I find lots of good papers about C. Elegans and Zebrafish models, not sure if there’s a need or location for these in this project.
I can do layman tldr’s for some of these if there’s interest, although the Abstracts are usually the best.
3 Likes
This would be an interesting discussion for research and theory. While Monty has the luxury of using very precise systems to determine location we still want to have a deep understanding of how the brain accomplishes this for the Thousand Brains Theory.
2 Likes
I chose those two because the first is published in a computation and AI journal and the second relates to spatial awareness neurology.
Hey @scidata! We highly encourage you to share some papers from the C. Elegans and Zebrafish models - a layman’s TLDR would be great (or even simply sharing some papers). It seems like the zebrafish paper you linked to applied some of the latest microscopy techniques (lightsheet imaging) on head direction cells, which is pretty cool. The C. elegans paper looks interesting as well - it’ll be interesting to know what theories have been developed for this organism since we have known its complete connectome since mid-1980s.
2 Likes
Human perception of a new color “olo”. Lasers are used on a cell by cell technique. This process aims for “programmable control over individual photoreceptors”.
A close approximation of the color ‘olo’ is provided in this introductory post.
1 Like
Thanks for the update, @scidata!
Incidentally, we at the TBP Team heard of the news some time ago and lots of us expressed that we wanted to “experience” the color Olo! 
1 Like
1 Like
^ That’s super cool! Thanks for sharing the video of the developing nervous system @scidata 
Having read and re-read Max Bennett’s “Brief History of Intelligence” which very convincingly explains the 5 breakthroughs that evolution of neural systems and brains produced over the past 600 million years leading to human intelligence, I just read an additional book by Iain McGilchrist “The Divided Brain” on the known principles of functional lateralization in the two hemispheres of the brain. I think both books should be considered obligatory reading for anyone truely interested in cognitive theory and the foundations of natural intelligence. These two books are very, very enlightning reading and give us the most valuable orientation possible. Evolution always provides insights into both form and function of organisms and nonetheless, brains across time and the animal kingdom.
The divided brain points out that there is a special evolutionary reason why we have two hemispheres across many species. The central reason is not function as was beleived for decades. It is the fact that we all need two distinct kinds of attention. As we all already know since LLMs and other ML models, attention is probably the most fundamental aspect of learning, even in model-free context of ML. Nature has given us two hemispheres, because embracing and understanding the world in a way that optimizes our chances of survival, requires two distinct kinds of attention. The left hemisphere uses a more convolutional approach which focusses on individual elements and features of our perceived reality. It is more reductionistic and looks at the atomic compontents of sensory input, while the right hemisphere focusses on connections of all these elements into a higher order pattern. It recognizes the uniqueness of a particular face belonging to one specific person, for example. Or it recognizes the higher order meaning of a poem or the true intent of a particular statement. The right hemisphere listens to the left hemisphere much more than the other way around. Both communicate. The left focusses on details and features and provides our right hand with fine motor skills. The right hemisphere sees the whole, the context, recognizes a particular melody and also is the seat of empathy and theory of mind. We understand purpose, intent and uniqueness with our right hemisphere. But both hemisphere are participating in all cognitive tasks. So it was wrong to think that language is located only on our left hemisphere. Language has a level of abstraction at word-level with syntax and atomic units of sematical meaning and language also has a higher order in which the actual intention of a text is recognized and common sense is applied to understand the entire picture of observed reality.
Knowing that we need and have two distinct forms of steering attention in parallel, should lead us to new ways in which we model our brain. We need two parallel forms of learning that communicate frequently the way our corpus callosum does at all times in our brains. Firstly, we need to design the two forms of attention at a neuro-computational level. I would like to know if any work is already taking place on these attentional structures and algorithms? This is foundational work that needs to be addressed and laid out, before we can achieve a high performance TBT model. Are there any opinions on this topic in this community? Or does it need more discussion and reflection?
I have a different point of view. I’d argue that lateralization is more of a hyperparameter of the system rather than a core architectural principle. It’s an evolutionary optimization, not a fundamental requirement. A supporting evidence is that individuals with only one functioning hemisphere (due to hemispherectomy early in life, for example) can develop normal or near-normal intelligence.
1 Like
I think lateralization has a lot of potential variance in individuals, because the plasticity of our brains is an even more foundational property of our brains. This allows for huge shifts and repurposing of specific locations for adaptation to traumatic and disruptive anomalies. But this does not mean that lateralization (with two fundamentally different forms of attention) are not essential properties of our cerebral cognitive abilities. Many of our most successful AI model have already incorporated asymmetric adversarial systems within a model.
A very important bit of supporting evidence that lateralization of two parallel forms of attention are essential to all vertebrates, mammals, and primates is that this form of bilateral organization has been shared and inherited across phylum, class, order, family, genus, and species. This evolutionary principle of brain structure and organization is persistent in most populations with higher cognitive activity. Birds control their eyes with two separate hemispheres and focus on different aspects of their environment. One eye allows fine calibration of motor movements, for example pecking at a small grain found in a pile of sand, while the other eye focuses of other more holistic aspects of things like threats.
To summarize, attention mechanisms are fundamental to learning as well as motor control. And hemispheric differentiation emerged early and has persisted in the later animal kingdom. It appears that these two forms of attention create different functional modes of operation. And these hemispheres then act in an adversarial-complementary mode. Together they achieve a much richer representation of reality than they are individually capable of.
2 Likes
I do want to add, thst I am not a Neuroscíentist, even though it has been my favorite subject to read and study for decades. So my opinions are not profesional. I just quote Scientists like McGilchrist and others, but cannot weigh their claims in contrast with others. It just does apear to be evident that our hemisheres work quite differently and this is universal to many species in our evolutionary path. How attention was derermined to be the driving factor for these differences is not clear to me, but seems plausible.
Just like we have clear differences in function from the PFC to the posterior lobes like a spectrum both for cognitive hierachy as well as a spectrum for control being located in the aPFC and perception in the rear parts of the cortex, we also have a left vs. right hemispheric specialization, and types of attention does seem to offer a good explanation for most known differences as related to behavior and trauma cases.
To close this point, I want to paraphrase an important comment from Iain McGilchrist: If lateralization does not play a fundamental role in brain function, then why is it that our brains and those of our ancestors and many other species, have such a prominent structural dicotomy in the first place. In neuroscience it is known that form follows function. And there is always a reason for every structural feature, even much, much less prominent ones.
1 Like
Thanks for your detailed answer. I think our views differ mainly in how we choose to model the brain: top-down vs. bottom-up.
To clarify, I see lateralization not as an explicit design choice, but as an emergent property of a bottom-up, self-organizing multi-agent system, where each of the 100,000+ cortical columns acts as an agent. In this framework, specialization emerges naturally from resource constraints and interaction dynamics, without the need to explicitly prescribe hemispheric roles. In summary, I see lateralization as a consequence, not a prerequisite.
If what you say is correct about those two distinct forms of steering attention in parallel, then I wouldn’t be surprised if individuals who undergo early hemispherectomy still develop both kinds of steering attention within their single functioning hemisphere.
2 Likes
Thanks for your very interesting views. I like your bottom-up versus top-down explanation of where you understand to fit in. I agree that this is the most essential question regarding lateralization, but I am not convinced that lateralization is just a late add-on.
If the statements made by much more recognized researchers than I, such as Iain McGilchrist are true, then your assumption could not be true. The kind of features distinguishing the right hemisphere are too foundational (which means bottom-up.) But I am not experienced nor qualified enough to make that call with any certainty. For me attention is foundational. But I guess attention itself could have differing hierarchical levels. In that case, a higher level of attention encompassing many large groups of CC, could plausibly be a later evolutionary add-on.
But if models of reality are learned within a CC and attention is crucial in that learning (build phase), then this tips the other way. The advent of two hemispheres in vertebrates does seem to be at least a supporting point for the argument that this specialization of hemispheres emerged early on.
1 Like
Just to clarify: I’m not saying that lateralization is a late add-on. It’s clearly a fact of vertebrate brain organization. I’m not challenging that point. From my understanding, my view is completely compatible with functional lateralization early in evolution (to be precise, I would generalize this concept to cortical specialization, which naturally encompasses hemispheric lateralization).
If we aim to model brains in a biologically grounded way, I think the real question is: how does lateralization develop?
- Is it explicitly hardcoded in the developmental blueprint—e.g., “one hemisphere will do X, the other Y”? Is it your view?
- Or does it implicitely emerges via bottom-up self-organization under constraints? (my view)
3 Likes
Thanks for your precise distinctions on which elements and concepts of the lateralization paradigm you believe to be in place.
I also really ike the fundamental question you phrased on how lateralization develops or emerges. I agree that it is the key question to answer. I am also not sure how this process is triggered or initiated. But I am quite sure genetics play a big part, while not the only factor in this case. The fact that other species in our lineage had lateralization is a strong indication that genetics are involved. Functional structures are mostly genetically hardwired., while not ignoring epigenetic factors.
2 Likes
Finding about the TBP project not too long ago, and with a programming background (not neuroscience), I’d like to offer my personal thoughts on the matter.
Focused attention makes sense. So from a programming perspective, when implemented in cod - some cortical columns could focus on some things and other cortical columns could focus on other things.
Doing some quick research it appears lateralization (the specialization of the left and right brain hemispheres) predates mammals and likely evolved hundreds of millions of years ago. So while it’s probably very essential for the brain, I don’t think it’s essential for higher level intelligence and thinking. (Meaning having it can have benefits to intelligence, but NOT having it doesn’t mean intelligence is not possible.)
Also there are well documented cases where an entire cerebral hemisphere was removed (hemispherectomy) in humans, and the person was still able to function remarkably well, especially if the surgery occurred early in life.
So while 2 hemispheres might have extra benefits to intelligence, it doesn’t appear to be an absolutely necessary thing.
Also from a coding perspective, I would not want to limit the whole AI system to just 2 hemispheres. For example, what if I want to build a robot with 5 arms, and 3 eyes? haha Maybe my AI robot would need 3 hemispheres…
So whatever AI system we build I believe it should be flexible.
4 Likes
The back-and-forth between hemispheres has a long history, good to see it discussed here.
“Papert’s Principle: Some of the most crucial steps in mental growth are based not simply on acquiring new skills, but on acquiring new administrative ways to use what one already knows.” - Marvin Minsky
Papert and Minsky wrote “Perceptrons” (1969, updated 1988)
I don’t have a copyright-safe URL for it, but the PDF is out there on the inter-tubes.
1 Like
I find both the order of the timing of the acquisition of lateralization (predating mammals ) as well as the recognition that lateralization is a fundamental administrative difference of available information, very significant. (And also agree it does not make more primitive intelligence impossible, if we have no lateralization in place).
Also the fact, that the loss of a hemisphere early in life (when a single hemisphere can still learn to do the jobs of two hemisheres) is better in cognitive terms than losing a hemisphere later in life, when the functions were already split across both hemispheres, supports the notion that two fundamentally complimentary forms of attention are needed for what we consider to be a healthy cognitive balance.
It must be noted, that hemispheric specialization is not essential for motor control which predates lateralization. Our left hemisphere on its own already masters the fine motoric control of a right hand in right-handed individuals and vice-versa in left-handed ones. So at an HTM-level of motor-sensory control, lateralization is not strictly needed, but for higher cognive “understanding” (disambiguation) of the object at the TBT-level of cognitive voting processes it may well be very fundamental. The left brain sees the elements of say an object or of a body of language (text) information. But it is the right brain that identifies the whole as a specific singular object, or in language it identifies the actual intention of the author of that language, rather than just seeing the literal meaning of that same text in the left-brain. It is only in the typical right brain that we see the singular identity of someone’s face, but in the left-brain we can consciously analyze the features of that face. The left is like Spock in Startrek and the right is like Captain Kirk. The left is reductionist and “conscious” of details. The right is holistic and conscious of specificity and of context. In my opinion this right side of our brains plays a fundamental role in the TBT voting process across the corpus collosum. Both hemispheres collaborate intensively, and as McGilchrist points out, the left hemisphere is barely aware of the right, but the right hemisphere is very aware of the left, because it integrates all the elements of output from the left. If we take this in conjunction with the functional differences of a PFC and the motor-sensory functions of our posterior brain, we can start to map out a more accurate layout of a TBT model.
The right brain may be more bottom-up cognitively and the left brain may be more top-down cognitively (in sequential terms of voting), but both have their control functions in the pre-frontal cortex and their sensory functions in the posterior sections of the brain.
2 Likes
zebrafish sensorimotor decision-making
how neural circuits integrate multiple, potentially conflicting, inputs during decision-making remains poorly understood
study provides a mechanistic brain-wide account of how a vertebrate brain integrates multiple features to drive sensorimotor decisions
They seem to focus on additive and winner-take-all methods of input integration (eventually deciding against the latter). Larval zebrafish move in discrete “bouts”. I’ve noticed that even the adults in many species do the same thing. You can actually watch the step-wise sensor-decision-movement cycle.
They speak very computationally, a fairly common theme in the zebrafish papers I’ve read (that’s largely why I read them). I found the luminance charts and figures difficult (maybe my old eyes). And the discussion of cortical neuron type and function was for neurologists, not programmers.
2 Likes