With my colleague Dimiter Kramer at @DeepInfra, we proved the Courtade–Kumar conjecture. We show that a delicate refinement of the conjecture, incorporating the output bias, admits an induction on dimension.
Paper: https://t.co/YjB9xUL9Eh
@mmitchell_ai it’s required on @ilkelimedyaio! it’s a WA state LLC with Turkish joint stock ops on top - my goal is to study employment, productivity & law as a hobby project 🥰
@benedictevans@nonparibus hahaha this is why it’s been really painful over the last year explaining a function that does “-1” of twitter engagement, which is whatever the inverse of bullshit is in math 😂 can someone please teach survivorship bias on twitter @elonmusk? #ilkelimedya
@Jason I disagree, I’ve been demonstrating how both can be true in WA state with @ilkelimedyaio since last summer - kiiiinda like what ethereum did for bitcoin. can we please at least slow down reinventing things from ground up? 😉
@mmitchell_ai@ProfNoahGian let me know if you’re interested in exploring double blind studies - the promise of the approach you outlined here is what got me into last mile AI experience measurement. I believe it’s possible to do this at scale with hybrid inferencing within trust boundaries.
@benedictevans I can’t wait for the avg person to fully understand what a function is, and how we now can train on the absence of signals in moments that matter. I’m very excited for what social media can be over this decade 🥹
@yishan I think of it more as continuous transformation / exponential improvement through superhuman compute. This may sound grim but one can say (much needed) AI regulation is technically introducing data boundaries to otherwise much more efficient worldwide compute power. ++
@bcmerchant people literally organized massacres over advertising AI designed in california and released globally. I think the most valuable conversations start by accepting that these events did happen and american tech companies & shareholders financially benefited from it.