Institute, as well as the holder of the Canada CIFAR AI Chair and the Canadian Excellence Research Chair in Autonomous AI. She has a PhD in AI from UC Irvine. Her research focuses on machine learning, neural data analysis, neuroscience-inspired AI, continual lifelong learning, optimization algorithms, sparse modelling, probabilistic inference, dialog generation, biologically plausible reinforcement learning, and dynamical systems approaches to brain imaging analysis.
Interviewer: Dr. Tim Scarfe
Pod: anchor.fm/machinelearningstre...
Prof. Michael Levin
en.wikipedia.org/wiki/Michael...)
www.drmichaellevin.org/
/ drmichaellevin
Prof. Irina Rish
/ irinarish
irina-rish.com/
TOC:
[00:00:00] Introduction
[00:02:09] Emergence
[00:13:16] Scaling Laws
[00:23:12] Intelligence
[00:44:36] Transhumanism
4 episodes in 1 week.. You are definitely spoiling me Tim.Thank you
I just love Micheal, he is one of the most sharp and geniuses of our generation in all aspects.
Except he, like all so called scientists, does not know what Thoughts are, where they originate, what they are made of, and even less masters them, as in silencing the inner dialgue at will, eyes open, and exploring that silence. And, without Thought, he has nothing at all. So there is a fundamental step missing here, either that, or science has abandoned Reason and Logic and rigorous methodology at the very first opportunity.
No wonder, our "Titanic" band plays wonderful tunes, but is going down just the same.
I have seen a wound heal instantly, in a wave of love. I now understand, more or less, how that happened. I now understand, how Jesus (and I am not religious at all, I just dont dismiss stuff offhandedly) could heal people in an instant.
Those who took the trouble to master our basic faculties, those who initiated (unwittingly no doubt) our religions, dominated our cultures for thousands of years. There is a power in that mastery. Self mastery (and I dont mean dissociation,or suppression) is hard. So what.Get to it. You owe us, big time.
Modern day masters, tell us of the mind of the cells. They knew this long before Michael showed up. (The Mind of the Cells, by Satprem).
It is the cowardice and laziness of modern scientists that is destroying our planet. Yes, I am glad with the eye surgery, I can see and drive again, but what good does it do me if I can only witness our global destruction?
@@gammaraygem We have the power of gods, but not the wisdom of gods.
The way out of global destruction is for us to see more, not less.
@@gammaraygem Get help
Michael Levin's minimal intelligence postulation put my jaw on the floor.
Levin and Rish!! Instant favourite, thank you Tim!
Michael is one of the real true geniuses alive today.
Irina and Michael are an excellent combination for exploring this incredibly interesting area. She has the ability of pulling out key nuggets and introducing novel and interesting points of view. And Michael is really an amazing and original mind. He is making amazing contributions to science. This was excellent. Thank you bringing them together in your channel.
Yes, applied functions effective & efficient AI is instructing better machine instruction akin to the neurological functioning which the human brain engenders, facilitates, creates.
That's an absolutely fantastic discussion! Thanks so much, I will return to it so many times. What a clear conceptualization of systems level and emergence alone.
Divine intervention is not the only alternative to emergence during so-called evolution.
Thanks for bringing in Dr.Levin 👍
Insanely amazing talk. There is so much inspiration in this type of talk. Thank you very much.
This was the most mind opening video I've seen in a long time.
First, it is really amazing to have two of my favorite Scientists in the same interview!
I found it very interesting how people related "Groking" behavior (unexpected Emergent Behavior) to things like Safety in AI, Robustness etc... and that there might actually be some ways to model the prediction of that. My belief is that there might be a subset that are predictable, but inevitably, there won't be a solution for all cases of Grokking.
Fascinating discussion re: human intelligence... and then throw in how that might be related to some sort of Grokking heheh.
Great discussion. Thank you again for making this happen!
Statistically speaking nothing should ever exist, there are more arrangements of matter and energy that would result in nothing than there are arrangements of it doing something interesting.
So the basis of this argument of intelligence is mathematical, what exists has to exist in an environment where there are a far greater odds of it not existing.
How does life A.K.A. intelligence overcome this?
The answer is: the whole is greater than its parts. And every composition is possible through harmony.
Another fantastic conversation. Thanks for bringing it to us.
Wow, I wasn't expecting Levin. Thank you very much
Only 15 mins in and gotta say, both guests are VERY eloquent. Love it so far
"The dark side of the Force is a pathway to many abilities some consider to be unnatural” - 48:35
Amazing🙏❤️🎯
Woohoo exceptional work as always!
Beautiful. Ty
This was a fantastic discussion.
Thanks!
OMFG THEY'RE TOGETHER AT LAST
Was the recent video with Walid and Keith taken down? It was great and I was half way through. Maybe there were too many good ideas. Haha.
Thanks for all of your videos. They have been very helpful for me as an AI layman to quickly start to understand the complexities of AI and their possible trajectories. As an analogy, it almost feels like primitive humans poking at a burning log wondering what exactly it is and how to use its power.
Love your videos! yay
Panpsychist such as myself have to show how you get from the most basic elements of reality all the way up to a human being - in every respect - not just cognitively but morally, socially etc too.
47:20 the bit about transhumanism is so awesome
20:00 Arithmetic
If I remember right, transformers are basically fixed-depth optimized version of RNNs. So of course this behavior should be expected. Math is thoroughly a description of processes.
I could have listened to them for hours...
i agree, a continuum . an infinite continuum unfolding
The circumstances that match this emergent behavior in training the LLMs in the open ai paper occurred because their was a point where the outcomes were fluctuating around a certain level, and the tendency was towards overfitting. The increase of parameters included in the training set caused the sharp increase in performance and remained there. The ability for models like gpt-3 to "generalize" when scaling the size of training. data is themodel rea over
I wasn't expecting this crossover 😳
Please let them speak more and more!!!
28:18 just listening today ( 1 month late). It occurs to me that levin and rish are talking about 2 different but related things: result of structure vs result of process.
What happened to Episode 101 ?
Some difference between an umbrella and eg neural lace amenable to external control!
Finally levin in here =D
How to define the integral components that make the light bulb moment! We all know that feeling when we go from huh? To Ahh! We call it the penny drops or eureka moment or the light switch moment it tangibly moves from one state to another state. Even an incremental movement feels like a movement a position change in thought. We even say ‘just sleep on it’ in the morning a perspective a thought or an answer can pop into your mind as the first thought of the day without even revisiting the question. So the mechanics that do that very much seem magical. One moment you don’t know one moment you do know.
I think it is very difficult to see intelligence as linear rather it’s inevitable or a pre condition of all things in life to be intelligent. Like a designer mark of intelligence is inside life period.
I think this intelligence from the microscopic to the human have pattern systems that are all based on the same collective cooperations. Which is why we can relate ourselves from one being to all things or we can feel ourselves inside our blood cells. Life embodies unifying principles from the invisible to the visible. However, conditions can effect growth of body and growth of thought.
But all of the conditions are full of limitless conditions actual and conditions possible.
I think even that which is not condition actual meaning it doesn’t have to exist as a physical phenomenon to effect thought. Thoughts can be made through non physical experience such as imagination.
Thought experiments are popping off in everyone’s mind all the time for different outcomes. Some might define this as projection this then starts melting the present with the future. That which appears impossible becomes possible through focus projection single-mindedness. Thought in itself is a Universe of complexity when you can flip a thought through multiple motions. When humans are communicating with each other they are flipping multiple signs verbal non verbal instinctive reactionary reactive proactive non reactionary every millisecond packed with information on multiple levels and that is not including the way some people react to what they call spiritual information. So with all of that amount of information flipping through in real time then being remembered as memory and sped up into future there is a ton of data computing in a human brain day and night. It makes sense why mind seeks stillness.
Or a state of not thinking some times we say switching off.
Brain wave change to a different frequency can still thought or data flow but then open up another world of meditation.
So quantity of data doesn’t embody intelligence it is only one form of it.
How data reacts to create eureka moments is not intelligence on its own just a method of calculating or solution seeking.
Being empty and still is not a more important intelligence but a state of intelligence
So actions can be experiences we learn from or simply defined as experience. The learning happens from the reflective process or conditions that allow for a light bulb moment which can be a solution that comes at you side ways you are not looking for the thought but it comes at you anyway. Simply Life does this.
To ask how is intelligence made to emulate it is to ask what is Life.
The multiples are incalculable the directions are limitless when looking at the process as components.
Components are ever divisible and there are quantum fields not even known by human science so that rabbit hole is endless.
But Life has emotional and caring meanings and it makes sense to be valuing that which is motivated by caring and kindness as building blocks for positive environments for learning experiences that grow unity.
I think answers are not in units or components but shapes and structures support systems and patterns. Geometry that balances and supports and relates elements through ratio
all things equalling value to each other. All is one one is All.
damn this is a good podcast
❤❤❤ Dear Dr. Rish and Dr. Levin! Thank you for a very informative lecture. May I ask you a question about the nature of our obligatory aging and death? Thank you in advance. Valery Chuprin, 71😅
Did anyone find the article about chemotaxis in individual molecules and non-zero competencies of sub-atomic particles discussed on 28:30 and 29:30?
Is there citation list. Would love explore all those papers professor Irina has mentioned
Superb talk! I'm a fan of prof. Levin ever since I saw him on the Lex Fridman podcast (not so long ago). However, I must disagree with him, that there are not sharply emergent behaviors on different levels of development of organisms. We might not be able to say exactly where the transition happens for the formation of cognitively capable agent from an embryo, but the problem might be with our observation tools and methods, not with the existence of sharp transition. As prof. Irina Rish pointed out, sharp transitions are being observed in nearly all complex systems, exhibiting new kinds of behaviors with the increase of the scale. With that said, I do not disagree, that single cells or organs do not have their competencies and forms of intelligence and problem solving. It's just that new capabilities seem to emerge sharply with scale or stages of development.
@6:40 he (Michael) has completely misunderstood high level causal control. He gives a descriptive account, which is always post-hoc. You cannot say the organism "caused" _this_ or _that_ until you identify "the organism". So it does not defeat the ultra-reductionist. Classical biology never can. Your goal-directed or top-down causation story is always just that, a story. To get to the heart of top-down causation and genuine agency you have to violate Lorentz covariance in one way or another, there's no way around it. Quantum amplitudes do not help here. You need backwards lightcone causal violation to get _any_ genuine top-down causation. This means Hamiltonian time evolution can only be an approximation (if, that is, you believe there is genuine top-down causation and you're not an ultra-reductionist telling fairy stories like Doug Hofstadter). Jaegwon Kim was the most honest philosopher about the causation problem for biology and classical physics, and he had no solution that was not epiphenomenal.
I know of very few theories that non-statistically violate lightcone causal structure, but the Susskind ER=EPR conjectures are one way, possibly. I do not see any others. No one is working on this however, not as an approach to how genuine top-down causation can be effective. (QFT cannot meaningfully violate causality, the QFT non-local correlators are statistical, not causal.)
Tononi and others have nothing. The "work function" at a given level is meaningless, it is just an aggregate of a reductionistic work function that's been chunked by you, the observer. The atoms still do all the causal "work," in that context. The "cybernetic" stories Michael tells are valid, I believe, but he has no explanation for how top-down causation can occur. He does not understand the physics. He thinks he does, but he doesn't.
If we take into account Wolfram Ruliad (Computational Universe) and simple rule leads to complexity eg universe, then this to me explains a bottom up design, indeed as Wolfram encroaches into the Physics Fortress (using computer based proofs) quantum behaviour and large scale phenomena can be fitted, it has a long way to go but his New Theory of Physics (and New Kind of Science) may help us shed light on your valid points. Bottom up, scaling feedback etc suits my thinking (thus far). Peace
What happened to episode 101
yeah honestly
It was up for a couple of hours, hoping it'll be back too... some really interesting bits on interpolation vs extrapolation, language ambiguity, discrete vs continuous and touching on QM. Only got half way. Some slightly fractious discussion but it was good.
@@luke2642 Some of it was a little bit too unhinged - I will stick the first 49 mins back up. Ping me on Discord if you want to see the other bit. I have edited it to tone it down and hopefully will release as another episode as it was great content
@@MachineLearningStreetTalk Thanks, only just joined the discord! I don't know how you find the time to do all this! I always hope commenting on YT will marginally help mlst regarding "the algorithm", and I do occasionally get really deep, considered replies. In general though, reddit niches tend to have much higher quality of debate, but, that'd be a whole new community to grow/manage, for you. Discord is invisible "unless you're on it" and searching/scolling back through time is a PITA. But builds community faster! Any thoughts/preferences?
@@luke2642 Cheers Luke, I just saw you joined! We do have a subreddit but I always forget to post to it, I should get back on that. The ML subreddit is often salty so I only post there on special occasions :)
there is no use if we keep discuss without action..... and if we had action but at the end we kill each other.... what this is all mean?
“Natural” isn’t a static thing. Completely agree with Irina and Michael! We are just in a new phase.
Natural also doesn't mean ever progressing. What is natural is what attempts to achieve equilibrium, which may result in variations over time due to entropic degradation.
@@dunebuggy1292 slight push back - it could mean ever-progressing if ever-progressing is possible. So I’d submit that “natural is whatever is possible”.
@@earleyelisha I think I agree with that. It makes sense.
Прости, прости мне.
Я не хотел обидеть твоих глаз.
Любовь величие моей гордыни.
Любовь моя смелее ветра и огня.
И я пою как птичка про тебя.
Скажи мне сладкая голубка
Зачем не смотришь на меня?
Зачем ты сердце мое ранишь?
Горит, горит моя душа!
Я не хотел бы расставаться. Нет.
Не сейчас и никогда!
Где нет тебя там пустота.
Я вижу облик свой.
И в пустоте вижу планету звезды космос…
И тебя.
Тебя.
Да, тебя.
Acknowledged.
Hmm. "we can do better then natural" but "the chances of .doing worse are not non zero." Levin has even expressed some concern of our species ending. It would seem the most likely cause of that would be in itself human caused. How befit/cost can be considered becomes an important societal question.
What is the smallest?
do experiment plz
Professor Levin and Professor Rish in their conversation for some reason did not point out the relationship of two organisms - a cell and a human, who grew out of it, their threefold organization. This is indicated by the Cellular theory, confirmed in the laboratory and visually. The conversation was about biology, about the cell, life, but for some reason there was no definition of the life process, its energy basis; the structure of living matter, its energy basis. People grow old and die in the 21st century, the age of the Internet, Space, Atom, Artificial Intelligence, as in the Stone Age. Is it because of this? There is no physical explanation for the nature of aging: what is it, where does it start in the human body, how does it develop, can it be stopped? Thank you.
I can't catch up with this many episodes.
Don't worry, I am having a spike in Feb, you can catch up next month 🤩
How did the first cell appear to begin with?
Likely imulsification or coagulation of aminoácido chains from within in a lipid-solvant medium mix with some environmental energy inputs stimulating propagation and function. That's my best guess. A membrane only needs a barrier between out and in, lock in key receptors can tumble together from proteins and poke outward from the primordial cell to snag nutrients mechanically.
yeah people have solid hypotheses but nothing verified or exact
@@paxdriver So, in other words, you have no idea. Lol
Strugong with the reality of consciousness, as cause. An intelligent universe.
The talk ended on, "once you pop you can't stop" was that hidden advertisement? 😂😂🤔 I bought Pringles...
The sorites paradox - it's not just for philosophers anymore. Or is it? At what point does Michael Levin stop being just a philosopher and start being a biologist? Can anyone run experiments that prove this has occurred?
starting @4:15 Dr Levin referred to some goal-directed agents in the cybernetic, not magic or supernatural sense? Huh?
My thoughts is that the Environment survival is the most pressing goal for goal-directed agents. Not magic or supernatural because perturbing, the experience introduced complexity properties to fill in possible patterns into existing knowledge spaces and new transformed spaces is subtle and passive process.
We can observe emerging behavior everywhere, from single cell to complex organism, from economics to societal structure - it's a common phenomenon, specially the system is complex. E.g. biological evolution and evolution of technological innovations - it's a non-linear dynamical system. Th3 DNN is no exception. More research and studies needed to fill the gaps.
In the field of AI in past 12 years is less science, more hacking - they share similar behaviors as hacker than true scientists - they call themselves AI-researcher. Now the true scientists realized that is not good enough for further progression of AI research in the future.
Build 1 and show we.
Where did the eukaryote cell come from? How?
Nick lane is good on this, he has plausible theories while being very open that it is still essentially a mystery
I love it when people say something is “not natural”…via a device on the internet
Mother Nature is the greatest mass murderer of all time
You can do better than emergence. It is Mandelbrot, iteration.
Love to ya.
Orangutans make quite ingenious umbrellas.
Did they explain how the first cell emerged? Also, how did the eukaryotic cell emerge?
A biologist called nick lane was on the lex podcast recently and he goes into this, it seems like we don't really have strong theories on these things yet
Yes. There are some big questions left unanswered. Including seemingly Michael's own kind of proof of some intelligence in the universe. He states himself that if he searches for things with zero intelligence, he gets to a place where nothing exists. Sounds like a clear signal that intelligence is fundamental, which would by definition mean that the universe has intelligence as a property. So it is a bit strange to hear him say that nature is horrible later in the episode. He is a biologist Afterall.
@@n8works when he said nature was horrible he was just being anthropomorphic for a second
@@xmathmanx yes. He was referring to a sense that science and technology have helped us ascend above nature. Nature is brutal but science has saved us, evidently. Even though he also argues that science and intelligence are creations of the natural universe. Somehow it's OF nature but also above and better than nature.
Instead of looking at the facts and seeing a universal intelligence that has created a unique human consciousness, of which we have yet to find any equal or superior, he seems to simply wave that fact away and say well since we've identified all of these aspects of intelligence common through all aspects of biology and have translated those to silicon, that we will somehow just find our way to recreating consciousness when it's the most unique aspect of what it means to be human. So much so that we can't even describe ourselves completely.
I hope I'm being clear in describing the challenge I see with this thinking.
@@n8works if you're saying the hard problem of consciousness is still not solved then I'm with you, if Levin is saying that human consciousness is an emergent phenomenon then I would disagree with him, but I don't think he is saying that
❤🧡❤🧡🧡❤
Regarding transhumanism we will see how far it will go and what other more natural solutions for social problems it prevents from implementing. Transhumanism+neoliberalism=fascinating recipe for desaster.
I am pretty sure the people he was talking about went with there mind stright to the abortion debate and thoght of it in phsiosopical terms of what makes someone human - so they wear thinking of it being able to think like a human rather then the cell of a human
If a universe composed of nothing could exist then it indeed could do nothing and thus has never existed and never will exist.
We need a new inquisition
🎊
Hmm. Shouldn't there be some discussion about trust? Weaponizing transhumanism? Whether it would need to be available to all, equally? It's a little disconcerting to hear scientists waving off the possibility of "screwing things up" too
This sounds all great....
Except he, like all so called scientists, does not know what Thoughts are, where they originate, what they are made of, and even less masters them, as in silencing the inner dialogue at will, eyes open, and exploring that silence. And, without Thought, he has nothing at all. So there is a fundamental step missing here, either that, or science has abandoned Reason and Logic and rigorous methodology at the very first opportunity.
No wonder, our "Titanic" band plays wonderful tunes, but is going down just the same.
I have seen a wound heal instantly, in a wave of love. I now understand, more or less, how that happened. I now understand, how Jesus (and I am not religious at all, I just dont dismiss stuff offhandedly) could heal people in an instant.
Those who took the trouble to master our basic faculties, those who initiated (unwittingly no doubt) our religions, dominated our cultures for thousands of years. There is a power in that mastery. Self mastery (and I dont mean dissociation,or suppression) is hard. So what.Get to it. You owe us, big time.
Modern day masters, tell us of the mind of the cells. By introspection, observation and rigorous self discipline. They knew this long before Michael showed up. (The Mind of the Cells, by Satprem). They also stated that Consciousness is a Force. IT expresses itself through matter, and evolution is its method. One was twice nominated for the Nobelprize, he could not care less.
Yes it is messy. So is a football game. You can get hurt. On the scale of the universe, our pathetic holocausts mean nothing. We rather not see those, but it is a game, and you can get hurt. If a cancer cell gets in tune with its intended frequency again, it instantly heals. Our evolution has purpose. Sorry about that. Just align with that frequency, and you´ll suffer less. 5000 years of religion how faulty and messy it may be, testifies to that. 100 years of science, BAD science, is destroying our planet, our habitat. Thinking you´ll technofix that is just too effing stupid. FIX YOURSELVES FIRST !!
Stop whining.
It is the cowardice and laziness of modern scientists that is destroying our planet. Yes, I am glad with the eye surgery, I can see and drive again, but what good does it do me if I can only witness our global destruction?
I am piqued by the level of emotionality where you claim to be acting against emotionality or human biases while doing the very thing. This notion that the human cannot have unique properties because of some reducible nature is itself an unscientific proposition. We don't know! At 43 minutes, Michael contradicts himself and you see it in how animated he gets in forcing the thought - and he says it; he doesn't believe in anthropomorphism because the human is not special and it is a transgression to believe so. Why? You have to show that there is a mutually exclusive nature to the empirical evidence. Otherwise, you're just expressing a belief. It doesn't matter if there is a substrate explanation shared at scale. Linearity of the substrate does not preclude emergence above it.
But anyways, more sardonic commentary on human ego and this and that. I love human ego. It's precisely why we're here in the first place!
Physicalism doesnt work either. You have the placebo effect which is explained by this bioelectric field, but its your consciousness as the controller. Theres one more layer and it is the fundamental layer underpining the universe. The Mystics were closer than physicalists.
They are onto something here but there are many basic errors with the frame of how this is being understood. However, this is unavoidable. Sometimes the solution is "no solution" but we never accept that outcome because of the way we understand information progression. Transhumanism isnt possible but the answer is so obvious and offensive. Ironically the most intelligent are the furthest from this understanding due to their lack of humility.
seems like Tim lost some weight!! you okay bud?
Fine thank you 🙏
The more urgent questions are "Can biology tell us what a woman is and did masks limit the spread?".
I fell in love with Irina I will propose with my eyes closed marry me beautiful and smart
00:00:10 Distinguished Professor. That's important. Your title should be corrected here.
The insult is greater than labeling a full professor as a lesser professor, such as an associate professor.
What? He was introduced as a distinguished professor right from the start, go and troll somewhere else
@@TimScarfe I'm talking about the white text on the black background. Do a small survey of distinguished professors, they would be annoyed at that alone if they were honest. Stop blaming me for your defect.+\
Michael is clearly after impressing and this annoys me. Biology is not a continuum, a developing embryo goes through stages of quantum leaps with transdetermination being the greatest example. Biological continuum applies only to post development stages.
It is exactly a continuum, even if it speeds up at certain intervals. You should study metrology if you don’t understand this.
@@CipherOne perhaps you do not know what is transdetermination. It is more like metamorphosis when catapiller becomes a butterfly, but even more dramatic break down of the continuum. The developmental programs switches completely in almost quantized fashion.
Yikes, this Michael Levin dude is bad news. He's practically barking at this woman, and accusing her of believing in 'magical action', or whatever that cheap smear that determinists aggressively hurl at anyone who disagrees with them. And seems to have a petty and intolerant tone about scientific endeavor in general. How is this guy so hailed? Doesn't sound like he's saying anything profound or insightful, and more likely he's largely a self-promoter....