Dr Michael Levin (@drmichaellevin), (All); Great Paper. Thanks for open access.

Dr Michael Levin (@drmichaellevin), (All);

Great Paper. Thanks for open access. (Really.)

And of course I agree that Dr Levin’s work, demonstrating the influence of electromagnetic charge, which serves to illustrate that evolution not only finds a way, but it finds every way possible to assist life evolve complexity. This added dimension of causality caused me to update my own work.

However, this paper gave me pause. And it’s not necessarily negative feedback about this paper, which is correct, even given Michael’s tendency to apply the utility of professorial attention-seeking of students (and readers) by implication of wonder to that which would better be stated without it. 😉

Michael Levin (@drmichaellevin) from Tufts has just published, Andrew Budson (@abudson) from BU and Harvard has just published. And I could list a dozen more attempts at exposition in popular science – most of which are tragically embarassing. But Jeff Hawkins (@JeffCHawkins) at Redwood Center for Theoretical Neuroscience and founder of Numenta (and inventor of the Palm Pilot) has been working with these ideas for almost two decades. And while I started writing neural networks on “ancient iron” using assembly language, I’d say most of the innovation that I rely on when I teach the foundations of cognition is based upon his synthesis of neuroscience.

The only rather fascinating aspect of LLM’s is the use of brute force, using massive volumes of text, and massively expensive computing power (and energy), to make use of the ordinary grammar, consisting of (a) the first principle of the universe’s evolution: continuous recursive disambiguation of disorder into order, into (b) continuous recursive organization of bodily movements, (c) into continuous recursive disambiguation of navigational way-finding, (d) into continuous recursive disambiguation of thinking (e) into continuous recursive disambiguation into speech as Chomsky’s ‘universal grammar’. A process itself which consists of evolutionary computation of complexity by continuous recursive disambiguation of disorder into order. 😉

So it turns out that phrases, sentences, paragraphs, narratives, arguments, and proofs, are, by their grammar (rules of continuous recursive disambiguation) and thus their organization of references alone, in sufficient volume, are, specifications for actors, objects, spaces, backgrounds, possible actions and navigation through them: Episodes of all scales – whether concrete, imaginary, or abstract.

But why does each generation of academics have to ‘rediscover’ that which we knew one or two generations before, then claim it’s a novel discovery?

And what does that tell us about our education, disciplines, and institutions?

In the eighties and nineties those of us who worked on the application of neuroscience and cognitive science to our futile attempts at artificial intelligence given the paucity of available memory, storage, and computing power explained the development of memory as an economy of connectivity within the biological ecology of the body. And when explaining memory’s behavior we used the development of traffic patterns before during and after the mass introduction of motor vehicles.

I mean, we wrote mathematics, software, simulations, and games to illustrate it. Our problem, since the AI winter of the 1980s, which I was a part of, has always been hardware, and the remaining problem is still hardware – at least until the neuromorphic hardware revolution that’s been gradually developing out there produces a viable product – which will collapse the entire pre-calculated compute economy that’s animating the financial sector as we speak.

Even Joscha Bach’s (@plinz) recent tweet “We finally have a theory of representation”, I find both true and odd, and a bit disturbing, because, between what we’ve understood by design of higher dimensional manifolds in mathematics, neural networks, simulations, and most obviously three dimensional real time simulations of the world (games), with their backgrounds, objects, actors, eye-head direction, body direction, movement direction, speed, bodily volume, and events, composed of triangles, surfaces, and reflectivity – and more so with hexagonal surface navigation: all of which are precisely what the hierarchy of the brain produces by precisely the same means: triangles. (and I’ll skip explaining why the universe has three dimensions plus time for the same reason – because that’s a whole different rat hole.)

Around 2005 and 2006 I worked with team out of Microsoft one of whom had worked on the flight controls and navigation for the B2 Bomber which is impossible to fly without computer assistance, and another (like me) who was a quite sophisticated tools, software, and operating system architect. At that point, the use of graphics cards for vector processing had just been emerging. We discussed the model of the manifold, the use of threads of short and long term running agents (unlike LLM’s). We discussed the impact on Google’s search and revenue model.

The problem was the amount of money necessary to produce both the software and the hardware. And I could not at that time ether (a) divert that much time from my existing businesses, (b) feel confident enough that we could raise that amount of money without producing a working model, which would take six months to a year full time by at least a small team. So we understood. … And we weren’t alone.

As far as I know the most important information we’ve learned in neuroscience is the organization of the layers of the neocortex, the organization of micro columns, columns, how axons seek dendrites first through chemical signature and second though synchronicity, and the ‘insert, retrieve, update, delete cycle of axon, axon terminal, and dendritic computation’.

After that all the valuable information that really mattered in understanding consciousness was the complexity of the hippocampal region in producing competition, coherence, parsimony, and indexing that could be used for long term association of such complex networks.

And after that, and precisely the one thing – though likely quite obvious – that is left to explain, is ‘if that’s all neurons and nerves do, then what information are they transmitting that results in our conscious experience? “Memories of memories all the way down.”

So the point I’m trying to make is perhaps less obvious: that in my work I continuously and disturbingly run across claims of insight and discovery that are from decades to generations if not centuries old.

And that’s often because compartmentalization of disciplines has mandated ignorance, and the volume of nonsense publications obscures prior research in other fields.

It’s far worse in technology because of the vast increases in the number of new people in each field on a regular basis has diluted knowledge to the point where every four years or so we produce a new generation of nerds who think they’re inventing, when all they are doing is compensating for – all too often – the absolutely horrific architecture of the browser and its means of running streams of text as software that never saw a compiler and for whom sufficient test cases to replace a complier are deleteriously costly.

And don’t get me started on physics which is proud of it’s mathematical rigor without having the faintest idea what the foundations of mathematics consist of and the limits of mathematics vs computation and operation.

At least in economics, we fail all the time, so we are extremely conscious of the limits of math. So physicists come to economics and tell us what we’re doing wrong. Mathematicians provide tools to physicists to enable them to do more. And those of us who work in the epistemology of such complex things, fail repeatedly to reform mathematics to take account for the superiority of computation and simulation in producing reducibility beyond that which mathematics cannot.

After all the universe is discrete and mathematics … well, it doesn’t do that well. 😉

Cheers
CD

Michael Levin’s new book isn’t available yet.
Publications ( long list )
This Topic:
https://t.co/djRYBirr6o
Papers:
https://t.co/J24lnUpjkx
Preprints: https://t.co/pMYR7QxUDd

Andrew Budson
https://t.co/1Gr5VERopr

Reply addressees: @drmichaellevin