Maxim Raginsky

This has been rigorously proved by Mitter and Newton in 2003 (although pointed out by orhers even before that in various contexts): mitter.lids.mit.edu/publications...

More words on this: realizable.substack.com/p/hayeks-abs...

Hayek's Abstract Logic 9000 The use of knowledge in Searle's Chinese Room.

realizable.substack.com/p/hayeks-abs...

More words on this: realizable.substack.com/p/hayeks-abs...

I’d assume JE saw an opportunity to cultivate this network and went for it.

It’s the edge.org connection.

It was a good workshop, though.

I attended a workshop on information and decision in social networks at MIT back in 2011, the plenary speakers' list was, shall we say, very Epstein-y: wids.lids.mit.edu/2011/

predictably creeprational?

tuck up your sleeves and loosen your talktapes, boyo

Nothing is real, everything is permitted.

the short stories are going to be lit

bsky.app/profile/mrag...

Historically, it came at a time when people started realizing that many ideas originating in automata and computation were conceptually related to the emerging notion of state-space models.

There’s definitely a trend toward over-formalizing things, although it’s not universal. The above definition, though, reflects the desire to unify the description of continuous and discrete systems within a single framework.

Are you talking about my lectures on SDEs? Yes, I am (slowly) finishing them.

<Homer Simpson voice> mmmm, bisque

I was partly trolling. This is from Kalman, Falb, Arbib, "Topics in Mathematical System Theory," but it's also true that mathematical control theory is not beating the allegations of being abstruse.

One important example is that optimal control generalizes the least action principle from Lagrangian mechanics, and a big distinction is the introduction of dynamical constraints (not all paths are allowed, only the ones your given control system could take).

lol true

It's useful, but you really need the open systems viewpoint which, with the exception of thermodynamics, is largely absent from physics. The discussion here is very useful and clarifying: mitter.lids.mit.edu/publications...

... those papers comes from physics instead of control theory, so it's a small wonder that *everyone* is hopelessly confused about the distinction between inputs, parameters, outputs, states, etc. In fact, those confusions still persist. Bottom line: read (control) theory. (2/2)

I'm re-reading various papers surrounding the debate between computationalist and dynamicist approaches to cognition (e.g., Fodor vs. van Gelder), and it's a hot mess. For one, of course Turing machines and finite automata are dynamical systems! But also, the dynamical systems formalism in ... (1/2)

Bingo!

I will just keep to "the maximum principle."

In today's lecture in my optimal control systems course, I will say the Latin phrase "ex ungue leonem." What will be the topic of the lecture?

At what point does sticking to an ever costlier bet turn into a degenerating research program, in Lakatos’ sense?

If it were a rational process, that would be one thing. But it's not -- it's driven either by herd mentality or by contrarianism for its own sake. Why not try hedging? Explore different approaches in parallel without staking your reputation on just one of them, reallocate resources adaptively.

I really find the language of “bets” grating. Silicon Valley technologists bet on scaling LLMs, Yann LeCun bets on world models, etc. This is neither science nor engineering, it’s all egos and ideology.