In this thread, I will argue that the conventional wisdom that Shannon's information theory is all about syntax and not about semantics stems from superficial reading. On the contrary, even his 1948 BSTJ paper is already concerned with syntax, semantics, *and* pragmatics. 1/14 Of course, it does not help matters that Shannon himself states that "semantic aspects of communication are irrelevant to the engineering problem" of communication on the first page of his paper. But you have to read the entire paper, not just the first page. 2/14

I am reading about the Piaget-Chomsky debate at Royaumont, and it is about so much more than "just" the faculty of language. M. Piattelli-Palmarini, How hard is the "hard core" of a scientific theory? (parts 1 and 2)

massimo.sbs.arizona.edu/sites/massimo.…

massimo.sbs.arizona.edu/sites/massimo.…

Currently reading this excellent paper by Brian Cantwell Smith: Even though it is a masterful takedown of Lenat and Feigenbaum’s CYC project, full of acerbic wit and withering criticism, it’s as relevant today in the context of all of the magical thinking surrounding GPT-3:

I’ll just leave this here Here we are:

late to the game, but what the hell:

1 like = 1 favorite book

go 1. The Feynman Lectures on Physics
2. Red Plenty by Francis Spufford
3. Against the Gods: The Remarkable Story of Risk by Peter Bernstein
4. Beetle in an Anthill (Жук в муравейнике) by Strugatsky brothers
5. The Time Wanderers (Волны гасят ветер), same

"Neural Stochastic Differential Equations: Deep Latent Gaussian Models in the Diffusion Limit" with Belinda Tzen arxiv.org/abs/1905.09883 1/5 Neural SDEs are simply deep latent Gaussian models in the limit of infinitely many layers, with appropriate scaling of the layer-to-layer noise. In our earlier work (to appear in COLT 2019), we have analyzed the expressiveness of such diffusion-based generative models. 2/5