Tarantino Wept.

Just getting started, the last phrase needs a completion, in the bullshit chapter:
This is where Large Language Models come in. These systems have no ground truth, no underlying model of physics, no rules of logic, and no embodied understanding of how the verbal symbols they so deftly manipulate.

If they can do it, more sophisticated continuous multi-scale processing on CPUs and GPUs has room for improvement without neuromorphic substrate revolutions. There is almost no history of even single level field computation doing the kinds of machine learning tasks common in network modeling.

I love this direction of thought but can assure you that computer and electrical design engineers spend enormous effort and compute budgets making sure that continuous multi-scale processes in semiconductors are modeled with those boring old symbols so that they are corraled into being digital.

And then he uses the term "neuroflaming" which in, spite of a heavy influence by WJF III in my work, I'm scratching my head and hoping it's a typo.

Transients! We love to see it. I think I need to look for a review covering recent work on transients in high dimensional nonlinear dynamical systems. It was a bit of a hot topic in the early 90s and I ran with it (i.e. why not compute and readout with transients rather than attractors).

Thanks, found the same explainer blog via a reddit post today.

@wang.social was it neural tangent kernels that you mentioned to me after the talk today? If so was there some specific new paper or development?

I miss reading to adults.

Not actually read yet but in my collection:

The Others: How Animals Made Us Human Paul Shepard
The Other End of the Leash: Why We Do What We Do Around Dogs, Patricia McConnell

I've enjoyed Stanley Coren's books on canine intelligence. Fun fact: dogs can recognize you from a great distance by your movement signature. (He's also a visual psychology and illusions guy)

“Outside of a dog, a book is man’s best friend. Inside of a dog it’s too dark to read.” Groucho Marx

"More singing" is almost always part of the solution, regardless of the problem. I love this woman for that observation.

Thanks for sharing your tribute and feelings with us. Remembering all my goodbye days, and remembering the occasional joy of being reunited in dreams (where occasionally I am a dog, or at least crawl under the porch).

I'm trying to interpret the "but" in your statement. Is it implying an error in their argument by not recognizing that fact, or that it's not really top down because an aggregate state doesn't justify the term top down? Something else?

More of this please. It was something of a footnote in my dissertation work but with a neural field approach to categorization and invariance, there would inevitably arise "stuck on" sites that would be interpreted as Jennifer Aniston cells if you liked that sort of thing.

Looking forward to diving in.