How far neuroscience is from understanding brains (2023)
pmc.ncbi.nlm.nih.gov
pmc.ncbi.nlm.nih.gov
I started a PhD in 2017 studying neuroscience + ML. I thought studying the brain would help me understand ANNs better. I was wrong. Ended up applying ML to analyzing EEG, MRI and similar.
But as many comments here have pointed out, studying neuroscience, for example, usually makes those analogies seem painfully inadequate. The same is true in philosophy of mind, for example.
Alas, state of the art in neuroscience / neural engineering is closer to bloodletting than a mechanistic theory of learning and memory.
https://www.slideshare.net/slideshow/georgia-state-universit...
Thanks for the read.
C. Elegans because every adult has the same number of cells in the same configuration.
Do those worm have no genetical variation? Even then how is this possible, surely the cell count varies over time as some cells die and others multiply?
It’s the only theory of how the brain works that I’ve come across that seems like it could be valid. Unfortunately, like other posters have already mentioned, neuroscience is incredibly complex and we just don’t have the tools to test it.
Even if it ends up being completely wrong, it’s a beautiful theory and well worth checking out!
I think the clinical shift to focus on 'the individual' instead of the ensemble has overlooked the importance of inter-brain synchronisation, the process of which we are only start to grapple [0]. A feeling all to familiar to fellow musicians, and no doubt anyone involved in activities in which timing coordination is key.
[0]: https://www.sciencedirect.com/science/article/abs/pii/S01497...
Love this.
> I think the clinical shift to focus on 'the individual' instead of the ensemble has overlooked the importance of inter-brain synchronisation
My sense is this is also the case for our cultural story of neuroscience. We can run massive analysis of the brain and what brain regions or neurons or transmitters are being used/activated and it all boils down to something we can't conceptualize. Even the principal components seem to be too complex for us to grock.
With this barrage of complexity I feel like we have a (bad but understandable) habit of reducing to the individual. We talk about the grandmother neuron or what region x is responsible for. We harp on the cog sci of individuals and we run experiments where we try to isolate them for "confounding factors" (other people).
I keep coming back to evolutionary theory as the middle ground between this over and under simplification of cognition.
On a deeper level, I also think the prior discrepency stems from how we perceive the world as individuals rather than a collective, our research methods implicitly reflecting this bias. In regards to synchronisation, you might find the following interesting:
"Brain Waves Synchronize when People Interact" [0]: https://www.scientificamerican.com/article/brain-waves-synch...
> Even if it ends up being completely wrong, it’s a beautiful theory
Not just that, but I think it's vitally important to be able to demonstrate how much we can potentially explain. There are some people that will never accept a physicalist model, but I suspect there are many who will happily accept it once they conceive of the possibilities. At least, that's the process I went through when originally learning about evolution by natural selection.
On a high level, I think Calvin’s theory is in the exact same vein as Dennett’s. Seems like we could sum them both up by saying the brain is paralleled and there is competition between thoughts?
What grabs me about Calvin’s work is he gets much closer to how this might work than anything else I’ve seen. Even cooler, I think he provides enough depth that we could potentially design a new architecture around his theory.
Still I can’t recommend it highly enough. Act I is enough to grock the concept and you can crush it in an evening.
Psychology is worse, much worse, and yet is used around the world to make many, many life-changing or life-ending decisions.
To me as a layman, getting interested in what makes people tick, how different issues manifest, how bad childhood and especially missing good father figure affects people proved invaluable.
But I dont mean clinical psychology, university textbooks etc but rather easy to digest popular stuff from ie Jordan Peterson (before he got addicted and crazy).
Understanding (ex)girlfriends, colleagues, parents, politicians, everybody and especially myself. People cant shock or surprise me these days.
Because that part is wrong, apart from the underlying inner working being dissected in labs (the cognitive science part), the therapy part it is well studied in the wild - and it works.
> and the results of therapy I come across always seem extremely underwhelming.
So you mean anecdotes? Of course you recognize the underwhelming results - the people where it worked well are less noticable, by definition.
> Maybe for some cases like severe phobias there is something to be solved, but do we really need to be paying thousands of dollars a year
That you have to pay money has been a decision of the society you reside in. It doesn't have to be this way, and in fact it isn't in most industrialized countries.
> for a therapist to say "hey, maybe if you'd invest more in hobbies, social relationships and a healthy lifestyle you would feel better"?
That is a strawman, not CBT.
"Just saying" the above is not what CBT is, though. I've had a lot of CBT and other kinds of therapy, and I've never had a therapist just tell me "so don't do that", or "do more x".
It's about analysing your own thinking(cognitive) and actions/habits(behavioral) in cases where those are keeping you stuck in a miserable situation, and identifying what you can do to change them.
Sure, if you put the most inanely over-abstracted construction on it like "just live healthier and you'll feel better", it sounds meaningless, but the implicit assumption there is that's something you can just decide to do, and succeed in. That's not the case for many people, and those people can benefit from CBT.
And peoples lives can also vary in quality. If you're raised in a nest of narcissists it is likely you'll have a very skewed view on relationships and your idea of normal could be considerably different from the average and have negative effects on your social relationships. Sometimes you just need a 3rd party to tell you that's not normal.
Coming from a family of narcissists myself they quite adamantly stuck to the idea that therapy was a total scam. Though my observing their behaviors it is my believe they are really saying "Therapy is a total scam because they didn't tell me what I wanted to hear".
As somebody who studies psychology.. If a psychological study says A, it’s more likely to be actually B. The quality of research is much, much worse than people think.
Again, of course, if you are a well adjusted adult, and no one in you life is clamoring for you to get therapy, then chances are, either you've pushed anyone and everyone away that cares for you that deeply, or you don't need it, both of which are totally possible.
As for if it's a scam? That word gets thrown around too easily these days. Do they cost a lot? Yes. Are you worth it? Also yes. That's not to say all therapist are equal, and someone that's great for you might be terrible for the next person, and some of them are just bad. (There are also those that are quite good!) As there are with all things. Going into therapy with specific goals on things you want to work on (eg irrational anger at some specific thing that's ruining your life) tends to help it not be just an ongoing extra expense like a gym membership, if that's not what you want. At the end of the day, they're stuck in the system as much as you are, and paying them also makes it feel more balanced in that they're being paid listen to you go on and on and help you with your ish. You wouldn't/shouldn't treat your friends so one-sidedly. That said, it's entirely possible to pay to see a therapist and not get anything out of it. I will admit to that as a possible failure mode, but it's a failure mode and there are certainly success stories out there of people turning their life around and becoming happy well-adjusted adults.
So somewhere there is evidently a lot of knowledge about effective, working psychology, even if that knowledge is used for evil rather than good...
Most of the social "science" are a scam, but since it gives peoples some "work" it keeps going. In reality it's only value extraction and they are basically parasites of the common people but since the people in that field have a lot of power ("educated", often female, etc) it keeps going. It's like the HR department; you would rather not have them but since there are laws around it you pay the racket and move on, not other way.
What I mean by this is that we thought we understood the body when we understood thermodynamics. The blood pumps through the heart, goes to the brain, nutrients feed the muscles, etc etc. Then we discovered how electricity works and we thought "oh! of course, it's all electric!" The heart pumps because of electricity, and the nerves are all electric synapses, etc etc.
I fully expect in the future we will look at quantum mechanics and apply that to our understanding of biology (if we aren't doing that yet), and down the rabbit hole we will continue.
That doesn't mean we can't do some incredible things with our current understanding of the brain, I work in neurotech at https://affectablesleep.com, where we are increasing the effectiveness of deep sleep. We know the mechanism we can use to trick the brain to increasing the synchronous firing of neurons which define deep sleep, but we don't know WHY it works.
Heisenberg would like to have a word with you. He wrote Reality and its Order during the war (world war 2 that is) and obviously he knew his quantum mechanics.
"From the point of view of present atomic theory, the organism appears to be an utterly improbable formation".
I can't tell if you are disagreeing with my initial comment or agreeing. Sorry for being a bit slow on the uptake on this.
Heisenberg begins with an abstract consideration of language and correspondence of expressions in a language to a given 'context' or domain of understanding. He then uses this framework to progressively consider distinct domains or contexts of knowledge (celestial mechanics, EM, QM, Chemistry, Biology ..) by noting the content of their respective languages and their relationship to one another. For example, Newtonian mechanics has the vocabulary of force, mass, acceleration, etc. Then he notes how Electro-Magnetism necessitated the introduction of new terms, most importantly (force) fields. He then notes, as an example, how 19th century mechanics (alone) could not explain the stability of the atomic ensemble, which is what gave birth to QM.
In all this, his main point is that each of these domains are 'closed' and complete in terms of phenomena that they are expressing, but are not universally complete. For example, basic (Newtonian) mechanics can fully explain the motion of an automobile, but if you want to understand what is happening with say the brakepads on a car when it is decelerating, we need to introduce the langauge of 'heat' and effectively Chemistry.
In terms of "organic matter", he muses on the fact that there is clearly something different about "living matter" and "dead matter", yet as of then (or now) we see that the same physical laws (chemical and QM) are operable in both modalities of matter, yet clearly something is different. He refers to the then current theories of vitalism and entelechy and also the thoughts of his mentor (Bohr) who apparently projected his overarching 'guiding thought' of "complementarity" to the question of living matter. At some point he also considers the possiblity that all 'organic matter' and "consciousness" may somehow be related.
As to "improbable", he takes pain to point out distinction between "objective" and "subjective" understanding of a system or phenomena. (This is likely due to his following in the footsteps of Goethe, whom he references.) QM can give "very precise" predictions regarding "a specific aspect" of a (complex) system, say momentum, but the other aspects of the system can only be defined statistically.
In all this, what is clear is that as of 1940-whatever neither Heisenberg nor Bohr thought QM as a language that is 'closed' over organic matter, to say nothing of consciousness.
Hope this clarifies.
That was quite some years ago already, mind you.
I've read this few years ago, how far we've come?
There are very few non-destructive ways we can get a look into a brain while it is living. And even research into ways we can look into a living brain is hindered by the fact we don't want to harm the person being observed.
I could see that making progress a lot slower
Animal brains are harmed for research all the time. Some species, like Zebrafish can be even imaged real-time at neuronal level.
As the article says, the problem is really that there are no good foundational ideas of what are the basic operating principles of the brain.
People are important because we can't tell a fish to perform an action or to think of an elephant, etc. So while we can image, we can't know what correlates with a high degree of certainty.
But even in other cases, a working brain is an entirely different object than a non-working one. Finding those foundational ideas and basic operating principles because we can't test a living brain. It's almost chicken/egg. Being able to experiment on a living human brain would probably give us much better ideas of the basic operating principles. And once we know those principles, we could probably move from experimenting on living human brains.
Like, we can dick around with muscle. We can force it to perform as it were living and receiving signals. And that's partly because you can cut open a living person and watch those muscles operate. You can manipulate those muscles in a living person. You can measure things, etc. It's much harder to do that with a brain. Because it's incredibly easy to make irredeemable mistakes.
I don't think I agree just on this piece, though you are making important points here. We can know, for instance, of visual or spacial awareness, reactions to certain stimuli or objects introduced in their environment, and so on.
You're not wrong on the importance of learning from active brain functions but I think we can at least make meaningful inroads with animals.
> Finding those foundational ideas and basic operating principles
I know a lot of people talk this way about brains and consciousness, that we need to first have our definitional elements before we proceed with research. But I think we often have, as the object of research, closing in on those elements, while gathering all kinds of functional and theoretical data from the mass of complicated noise that is brain structure and activity. We can build out correlations, mechanisms, and so on and close in on foundational elements.
It may even be the case that foundational organizing principles turn our understanding inside out in important ways, but I have never understood the idea that we are stuck if we can't start with such things in hand.
I'm talking about very specific correlation. Like when I type these words, exactly how and what fires to achieve that. Which neurons and pathways are the thought "hit these keys", what's the signal to the fingers etc. We know "When this area is active, this group of actions can happen".
It's hard to get good resolution.
And I believe it would be easier to replicate consciousness if we knew how it worked, I also believe it's possible we eventually figure it out. It will just be much harder. And we may not know when we do it either. I believe that is where we will need common, agreed upon definitions.
But defining intelligence and consciousness is difficult because you have to define it in a way that includes everything we agree is conscious and excludes everything we agree is not. And we'll also have to agree what is and is not conscious. It's a way more involved problem than a lot of people give credit.
I think this is extremely important, but I would say that methodologically, messy definitions can nevertheless be stepping stones that bring us closer to full comprehension of definitions. I suppose I just have a specific hobby horse in this context and I don't know that I'm responding to you so much as a general idea I've seen elsewhere, that you're rhyming with, which is this idea that research has to start from agreed upon definitions.
I would say instead that there are such things as characteristic instances that embody lots of the stuff we are interested in, such that we can say "whatever consciousness is, it involves that" and we can just point to something that's a wild, underdefined phenomenon, and make it subject of study. Without a definition, I can still understand that states of mind that involve, say, wakeful awareness and intention are going to be about consciousness. We can know necessary conditions even if we don't know "sufficient" conditions, and the former are a stepping stone to comprehending the latter.
As a simplistic analogy, evolutionary designs of FPGA boards can end up relying upon idiosyncratic properties of the board(s) and create circuits that "shouldn't work" based on an idealized electrical circuit model. And they may not be transferable to other boards. In other words, to "understand" some evolutionary FPGA circuits, you need to "observe" more than just the gate configurations and idealized schematic.
Brains are not FPGAs or even circuits, but I think the analogy holds. They're not _just_ idealized representations of spiking neural networks.
DeepMind has been doing work along those lines with 'AI BRAIN' https://medium.com/@daneallist/unlocks-secrets-of-real-brain...
The basic problem here is that we don't understand what a neuron does, so how would we model it?
We understand lots of aspects of neurons and can model those. See the wikipedia page for some examples of well-understood principles.
It seems the DeepMind rat brain didn't really try to get close to biological neurons but still produced outputs close to a real brain. (vid from here for about 2 mins https://youtu.be/hIuXXvWY6-4?t=125)
Studying birds gave us some data, but mimicking them wasn’t what got us jumbo jets.
Probably an unavoidable property that emerges from the sheer and perverse complexity of the human brain, its 100 billion connections and the unimaginable amount of interactions between them, modulated by neurotransmitters + their reuptake, length, amount, location, quality and condition of pre/post-synaptic receptors, axons and other nano structures in the brain...
Comparing this disgusting moist, fleshy and electric masterpiece of nature with something primitive like a """neuronal""" network or LLM was always ridiculous.
I don't think complexity alone leads to consciousness. Rather, consciousness is the mechanism by which information is integrated in a relatively resource-cheap fashion: dump all this info into a shared mental workspace, add some reflective awareness of that info/workspace, and bam, you can both process the information in an integrated way, and you're also now aware of it all. The higher-level information processing and the awareness go hand-in-hand.
There're good reasons why most of the leading theories of consciousness focus on information integration. (Baars' Global Workspace Theory, Graziano's Attention Schema Theory, Tononi's Integrated Information Theory, etc.)
Complexity alone causing consciousness, much like an infinite amount of monkeys coming up with Shakespeare's complete works, is a far flung pipe dream.
The hardware doesn't guarantee the software, but the software can't exist without the appropriate hardware is probably a better way to look at it.
There's so much life out there that survives just fine without consciousness that it seems like a very narrow stroke of luck that it occurred at all.
[1] https://en.wikipedia.org/wiki/Artificial_general_intelligenc...
I think you're pointing out a bit of a chicken vs. the egg situation here.
We have no idea how intelligence works and I expect this will be the case until we create it artificially. Because we have no idea how it works, we put out a variety of metrics that don't measure intelligence but approximate something that only an intelligent thing could do (we think). Then engineers optimize their ML systems for that task, we blow by the metric, and everyone is left feeling a bit disappointed by the fact that it still doesn't feel intelligent.
Neuroscience has plenty of theories for how the brain works but lacks the ability to validate them. It's incredibly difficult to look into a working brain (not to mention deeply unethical) with the necessary spatial and temporal resolution.
I suspect we'll solve the chicken vs. egg situation when someone builds an architecture around a neuroscience theory and it feels right or neuroscientists are able to find evidence for some specific ML architecture within the brain.
You mention that people feel disappointed by ML systems because they don't feel intelligent. But I think that's just because they emerged one step at a time, and each marginal improvement doesn't blow your socks off. Personally, I'm amazed by a system that can answer PhD level questions across all disciplines, pass the Turing Test, walk me through DIY plumbing, etc etc, all at superhuman speed. Do we need neuroscience to progress before we call these things intelligent? People are polite to ChatGPT because it triggers social cues like a human. Some, for better or worse, get in full-blown relationships with an AI. Doesn't this mean that it "feels" right, at least for some?
We already know that among humans there are different kinds of intelligence. I'm reminded of the problem with standardized testing - kids can be monkeys or fish or iguanas and we evaluate tree climbing ability. We're making the same mistake by evaluating computer intelligence using human benchmarks. Put another way: it's extremely vain to say a system needs to be human-like in order to be called intelligent. Like if aliens visited us with incomprehensibly advanced technology we'd be forced to conclude they were intelligent, despite knowing absolutely nothing about how their intelligence works. To me that's proof by (hypothetical) example that we can call something intelligent based on capability, not at all conditional on internal mechanism.
Of course that's just my two cents. Without a strict definition of AGI there's no way to achieve it, and right now everyone is free to define it how they want. I can see the argument that to define AGI you have to first understand I (heh), but I think that's putting an unfair boundary around the conversation.
Minor(?) plot point spoiler ahead for season one:
Later on the story the main character further learns that the system is actually a mainframe made up of living human brains connected together. Hence the relation.
Neurological connective tissues and their corresponding signals may be more of a symptom of electrical wave action rather than a signaling nexus.
Everything is a wave. Everything is a probabilistic wave form with probabilistic outcomes.
Once science understands these waves and their myriad forms, we may have a greater understanding of our own wires and grey matter signaling.