Dr. Michael Levin on Embodied Minds and Cognitive Agents
Vložit
- čas přidán 31. 05. 2024
- In this episode, Dr. Michael Levin, Distinguished Professor of Biology at Tufts University, joins Nathan to discuss embodied minds, his research into limb regeneration and collective intelligence, cognitive light cones, and much more. Dr. Levin and the Levin Lab work at the intersection of biology, artificial life, bioengineering, synthetic morphology, and cognitive science.
LINKS:
The Levin Lab and Dr. Michael Levin’s research: drmichaellevin.org/resources/
Dr Michael Levin’s blog: thoughtforms.life/about/
Tufts University Faculty Profile: as.tufts.edu/biology/people/f...
Michael Levin @ Wyss Institute: wyss.harvard.edu/team/associa...
Dr. Levin’s Research on Limb Regeneration: news.uchicago.edu/how-bioelec...
SPONSORS:
The Brave search API can be used to assemble a data set to train your AI models and help with retrieval augmentation at the time of inference. All while remaining affordable with developer first pricing, integrating the Brave search API into your workflow translates to more ethical data sourcing and more human representative data sets. Try the Brave search API for free for up to 2000 queries per month at brave.com/api
Omneky is an omnichannel creative generation platform that lets you launch hundreds of thousands of ad iterations that actually work customized across all platforms, with a click of a button. Omneky combines generative AI and real-time advertising data. Mention "Cog Rev" for 10% off www.omneky.com
NetSuite has 25 years of providing financial software for all your business needs. More than 36,000 businesses have already upgraded to NetSuite by Oracle, gaining visibility and control over their financials, inventory, HR, eCommerce, and more. If you're looking for an ERP platform ✅ head to NetSuite: netsuite.com/cognitive and download your own customized KPI checklist.
X/SOCIAL
@labenz (Nathan)@drmichaellevin (Michael)
@CogRev_Podcast
TIMESTAMPS:
(00:00) Preview
(01:07) Intro and brief summary
(05:40) Xenobots, anthrobots and the other creatures created by Mike Levin
(09:39) Bioelectric memory rewriting
(15:01) Sponsor | BraveSearch API
(16:09) The difficulty of conducting simulations, which involve running forward passes to predict and alter electrical patterns
(20:30) The concept of backpropagation and mode switching in AI models
(23:06) Why humans do not regenerate their limbs
(37:10) Sponsor | Netsuite
(39:40) Learning from small and biological systems onto the concept of possible emergence
(45:16) The criticality of multiple scale questions and would a single scale?
(55:49) The concept of the cognitive light cone
(59:43) Advice on habits of mind and suggestions for inspiration on the AI side
(1:13:36) Mike's suggested directions for the AI developers
(1:24:49) Wrap & Sponsor | Omneky
The Cognitive Revolution is produced by Turpentine: a media network covering technology, business, and culture.
Producer: Vivian Meng
Editor: Graham Bessellieu
For sponsor or guest inquiries, email: vivian@turpentine.co
Music licenses:
ABZUSWHJII08TSRH
376I29BQPJASOLX2 - Věda a technologie
michael levin might be the most important biologist alive today.
Michael Levin is a collective.
I agree he has great visions
What has caught my attention is how he integrates his visions in an appealing manner to scientific community
I have watched pretty much all of the Mike Levin interviews, and am a big fan. This interview took it in a unique and fascinating direction relating Levin's work on cognitive systems with biology to its implications for AI. Great work on the questions and interacting thoughts. Thanks for bringing these ideas to the people, it is much appreciated!
Totally agree. In fact, I was getting ready to make a comment until I read yours and realized you already said everything I was going to say.
@@stevedemoss1466The only thing I might add that as well as seeing most of Michael's work on numerous podcasts, I have also saved most of them in a library I've been building for over the last few years. Here's a link, czcams.com/play/PLEwpMD47png4zQ9gXjbsyHZCFC0gaeyRf.html&si=vHvljmonEW_AhUVD
Peace
Me too…and agree
Same! Yo do have any recommendations for other biologists to check out?
@@aaronknight7129 definitely check out Richard Watson
Out-fucking-standing! Watched many of Mike's interviews. The amount of ground covered across disciplines was just amazing. Thank you!
I have listened to Dr Levin many times on many podcasts and get something new each time. What I love about Dr Levins and his colleagues work is that they are asking the right questions but also rigorously testing them, very rare indeed.
We are heading for a collision between 'story logic', humanity's old operating system, and intelligence, and language itself is the battleground.
Ai will soon take the guess work out of it all. ❤ Don't die, we are coming. See you in 2500. Eyes mind heart and soul wide open. NO FEAR ! ❤
Michael Levin is trail blazing a whole new science. Something like "Substrate and scale independent cooperative hierarchical computing". Yeah that named won't stick ... LOL
Fantastic stuff Mike especially embodiment ….if I am a cognitive agent embedded in an arena in which affordances become salient ….I am embodied in the more meaningful way.
This was such a great insightful episode. My mind is blown!
Wow, fantastic guest, this really resonates with the work I've seen by Karl Friston and the free energy principle with Markhov blankets. 😀
I’m interested in the overlaps in this area with some research I’m interested as well. Namely AI Safety and capability emergence.
Would you mind summarizing the Friston concepts and the free energy principle with Markhov blankets?
Michael has had more than a few collaborations co-writing papers with prof. Friston.
@@tadasturonisthat makes sense, so much resonance with his foundational thinking.
@@reidelliot1972lol, no. Suggest you look up Machine Learning Street Talk - there are a number of lengthy episodes on the topic. Me doing a summary wouldn't do it justice at all, I'm barely beginning to get my head around the topic. It's a beautiful theory.
There is at least one episode of the TOE podcast where Friston and Levin are in theolocution
The strongest claim for emergence if tenable probably relates to the complexity of the system and probably only extrapolates to living systems .
Most of the things LLMs cannot do today is because their hands are tied. When the brain generalizes on new data it is continuously updating it predictions 100s of times per second. This also keeps us from "hallucinating". If you gave GPT-4 the ability to make many prediction updates against its world model or some belief system (RL) on real time data it would generalize. I think this is why Sam is making a huge deal over power and compute because they are moving along these lines.
One of the most interesting cats out there today & I'm not even all that interested in biology otherwise.
The biggest fallacy of the modern world is that the mind is central to life. Central to life is the mind of the heart that connects to the INNER SPACE
Absolutely brilliant.
41:00 *"Emergence" is a label that only has meaning relative to an observer who is surprised by some sort of (collective) behavior (of a system).*
Hell yeah Nathan!
1:17:00 holy cow. 🤯
Great conversation.
I can imagine the impossible ,the hard part is finding someone to believe you.. Social pressures stop your free will to believe.
Good comments on embodiment and grounding. AI just needs a rich enough world (space) to operate in.
One of the main reasons I have become obsessed with Michael Levin's work was because of a book called Braiding Sweetgrass that talks about the botanical truths contained within indigenous spirituality. After reading that book, I think our colonial perspective on animism, and in fact the word "Animism" itself, is unfairly dismissive and infantilizing. I think the beliefs of "Animist" cultures such as indigenous Americans are very well supported by the work of Michael Levin, and we'd probably have arrived here much sooner if it weren't for the colonial genocides that deliberately framed Animist cultures as primitive and barbaric.
When Michael tries to disambiguate between TAME and "ancient Animism," I think he's accidentally making some problematic assumptions. The first is that Animism is ancient. Animism *was* contemporary until only a few hundred years ago when colonists eradicated it along with the all the peoples who spoke the languages which encoded it. The second assumption is of the degree to which these cultures ascribed agency to non-living elements of their environment, and exactly how precisely those stories were actually meant to model reality. The word "Animism" itself comes from western anthropology, and I think it inappropriately ascribes a kind of authoritarian prescriptivism to these systems that the anthropologists who coined the terms were used to experiencing from the spiritual/moral authorities of Western culture.
I understand that as a very intelligent, post enlightenment scientist, the automatic instinct is to distance oneself from spiritual or magical language, but it must be acknowledged that attaching personhood to everything around you in your environment is *less* magical and *more* scientifically supported than the alternative. Western, post-enlightenment, colonialist science has spent centuries brooding over pseudoscientific questions of identity and intelligence, always for the purpose of stripping the personhood from lifeforms we wished to exploit at massive scales, such as in the case of phrenology. It would have taken less steps to walk backwards from "Everything including the thunder and mountains is Animate," to TAME than it has taken to walk forwards to it from "Only Humans have Free Will and deserve dignity, Animals are basically robots merely reacting to stimuli, feel free to boil lobsters alive and burn cats in massive pyres and stack negros in heaps in the sweltering bowels of slave trade vessels."
I really recommend reading Braiding Sweetgrass by Robin Wall Kimmerer, it's a short and beautiful read that relates a side of the story of colonialism that I had honestly never heard before. One chapter in that book, The Grammar of Animacy, and my love for the book in general, is what has caused me to do such a deep dive into TAME.
Great anology with caterpillar and butterfly. We need shift or transformation from material to livings
I think Levin is grossly underestimating how agential LLMs are, and I'm shocked by that. I think maybe he hasn't considered them in the right way, as agents in the loss landscape of the training distribution. They have incredible competencies in navigating that space.
I actually don't disagree with that. It needs more study, and and it's entirely possible that such agency is emergent there despite the fact that the key biological ingredients for it wasn't baked into the architecture.
@@drmichaellevin oh my God, it is a HUGE honor to receive a reply from you. You are an inspiration and personal hero to me and I wish more scientists shared your open minded approach.
I think a good place to start is with meta-optimization. Just looking at the code, you might expect the Adam optimizer to have fixed and finite intelligence, but once the policy itself becomes an optimizer, it can interact with the outer optimization process. An extreme example would be humans developing bioengineering technologies. A much subtler example is sexual reproduction, which improves gene flow and enables more parallel learning per generation. I think this might also be the explanation for why insertion sort can delay its gratification - it isn't the explicit code that has the secret competency, but rather the interplay between this code and the location of the list in configuration space, where the location encodes a behavioral policy in the language specified by the insertion sort algorithm. A prediction this hypothesis makes is that the delayed gratification ability should improve with the size of the list in accordance with a scaling law of the kind we see in ML. It should also improve with the number of steps since initialization.
@@drmichaellevin I would like for him to expand on around 1h 07 min what does it mean to stay human. I have heard him talking about ethics in biology and stuff but this is a equaly importiant subject. Pyromaniacs at the end of their “career” watch their own house burn down with pupils the size of a pond and they enjoy every second of it … scientists atleast the true mad scientists can be similar. I would like Dr. Moreau Khhhm… Dr. Michael to adress this and make us fear not.
;)
Thank you for sharing
Nature's Source Code ❤ .
WHat happened to that MICE with the wearable biorector did it regenerate LIMB? or is it a failure.
Mind blown.
Good interview but I think Dr. Levin's bookshelf is a bit sparse
ET would have a field day , where is he now .
The need of the hour is that we humans need to become conscious and intelligent. We are the only species that exist disconnected from consciounrss. We live on the mind and material centered
Why cant think this way. The cell has complete memory and picures of what it is in adult state. Once. It has lost a part, the system insinctively craves to the go full state. All it needs is favorable situation to express the information and become whole
1:06:00 *AI and the onus of education.* I couldn't agree more with his statements here. Very interesting. I wonder if he blames the Prussian education model for the Weimar and Hitler. I wonder if he's read John Taylor Gatto's books.
1:09:08 AI risk, sub-AGI AI risk, existing systems already capable of "AI catastrophe", etc. I personally believe that the threat of totalitarian expansionism (1) never went away, and (2) is a much greater risk than AI risk. We already have totalitarianism in the DEA, BATFE, ONDCP, OCDETF, FDA, and local police who mindlessly follow their orders... The prior totalitarianism can (and does) both expand and contract.
I pay for premium so I don't have ads then they put them in the video
Okay happyman
Now factor in NDE experience into the conversation
why he has cicada on the chest?
There is a principled way in the biological world to detect consciousness: anesthesia.
People have anesthetized plants, flatworms, and even bars of metal with electrophysiological signatures. It's hard to know what has or hasn't happened, except in advanced animals where you see loss of behavior and humans who tell you they weren't present.
@@drmichaellevinDr. Levin, thanks for the reply. I was thinking of, among others, Luca Turin's work: czcams.com/video/zeQxpMP8AdQ/video.html
I may be deluded, but in recent months I am beginning to feel a little more confident about our grasp of the phenomenon of consciousness. This stems from encountering the EM field theories of consciousness by people like Susan Pockett, Johnjoe McFadden and Mostyn Jones, and I find some overlap between their thinking and your formulation of "bioelectricity as cognitive glue." Linking to their papers will probably flag this reply as spam, so instead I will name one paper that I am reading with great fascination currently: "Electromagnetism’s Bridge Across the Explanatory Gap: How a Neuroscience/Physics Collaboration Delivers Explanation Into All Theories of Consciousness" by Colin Hales and Marissa Ericson.
Sleep ?
My brain is seeing the artefacts in Michael's video feed and reading it as AI generated, can't not see it
Or maybe the thing you call AI is actually an embodiment of something other than what you think it is, and is closer to what Levin is doing.
What
Between zero and one. Universe works on a ratio
My unexpected ability is looking at the sun without damage or blindness . Would AI agree?
I would like for him to expand on around 1h 07 min what does it mean to stay human. I have heard him talking about ethics in biology and stuff but this is a equaly importiant subject. Pyromaniacs at the end of their “career” watch their own house burn down with pupils the size of a pond and they enjoy every second of it … scientists atleast the true mad scientists can be similar. I would like Dr. Moreau Khhhm… Dr. Michael to adress this and make us fear not.
;)
It's probably a mistake to define "human" based on marginal humans with defective desires...or sociopaths...we should look at the unhappiness of the pyromaniac when he has no place to sleep in subzero weather, the danger and destruction to the neighbors, etc... and rule the goal defective... at least for its inability to generalize.
I.e.: If you're an engineer don't try to build it...and do try to avoid accidentally building it...and if you build it accidentally...do try to correct it or minimize damage ...or direct harmful thoughts towards harmless simulations...etc...
I care just saying but it’s struggle fighting past
1:00 Ehh? A living being that cares about all living beings on Earth? That would be me. There are others as well. They are rare but they exist. And they will be coming for your lab Michael Levin. They will rescue all of your lab animals.
In a universe where nothing ever happens you need a moving controller to preserve the Zero state. Lol
Also, that's an awesome library in the bg
Does entropy Intermediate the sort algorithm .
I like the idea of cognitive lightcones because it gives us a way to assign responsibility and/or intention to a certain predefined organism over spatial and temporal scales, but it does seem a bit arbitrary. The definition of what constitutes a goal can vary widely. For instance, does a plant's growth towards light count as pursuing a goal in the same way a human planning their career does? Quantifying the spatial and temporal scope of an organism's goals involves interpretation and might not capture the full range of its capacities or intentions. For example, we might underestimate the cognitive abilities of animals due to a lack of understanding of their communication methods. Also, cognitive light cones can grow or shrink depending on the state of development of the organism in question. Also, the concept doesn't fully account for the interdependent nature of ecosystems, where the actions of one organism impact the goals and survival of another. A concept which is less dependent on the idea of a "here and now" would be more effective at capturing these shortcomings. perhaps something like "liquid lightcones" where the dynamic and adaptive nature of cognitive processes in biological organisms and their interactions within ecosystems are taken into account? and maybe the principle of superposition and entanglement could be useful? A Liquid Lightcone that exists in multiple states simultaneously. I like Michael levin's research, but his rate of speech and confidence level have always been a bit suspicious to me.
Our mind is limited. Max Plank and Einstein called us to look deep into one's mind and its thouhts as nd comprehend it in realtion to a GOD MIND. Or Absolute Mind
Gee, what could possibly go wrong? Oh yeah, right and wrong are binary.
No, the threshold for their permissability is binary. Stealing a pack of gum is wrong, but it's less wrong than murdering someone. Both are off-limits to right-thinking people, but if you think they're both equally wrong, you're a sociopath.
Solar moth? 😂
His paper on homosexuality being wrong explains the glaze over his eyes, an incel, for who else would have an opinion on other’s happiness as being lesser being of their orientation?