A new model of intelligence. How the principles of genetics, medicine and neural networks have been shaken
The proposed concept literally walks the line, offering ideas as disruptive and philosophically profound as possible in order to remain within the bounds of scientific research. Michael Levine's work on cellular intelligence, bioelectrical communication, and what a model of intelligence might look like is fundamentally shaking up everything we knew about the brain and consciousness.
Disclaimer: here I often post articles about the brain, biotechnology, and philosophical problems. This article is a free and adaptive translation of a rather interesting and bulk material.
Disclaimer two: The article will often contain clippings from interviews. It’s too eye-opening to design canvases of text as quotes. Therefore, before the direct speech it will be stated that this is the speech of the author of the research.
What is so attractive about this model of intelligence?
This article will present some wild ideas, so it's important to understand who Michael Levin is. And that he's not a fringe or a crackpot. Michael Levin is the director of the Allen Discovery Center at Tufts University, the Tufts Center for Regenerative and Developmental Biology, and co-director of the Institute for Computationally Engineered Organisms. He co-author of several hundred research paperswhich has been cited more than 30,000 times, and the list of outstanding awards is quite impressive.
All this serves as a kind of guideline that should be kept in sight. After all, this review will be very unusual.
It is also important to note that Levin remains one of the leading researchers in a broad and expanding field. This is rather a visionary piece, offering a rather original view of the development of a huge scientific layer.
Cellular Intelligence as a New Model of Intelligence: Redefining the Understanding of DNA
There is a certain general idea about DNA. It comes down to the fact that DNA is the model by which our bodies are built. A kind of compressed set of instructions that guides stem cells. According to the same instructions, cells understand where to grow and what to become in order to assemble a full-fledged organism. IN interview with the “head of TED” Chris Anderson in 2021 Levine explains one problem with this concept.
Below is Michael Levin's direct speech.
When a tadpole becomes a frog, it has to change the shape of its snout. The eyes, the nostrils, the jaws, all these elements have to move. It used to be thought that the body had some kind of hard-coded set of sequences. According to them, all these elements move, and eventually the tadpole becomes a frog. But a few years ago, we discovered a rather surprising contradiction.
In the scientific community, there is such a phenomenon as “Picasso’s tadpole” – these are tadpoles whose jaws, eyes, nostrils can be shifted to the side… That is, initially all the elements are “out of place.” Eventually, these tadpoles grow and acquire regular frog faces. Organs may initially be in the wrong position, but they still end up folding in the normal order.
So this system, like many living systems, is not reduced to a set of programmed actions. Its operating principle can be described as: the desire to reduce the error between what is happening now and what is the “correct configuration” of the frog's face. This type of decision-making involves flexible response to new circumstances. In another context, we would call this a new kind of intelligence.
Mini conclusion. “Picasso tadpoles” with distorted facial features successfully remodel themselves into normal frogs, demonstrating a flexible type of “intelligence” for solving problems. This is fundamentally different from the idea of a rigid set of instructions embedded in DNA. But it partly correlates with genetic development of cognitive skills.
QuestionSo if DNA doesn't communicate with these cells through biochemical signals, how do the cells know what stage of development the body is at and what to do next?
Bioelectricity as a way to power the model of intelligence
Cells definitely communicate biochemically and through physical forces. But something else extremely curious is happening. Essentially, this can be called bioelectricity. Moreover non-neuronal bioelectricity. It turns out that all cells—not just neurons, but all the cells in your body—communicate with each other using electrical signals.
Michael Levin
Levin and his team began tracking electrical signals in frog embryos as they developed. In the process, they found that as soon as the cells of the new organism began dividing, they immediately formed electrical networks. Long-term observation led to a bold conclusion:
DNA can no longer be considered the software or instructions that cells carry out to create our body. It's more like the hardware that powers the intelligent system.
People don't like computer analogies in biology. They're not perfect. But I think this analogy gets the point across pretty well. What the genome does, how it codes for cells, is the hardware. The genome tells each cell what microscopic hardware it has to work with. And the hardware is taken over by the proteins that the cell has. And everything that happens after that is software.
Moreover, this equipment can be reprogrammed. The genome does not directly determine your shape. It does not form the contents of memories within your body's networks. It gives us amazing hardware that does some things by default, right out of the box. But it is also easy to reprogram.
Cells form electrical networks, and these networks process information, including memories of patterns. They include seeing large-scale anatomical structures, where the various organs will be located, where the different axes of the animal will be located… The knowledge that there will be a head and a tail in front and behind is literally held in the electrical circuits of large tissues in the same way that the brain stores other types of memories and learning.
Based on interview materials from 2021
The point is that this structure of electrical connections can be hacked. Literal biohacking.
Experiments with multi-headed flatworms
You can take a broader view and give this intercellular network stimuli or inputs, just as you would reprogram a computer. What you can do is make this network do something that it doesn’t do at all. These bioelectric signals give us an entry point directly into the software that controls the body’s anatomy at scale, and that offers a very different approach to medicine than trying to change the processes inside each cell.
Michael Levin
For the demonstration, Levin and his team chose a planarian, a flatworm that lives in fresh water and has the remarkable ability to completely regenerate severed body parts. These creatures have an entire brain, as well as many internal organs. And you can cut the worm into a number of pieces, and each piece will restore everything that it lacks to form a complete organism.
This is a system in which each particle knows what the whole planarian looks like. And it can restore the right organs in the right places, and then slows down this process.
Michael Levin
Levin and his team cut off the heads and tails of these flatworms, measured the electrical gradient between the heads and tails, and ran several experiments to see whether manipulating that electrical gradient would cause the cellular networks to build the wrong parts.
An explanation of how this model of intelligence is reprogrammed
This entire section is direct speech from Michael Levine.
We don't use electricity. We use ion channel proteins that every cell uses to tune its electrical state. It's like turning tiny transistors on and off. They can switch this circuit into a state that says “make two heads” or “make no heads at all.” And what you see here are actual worms that either have two heads or none. And that's because the electrical map, what the cells use to make decisions, has been changed.
These worms are based on a crazy experiment. You take one of these two-headed worms and chop off both heads, leaving only the normal middle fragment. It is extremely important to understand the fact that these animals were not genetically edited. Their genomic sequence is completely wild type.
The standard paradigm says that if you:
Got rid of that extra fabric.
The genome was edited.
Created optimal conditions for growth.
Then in the end you will grow a completely normal worm. The most amazing thing is that this not happening. These worms, if cut again and again in their usual environment, continue to regenerate as two-headed. But this is not the end.
Normal, headless and four-headed planarians are all viable. The memory patterns by which these animals regenerate after damage have been rewritten. And in fact, we can rewrite it again and return them to single-headedness without any intervention in the genome.
Mini conclusion. The information structure that tells worms how many heads they should have is not in the genome, but in an additional bioelectrical layer.
Now we have the ability to rewrite other factors. And this, by the way, is also a key definition of memory as an element of intelligence. Memory must be long-term stable and rewriteable. And we are now starting to crack this morphogenetic code to ask how these tissues store a map of what to do, and how can we go and rewrite this instruction for new results? What will be born as a result?
Birth of the Xenomorphs
Xenomorphs may seem like something distant and cosmic, but they are already being used in laboratories. Here first example works by Michael Levin and his colleagues. Xenobots — groups of cells of one organism with the possibility self-organize into entirely new beingswhich use cellular intelligence and bioelectrical communication to develop unique behaviors.
Don't want to create them yourself? No problem, these same cells can be design and program from the outside.
Organ regeneration and a new model of intelligence
What's also great is that hacking organisms using this approach may actually be much easier than it seems. We don't need to understand how to build a hand, an eye, or a brain at the molecular level. Cellular intelligence already knows what to do. We just need to trigger the organ's construction at the macro level.
Levin gives the example of the construction of eyes in the intestines of tadpoles. Below is the author's direct speech:
By running eye-building routines on the body's “physiological software,” you can very, very easily tell it to build a complex organ. This is important for the fields of biomedicine, because we ourselves do not know how to build eyes at the micro level.
Targeted changes in voltage in ion channels in frog embryos cause the formation of completely ectopic eyes in unusual locations, such as the intestine.
Direct speech from Michael Levin
I think it will be a long time before we can grow functional human eyes, hands, and other organs in bioreactors or some other exogenous way. But we don't need to, because the body already knows how to do it. These subroutines are triggered by specific electrical patterns that we can detect. This is what we call breaking the bioelectric code. And we can create eyes, we can create extra limbs… We can create extra hearts… We are starting to crack the code to figure out where in this software are the subroutines that we can run.
For example, we can show that in adult frogs, which usually do not regenerate their legs, 24 hours of stimulation, using the methodology, allows them to grow a new pair of legs in a year and a half.
The idea is not to micromanage the process, but to convince the cells that this is what they want to do – and that they have all the competence to do it.
Revolutionary implications in regenerative medicine
If you can regrow any body part you want from scratch, then a new form of medicine is available to you. And with frankly creepy potential. And this potential in the field of cellular intelligence and bioelectrical hacking is enormous.
When you think about it, most of the problems in biomedicine—birth defects, degenerative diseases, aging, traumatic injuries, even cancer—come down to one thing: cells don't build what they're supposed to. If we understood how to interact with these colonies of cells, how to rewrite their target morphology, then we could ensure the disintegration of tumors, correct birth defects, and induce regeneration of limbs and other organs.
This is something we have already done in frog models. The next truly revolutionary step is to transfer this to mammalian cells.
Going Deeper: Molecular Intelligence
Levine's group expanded its focus beyond the study of living cells and found that the patterns of routines that can do some of the work appear to be traceable down to the molecular level.
Direct speech by Michael Levin
All these levels consist of subagents who solve problems in different spaces: anatomical space, physiological space, wherever – and they have different competencies and different plans. Each level uses what I call agent material. Something that allows you to design an organism completely differently than passive or even active matter.
And this even applies to cells. I mean, we study the learning ability of molecular networks. Let alone whole cells, even molecular networks probably have at least six different kinds of learning ability.
It's important for us to have some way of talking about molecular systems as intelligent systems. That will allow us to fully describe the principles of scaling. We all start life as an unfertilized egg and a single sperm. It's a blob of chemistry and physics. And there's no point where you can say, “Okay, you were a bunch of physics, but now you're a real mind.” So we need a paradigm for how intelligence scales and evolves from simpler forms.
The kind of intelligence that I’m talking about is the kind of intelligence that William James defined back in the 1890s as “the ability to pursue the same end by different means.” So it’s really navigational intelligence. It’s a highly observable, highly empirically testable problem-solving ability. I’m not defining it as consciousness, and I’m not talking about self-aware meta-intelligence. I’m talking about the ability to navigate complex, chaotic space and achieve goals despite all the new factors that are happening all around you.
And this may explain our unusual state of mind, stuck in point of critical existence.
An explanation of how intelligence arises at different levels: from molecular to cellular, from the organism level to the swarm level and even the ecosystem level.
TAME: A Technological Approach Explaining the Intelligence Model
At some point, Levin and his team started finding testable examples of flexible “intelligence,” and more and more of them were coming in. So the scientists had to develop an approach to make sure they weren’t seeing a “Deus Ex Machina.”
Direct speech from Michael Levin
Two thoughts regarding our developments.
The first and most important. Having such diversity of intelligence is not a philosophical statement. This is an empirical, testable, experimental statement. If you think that some system has some intelligence, you hypothesize that there are problems in the space, the purpose that drives the system, and the competencies that you think the system has, and with which you are going conduct experiments.
It's not like that, we're not drawing castles in the air, we have very specific hypotheses about the abilities of different kinds of systems to solve problems. It's impossible to simply say that all objects in the world have “souls”. However, you can't prove that there isn't something more in our cells than the cells themselves.
Second thought. When you make a claim that a system has intelligence, you are essentially putting the system through an IQ test because you, as an observer, are saying, “This is what I have observed this system can do.” The problem is that you have already overlooked many other factors.
Consciousness and Artificial Intelligence: Where to Draw the Line
As with many cutting-edge areas of today's wild world of science and technology, Michael Levine's work teeters uncomfortably close to the edge of the unknown and the realms of philosophy and religion. Anyone who's watched the series “Fringe” – let me know in the comments)
Indeed, when it comes to artificial intelligence, where researchers are struggling to understand what exactly separates a sufficiently advanced AI from a conscious being, Levin's work approaches the same barrier, but from the opposite side.
Levin's TAME framework can evaluate the latest advances in AI and machine learning. According to the scientist, his team has some arguments about consciousness and similar patterns. Moreover, scientists cannot prove that there are not yet machines that have a significant – in some cases, human-like degree of operational intelligence. But this answer lies rather in the area of cyberpsychology.
In the magazine materials Frontiers in Systems NeuroscienceLevin argues for a “more inclusive structure of knowledge.”
Direct speech from Michael Levin
Whatever the differences between us and some future AI architecture, the result will not be the same as we see it. Here are some bad definitions:
“It’s just a machine, it operates according to the laws of physics and chemistry.” Well, it's the same as you and I.
“I know what it is because it’s linear algebra, and I’ve derived the algorithms for it.” But we find properties like learning and memory in systems as simple as a few genes turning each other on and off. A network of differential equations representing genes turning each other on and off – let alone an entire cell, let alone an entire genome – is nothing more. And the first biocoder may have been Ivan Pavlov.
We discovered unexpected possibilities and solutions to problems in such a stupid sorting algorithm as bubble sort. Strict determinants: six lines of code, nowhere to hide, no magic. But if you look at them the right way, you'll discover things you didn't even know existed. Something that is literally not in the algorithm, but that it is capable of.
Indeed, Levin decided to draw his own line between living cellular intelligence and artificial intelligence.
A few months ago, I started writing a paper that laid out very clearly half a dozen factors in biology that are really important for making real agents that matter in a moral sense. Like, “Here's what biology does that none of our computer architectures do.”
And I stopped. And I'm not going to write this article. Not that it will help – sooner or later someone else will do it. But I don't want to be responsible for this nonsense. To whatever extent I may be right, there is a huge possibility that there are features that I think give rise to the existence of true intelligent beings that we would need to take care of… In that sense, I don't want to be responsible for creating trillions of them and have no control over how they are treated.
What's it worth? supercomputer-based consciousness or a neural network built on based on living parts of the brain.
New forms of intelligent life?
Here we come to a key point: is it possible to create new forms of conscious, intelligent life using the ideas and methods that the author and his colleagues have developed. And here Levin is clear.
Direct speech from Michael Levin
I think it's absolutely possible. The idea that something special in the mind can only be created by a blind, inventive agent that generates random mutations and selects specific sequences? I don't see why this process should have a monopoly on creating “real intelligence”.
I know there are people who disagree with this. Richard Watson is one of them. But I think there are many ways to prove I'm right. I think we already have a good basis for figuring out what the actual methodologies and components are that we need. And they have nothing to do with products that were born from protoplasm, or from any other elements that we assume are associated with biology. I'm absolutely sure they can be implemented in other environments.
When I lecture on this topic to ordinary, casual audiences in traditional fields of science, much of the material is something that the audience has never heard of, or something that seems completely wrong on a philosophical level. So I'm not sure where we exactly are in the transition between “it's impossible” to “it's completely obvious.” I think this is the path we're on, but I'm not sure where exactly we are… Assuming we all live long enough, this discovery will change everything we know. And so it should be.
Our future is freedom of embodiment. In the future, I see our children being told stories about the past saying, “You've got to be kidding me. Do you mean that someone was born, and simply due to the vagaries of some cosmic ray, a glitch in the DNA, he had to stay and die in the body he was born in? Maybe their goal was a higher IQ or a longer life, but no, they got lower back pain and astigmatism and then died at 70? This can't be true. Nobody can live like this!”
This is the future I see: and from that position our current situation becomes ridiculous. And that's how it should be. It's really funny.
This is a very strange, promising and complex material. Some of its elements seem fantastic, but visual experiments prove the opposite. It is unclear what such technological development may lead to. Perhaps we are already standing not at the threshold of technological singularity, but literally at its beginning. And we do not realize it yet.
As always, more materials on the topic of the brain, consciousness, controversial areas and boundaries between the nature of the human body, consciousness and something more – comes out in the telegram channel. Subscribe so you don't miss the next materials!
