A shame, since the conversation was almost interesting. — Banno

:yawn:

This is super cool. So many neat intersections. It'd be interesting to look at the original paper, but it'd probably read funny to me -- i.e. the pop-version is probably just as good.

One thing the pop article made me wonder about was the notion of coherent/white-noise in relation to a dishbrain. The technical side going on in this set of sentences:

To allow the brain cells to play the game, the computer sent signals to them indicating where the bouncing ball was. At the same time, it began monitoring information coming from the cells in the form of electrical pulses.

"We took that information and we allowed it to influence this Pong game that they were playing," Kagan says. "So they could move the paddle around."

At first, the cells didn't understand the signals coming from the computer, or know what signals to send the other direction. They also had no reason to play the game.

So the scientists tried to motivate the cells using electrical stimulation: a nicely organized burst of electrical activity if they got it right. When they got it wrong, the result was a chaotic stream of white noise.

is where the meat of the argument would be. "That information and we allowed it to...", "sent signals indicating", and "chaotic/organized electrical activity" -- in some way I wonder if it'd been any different if the dishpan was hooked up to a series of lights and if the experimenter had just shocked the cells or anytime it guessed the wrong light -- but there's a lot of intentional conduct being implemented by the experimenters, and I wouldn't know how to tell if it's the cells in the dish-brain or the cells in the bone-pan that's making the inferences.

in some way I wonder if it'd been any different if the dishpan was hooked up to a series of lights and if the experimenter had just shocked the cells or anytime it guessed the wrong light — Moliere

These sorts of experiments have already been done long ago. Neuronal networks don't seem to care if they're electrocuted, burnt, stamped on or otherwise 'punished' (this is not strictly true, but it sounded better rhetorically). The point about the noise wasn't to provide some form of negative feedback (picking white noise at random), it was to prove that unpredictable feedback was indeed a negative feedback constraint (as opposed to, say dopamine circuits in a larger cortex).

The playing of Pong was the trivial part of the experiment (though important to the demonstration of the set up), the important bit was that the cells avoided uncertainty (noise) even at a level of network complexity so low that other feedback systems, such as neurotransmitter availability, are still miles away.

It demonstrates Friston's model because uncertainty avoidance has been shown to be capable of driving feedback, and as such other forms of feedback seem more like multipliers than drivers (which is what Friston's model would predict).

uncertainty avoidance — Isaac

Most people avoid tense situations. Repo man spends his life getting into tense situations.

My point is that explanations in terms of intent do not apply to dishbrain. Talk of intent is part of a different language game. — Banno

But there is no evidence that the universe itself had any intent in the forming of sentient life. It happened because it could happen. The mini brain was created from human intent.

My suspicion is that for some act to count as intentional, the organism might in some sense have done otherwise. — Banno

Do you not think that intention requires the system to be self-aware and be able to 'judge' based on phenomena such as instinct, rather than simply choose from a list of alternate actions based on a conditional input?

Mini brain/dish brain is a very interesting advance in the world of organic computing.
The organic computer combined with the quantum computer will I think become very important to the future direction transhumanism may take.

Cool, much thanks -- I was focusing on pong being played, for sure, and you've helped in pointing out that's not the point, but rather to demonstrate a model's prediction of neuron-network behavior.

heh, yeah, glad I just asked :D.

So no instant-karate-brain-chips in the next year or so, I suppose, but one step closer to the cyberpunk future.

to my eye the question is, if we choose to say that the dishbrain intends to move the paddle to stop the ball, have we extended the use of "intend" too far? So far that we have lost some worthwhile distinctions. For instance, we commonly only attribute culpability in cases of acting intentional - is the dishbrain now culpable for any negative consequences of its intent?

Isn't the language around intent distinct to that around use, including, as Josh says, normative features — Banno

I would be comfortable in saying that if what the dish-brain is doing is generalizing from particular events, and thus using memory to anticipate for the sake of maintaining normative directions ( the norm being whatever consistency of coordinated neural activity is repeating itself ), then this represents a primitive form of intending. I am defining organismic intention here as a responsiveness to its environment involving capacity to discriminate and respond flexibly and appropriately to aspects of its circumstances that matter to its self-maintenance.

But what purpose does this coupling serve? Is it a striving to avoid the white noise as an aversive stimulus or negative reinforcer?

Where did such a preference come from? It can only be a relic from the genetic bauplan of what makes a neuron useful in an actual embodied relation with its world — apokrisis

When you ask where the dish-brain’s preferring to avoid white noise comes from, might we not distinguish between a specific and general source of motivation? In the most specific sense, motivation is tied to pre-determined sensitivities to environmental features , as well as a pre-determined capacities to act on those features. As a result the world looks and matters very differently to a bacterium or an ant or a human. The particularities of preference ‘come from’ these unique capacities to sense and respond, but in a more general sense preference is always for the sake of the continuation of whatever coupled system of organism-world interactions is repeating itself as a normed process.
So specific preference is dictated by direction of sensory-motor ‘use’, and use , beyond specific goals , is in the service of the preservation of self-consistency of change in the face of potential interruptions and perturbations. I don’t think the striving for self-consistency needs “a genetic bauplan of what makes a neuron useful in an actual embodied relation with its world”. Rather, self-consistency is what Piaget called the most general organizing principle of life.

Perhaps I should make my quibble more explicit. The temptation is to treat intention as developing by degrees, from wiggling ameba through to a fully autonomous sentient beastie. I'm suggesting considering instead a difference of kind, rather than degree. Intent plays a role in various aspects of our life other than the experiments so far discussed.

A thought experiment, if you will. Suppose Fred trains dishbrain to recognise the face of someone he wishes to be rid of. Further he links dishbrain to a gun and trains it to target and shoot - a step not too far from playing pong.

Now if we are to say that in some small degree, dishbrain intends to shoot, are we to also say that dishbrain is to some extent culpable? Is dishbrain an accomplice to Fred's plot?

I think that stretches credulity. I think it worth recognising a difference between our language around cause and effect and our language around intent and responsibility.

Hence my preference for something along the lines of anomalous monism, in which our intentions are not reducible in any direct way to physical states.

@Isaac? Any thoughts?

(Again, my complaint is not against the experiment, but the way in which it is reported as "teaching a dishbrain to play pong", with all the intentional implications thereof.).

Hence my preference for something along the lines of anomalous monism, in which our intentions are not reducible in any direct way to physical states. — Banno

Doesn’t fly. Old hat.

Biosemiosis fixes things by showing why life and mind are based on this epistemic cut. We now have a testable theory of the modelling relation that accounts for how it works, whether we are talking semiotic codes at the level of genes, neurons, words or numbers. Intentionality can be understood genetically, neurally, verbally and numerically.

No point trying to breath life into ancient history. :grin:

Or perhaps this addresses stuff well outside your system.

I commend Mary Midgley to you, as antithetical to your scientism. Challenge yourself.

Challenge yourself. — Banno

...says the guy who never does...

Cheers. :meh:

We now have a testable theory of the modelling relation that accounts for ... — apokrisis

Is it "accounts for ..." or is it "may be able to account for ..."?

You and @Joshs (but @Isaac I think to a lesser degree) are always talking as if this is all a done deal, as if all we needed was the theory, as if having a theory you can imagine testing is the same as testing and confirming it, as if saying you could in principle fill in the details of an account is the same as actually filling in those details rather than the giant IOU it actually is.

All well and good if you're laying the groundwork for a research program, but am I wrong to think there's rather little in the way of observation to support all this theory? I'm not denying the elegance of the theory, or at least the bit of it I understand, but does it have anywhere near the body of support that, say, QM or evolution by natural selection has?

To be clear: not critiquing whatever science there is here, which you are vastly more competent to judge than I am. I am questioning whether you are in so secure a position that you are entitled to be as dismissive of doubt as you generally are. Insofar as people ask questions in order to better understand what you're pitching, of course you should be answering, 'yeah that's it' and 'no that's way off' — you're the world's leading authority on your own position.

But this forum is not exclusively dedicated to your views, so insofar as you or @Joshs answer the sorts of questions philosophers talk about with "Shiny new theory says X" without assuring us that shiny new theory has much claim to truth, why should we listen to you? You guys have preferences among theories, good for you; let us know when you have overwhelming evidence.

Is it "accounts for ..." or is it "may be able to account for ..."? — Srap Tasmaner

Yes, you are being challenged to challenge precisely this. So tell me, how does biosemiosis fail the task? What does it lack or overlook exactly?

If Dishbrain is meant to be a test of the Bayesian Brain approach, then at least the "philosophical" reaction ought to take its modelling relation thesis seriously and not just wander off into the thickets of yesteryear's armchair AP musings.

but am I wrong to think there's rather little in the way of observation to support all this theory? — Srap Tasmaner

Spectacularly. You are talking about what has arisen in first theoretical biology, and now theoretical neuroscience, as a meta-theory.

As I have said often enough, biosemiosis as "the mechanism" of bios – life and mind – has really hit pay-dirt now that we have the tools to see what is happening down at the nanoscale of the body and brain's molecular machines.

https://thephilosophyforum.com/discussion/comment/68661

You guys have preferences among theories, good for you; let us know when you have overwhelming evidence. — Srap Tasmaner

But how can you assess the evidence if you haven't understood the paradigm shift?

But how can you assess the evidence — apokrisis

Wouldn't dream of it. I'm not competent to.

But if, on my next day off, I wandered over to the Life Sciences building at the local state university, and asked everyone I met there about biosemiosis and Friston and Salthe and all the rest, they would all assure me that it is universally accepted — except perhaps for a handful of dinosaurs on the verge of retirement — and as well-supported as, say, evolution.

Is that what you're telling me?

Well phrased. I share your disquiet at what appears to be quite articulate, erudite handwaving and overreach.

But if, on my next day off, I wandered over to the Life Sciences building at the local state university, and asked everyone I met there about biosemiosis and Friston and Salthe and all the rest, they would all assure me that it is universally accepted — except perhaps for a handful of dinosaurs on the verge of retirement — and as well-supported as, say, evolution. — Srap Tasmaner

Metatheories don’t have to invalidate each other. Biosemiotics ties together the physical, biological and psychological on the basis of a unified overarching scheme that doesn’t so much contradict rival approaches as attempt a grand synthesis that it hasnt occurred to others to try before ( that is, not before Peirce, Piaget and a handful of others). I think what the profs at your local state university will say about Friston and biosemiotics is that in every era of science there emerges a certain network of powerful organizing ideas that inspire, in different ways , a young generation of researchers in a variety of fields, regardless of whether they embrace every aspect of these frameworks. This is the role that free energy and biosemiotics is playing today in interdisciplinary work across the spectrum of biological and neuropsychological theorizing. (I could add enactivism to this list). It’s not a question of such ideas simply being proved or discarded. They are more formidable than that. In some form or another they will remain with us, and even when they give way to a new set of motivating assumptions, their role in making the next steps in thinking possible will be evident.

insofar as you or Joshs answer the sorts of questions philosophers talk about with "Shiny new theory says X" without assuring us that shiny new theory has much claim to truth, why should we listen to you? You guys have preferences among theories, good for you; let us know when you have overwhelming evidence — Srap Tasmaner

Many competing theories can justifiably claim empirical validity and truth. Each can have evidence supporting it. It is thus not simply a question of whether one metatheory is more ‘true’ than another but whether one is true in a more useful way relative to certain aims and problems. You would have have a hard time finding a theoretical biologist or neuroscientist denying that biosemotics or free energy are devoid of evidence. Your preference must be made on the basis of whether the conceptual shift required to make the evidence supporting these ideas attractive is one you are willing or able to make.

Is that what you're telling me? — Srap Tasmaner

Of course not. That would be like asking house painters what they thought about abstract expressionism.

The meta-theory of the biological and neurological sciences is still deeply reductionist. Especially where it is dominated by medical and pharmaceutical funding.

You would have to strike on folk who indeed research at the meta-theoretical and inter-disciplinary level – like the Salthes and Fristons of this world.

That is why this is as much about metaphysics as physics. What is Friston saying when he claims to have created Bayesian mechanics as a new branch of physics? What is Salthe saying when he rejects Darwinian evolution?

You are talking about the tiny few who are seeking to shift paradigms rather than the vast majority who simply hope to assimilate more facts to the prevailing theoretical structure.

Like me, you would have to seek the right people out. This isn't about a weight of evidence. It is about a willingness to stand apart from the herd.

Salthe paid the price career-wise for being too openly metaphysical. Friston stuck closely to the mathematical route and has been rewarded. He only started to make his bold claims a couple of years ago. And he couches them in terms of engineering. Society is going to get paid in terms of technological results – as DishBrain is trying to sell.

Even in science, there are strong constraints on what you are allowed to believe. And the money is in the reductionism that produces the machines.

So if your local life sciences department is in general anti-reductionist, then something very weird is going on there.

Thank you both for forthright answers.

@Joshs, your describe something that sounds like a research program a la Lakatos, which seems pretty reasonable. My grasp of the philosophy of science is just strong enough to find that plausible but not strong enough to quibble. — I do wonder though, whether we are the right audience for your cheerleading, not being scientists who may do research developed around the core of your suggested program. Why us?

I don't know what you're describing, @apokrisis, but it sounds somewhat more — I'm not sure how to put this — messianic.

That would be like asking house painters what they thought about abstract expressionism. — apokrisis

I'm sure this is meant to suggest what we benighted AP folks call a "category mistake," but it sounds a lot like contempt.

It is about a willingness to stand apart from the herd. — apokrisis

At least there is, it seems to me, a natural audience for what Josh has to say; it's just not us. You, on the other hand, are a voice crying in the wilderness. No point talking to the herd; they won't listen and wouldn't understand if they did. So why us?

A string of inaccurate ad homs. Why did you ping me for my response I wonder?

Most people avoid tense situations. Repo man spends his life getting into tense situations.

I'll take obscure 80s crime flicks over Freud.

no instant-karate-brain-chips in the next year or so, I suppose — Moliere

No, but they might progress to pac-man.

I think it worth recognising a difference between our language around cause and effect and our language around intent and responsibility.

Hence my preference for something along the lines of anomalous monism, in which our intentions are not reducible in any direct way to physical states. — Banno

I agree with the anomalous monism, but I think where I might be slightly more sympathetic to @Joshs's position is that, if I've understood it correctly, anomalous monism allows for this overlapping layer of language (and associated consents) measured by felicitous usage rather than correspondence with some physical description of the world. Here your quibble about 'intent' as used in law is valid as it doesn't matter one jot what the neurons of the defendant are doing, it matters only what the defendant, as a person, is doing.

My counter-quibble is that sometimes it does matter what the neurons of the defendant are doing, cases have bee acquitted on grounds of mental incapacity and the expert witnesses in those cases are often neuroscientists attesting to exactly the kind of reductionism about intent that Dishbrain is the extreme end of "He cannot have intended to murder X because the neurons forming that intent were disrupted in their activities by a tumour".

So there is, I think, an uncomfortable leak in the otherwise watertight irreducibility of our terms like 'intent'. What to do about that leak, I don't know, but I think those acquitted of crimes they had no 'intention' of committing are probably grateful for it.

As to Dishbrain, I'm personally quite comfortable in saying that it didn't intend to play Pong, having no concept of 'playing Pong' toward which to direct its actions.

For me, the idea of Dishbrain 'intending' to avoid uncertainty is the exact opposite of what the experiment shows (and what Friston predicted). The experiment, insofar as it supports the free energy principle, demonstrates that it is a physical necessity that self organising systems such as a cortex of neurons will act on the world to avoid uncertainty. That's the beauty of the theory, they can't not (and remain self-organised). I don't think such a physical necessity matches they way we normally use 'intent'. It might show that there's no physical basis behind our word at all, but that's fine, I don't see there needs to be.

if, on my next day off, I wandered over to the Life Sciences building at the local state university, and asked everyone I met there about biosemiosis and Friston and Salthe and all the rest, they would all assure me that it is universally accepted — except perhaps for a handful of dinosaurs on the verge of retirement — and as well-supported as, say, evolution. — Srap Tasmaner

Neuroscience is oddly divided, in my experience. If you did take such a survey, you wouldn't so much get a dozen different meta-theories as you'd get one or two meta-theories and 90% shrugs, or a muttered "we poke X, Y happens - we're now developing a drug to poke X".

Neuroscience can be incredibly, frustratingly mechanistic. Mainly, to be fair, because the brain is so fiendishly complicated that trying to distil it into any kind of meta-theory is a nightmare whereas treating it like a mechanical 'black box' can yield a surprising amount of well-cited papers along the "we poked X and Y happened" lines.

As to the 10% remaining (probably less, to be honest) - would they all agree with Friston? Well, in my experience (which is limited to a couple of British Universities and the journals I read, so not representative), there'd be little disagreement about the fundamentals such as the basics of predictive modelling in general. The hierarchy of ascending cortices Friston uses is disputed - some take a more multiple equal processing approach. The degree to which cellular-level uncertainty avoidance manifests in behaviour is often swamped, not by competing meta-theories, but by a 60 year-old obsession with the role of neurotransmitters in affect (low on Serotonin, anyone?). So I think basically you'd get little disagreement at the neural network level, but a lot of muddy water when extrapolating out into behaviour. Once the super-stardom of neurotransmitters fades, I think we'll see support more in line with the bigger theories, but it's like convincing music aficionados that other bands are at least as good as the Beatles. Hard work.

One of the aspects fo the free energy principle that's often misunderstood is that it describes the consequences of self-organisation as much as it models reality. As such, although there are predictions and experiments such as this one to support them, there is also a big chunk of the theory that simply not disputable - not because of the weight of evidence, but because, like mathematics, it's just re-arranged the equations to say something interesting. You can dispute that it's interesting, but you can't dispute that that if Y=X then also X=Y, if you see what I mean. Friston's basic free energy gradient equations are just re-arrangements for Bayesian optimisation of entropy gradient climbing expressions. It's more a description of what it means to avoid entropy (remain organised) than a modelling assumption, in that sense.

didn't intend to play Pong — Isaac

:cool:

Let's start with the basics ... what is a brain?

Once the super-stardom of neurotransmitters fades, I think we'll see support more in line with the bigger theories — Isaac

That sounds plausible. Also wouldn't be surprising if the role of neurotransmitters was overestimated because we got to them first, because there were tractable problems about their role we were able to address before we could make any headway on neuronal networks. That means there will be a step coming at which neurotransmitters will likely be given too steep a discount, before the pendulum swings back to finding a harder-to-reach role for them.

This is all just science journalism though. One thing I always have in the back of my mind is that science can cheerfully proceed this way, iterating and refining, and at multiple levels — individual experiments can be repeated but done better, theories can be replaced by other theories within the same program or paradigm, programs and paradigms can be replaced by others. The latter shifts can be difficult to explain, but should engender, always, more and better science. A scientist can expect her field to move over the course of a career, and must expect to say, "When I was in grad school thirty years ago, we all thought ..., but now ..." Philosophy moves, but not quite like this. For what sorts of values for X would a philosopher say, if X, then we'll all have to start thinking about Y differently? It happens, but much less often, so there must be a different mechanism here. (Assuming it's something besides fashion.)

there is also a big chunk of the theory that simply not disputable - not because of the weight of evidence, but because, like mathematics, it's just re-arranged the equations to say something interesting. You can dispute that it's interesting ... — Isaac

Right, and that has to do with interpretation, but this is so complicated, because there's science-engendering generality and interpretation, and there's public-facing, also general (at least because detail-poor) interpretation, which is in some ways close to application. Scientists who can be very clear about 'what this means' for the field, can be very wrong about 'what this means' for non-scientific purposes. Down in the valley of the nitty-gritty, the generality of the program is still a constraint, but non-scientific generality is worse than useless; in philosophy, those two sorts of generality should be nearly the same (because of "saving the appearances"), and when they aren't philosophers say the same sorts of things scientists say: you think you have knowledge but here's what's really happening when you think you know something (philosophy); you think you see things but here's what's really happening when you think you see something (science).

It's more a description of what it means to avoid entropy (remain organised) than a modelling assumption, in that sense. — Isaac

And maybe in that sense no more than a reworking of Kant, who knew enough to expect the subject of knowledge to have a sensorium, but not enough to expect the subject of knowledge to be self-organizing, with all that entails. Such an enriched Kantianism might be interesting, but it's not really comparable to the original. You could as well say that the subject of knowledge must have arisen through evolution by natural selection.

It's just not at all clear — and @apokrisis is right about this — at what point we are really passing from a priori to a posteriori. Kant's conditionals are supposed to be awfully strict, meaning they are intended to rely to the greatest extent possible only on logic and not on how the physical world happens to be.

Point being, it is interesting to know how an organism might acquire knowledge, not least because we are organisms. It is less clear that only an organism can be the subject of knowledge, but if you're a biologist that's exactly what you're going to assume because you only study the natural world, not the possible world. Kant's concern was knowledge, not the knowledge of organisms.

But there is this grey zone, and I take it this is where apo's grand synthesis lives, in which we consider what a possible natural world could be, and that means, to begin with, showing that the actual natural world can be described without remainder as such a world. It's just not clear to me what else this is: it's definitely not science, because you can only do science with the actual natural world , and this level of generality is somewhere above what usually defines a research program; but it's not just interpretation either because there's more here than the science of the actual natural world.

Friston's basic free energy gradient equations are just re-arrangements for Bayesian optimisation of entropy gradient climbing expressions. It's more a description of what it means to avoid entropy (remain organised) than a modelling assumption, in that sense. — Isaac

Well put. His achievement is to take the general prediction-based view of “brain processing” and turn it into differential equations. So he gives a mathematical basis that offers some unity to the field.

Outside of neuroscience, the view had taken hold that the brain must be some kind of input-output, data-crunching, computer. The mathematical basis was Turing Universal Computation. The metaphysics was homuncular Cartesian representationalism. Philosophy of mind was in love with the modular cognitive science model.

But Friston offers a maths more suited to the actual neurobiology of the brain as a model of its environment that acts to predict its inputs. This is the metaphysics of Peircean semiotics or Rosen’s modelling relation. And it fits the general enactive or embodied turn of cognitive psychology and philosophy of mind.

Mind science was tracking down this road since Helmholtz until the computer revolution derailed it in the 1950s. A new mechanistic paradigm was forced on to it. And now it has returned to that more naturalistic paradigm.

Mind science was tracking down this road since Helmholtz until the computer revolution derailed it in the 1950s. A new mechanistic paradigm was forced on to it. And now it has returned to that more naturalistic paradigm. — apokrisis

I don’t know that mind science had much of a chance between Helmholz and the rise of embodied approaches in the 1990’s. After turning its back on James, it quickly embraced positivist behaviorism. It took the computer revolution, along with Chomsky , Shannon and Bruner , to make talk of internal intervening variables acceptable again. Mind science had to wait another 40 years to free itself from its rationalist bias of first-generation cognitivism.