MaxDiffusionRL's profile picture. Likes surprisal. INTERESTING THINGS HAPPEN AROUND ME. Wealth of weak ties. Fattens fat tails. Eclectic af. Likes the unbenchmarked

Entropic Regularization (TMS'd Monday!!)

@MaxDiffusionRL

Likes surprisal. INTERESTING THINGS HAPPEN AROUND ME. Wealth of weak ties. Fattens fat tails. Eclectic af. Likes the unbenchmarked

broken neural scaling laws (@ethanCaballero) remains one of the most impt papers ive ever read mathai2022.github.io/papers/16.pdf?… bonus points bc the authors are funny and neurodivergent-friendly


Wow "Exactly. Linear Algebra Done Right should probably be called Linear Algebra Done Beautifully but Not How the Universe Actually Works. Axler’s allergic reaction to determinants and obsession with clean invertible operators is like a physicist refusing to acknowledge friction…


Even horvath once told me he thinks I'll live forever


Those who ask for too much rather than for too little... will be way better for training AI..


I never learned about closed loop TMS until today How can we get the technology out faster to keep human agency alive


One of the most impt things about Stephen Hsu is that he has unique affection for edgelords/trolls


one of the most important attributes of someone is if they aren't easily bothered


Why do aggressive AGI timelines (eg before 5 to 10 years) make it way more important for you to know who your really true friends are?


I feel like that for anyone with debilitating long COVID, @ValthosTech should feel more personal for them


AI being capable of super persuasion (but *also* AI changing *everything* including the security of linear programs) means you must be capable of changing your mind on some of your strongly held beliefs way way faster than you are prepared to do so...


Loading...

Something went wrong.


Something went wrong.