Дорогой @mikhail_dvorkin, когда я году в ≈2013-ом слушал твои пары по алгоритмам в CSC, ты обещал написать книгу «Что будет, если P ≠ NP, и если P = NP».
В свете последних достижений научного мира я прошу поторопиться, так как шансы не увидеть одну из двух частей растут.
meet @mostik_ai!
what happens when you put 12 PhDs in one room for four months? first place on the ARC-AGI leaderboard, which I can't say much about while the competition is still running. and this, which I can.
everyone's arguing about whether open models will catch up to frontier models. we think it's the wrong question. here's the one we pose: why does a frontier model have to generate your answer at all, when the only thing you need from it is the reasoning?
we do this by enabling models to communicate in latent space. through our protocol, hidden states pass straight from a frontier model into a small one running on your infrastructure -- no text between them, and neither model is fine-tuned. two models from different families, sharing reasoning, both left untouched.
how do we know it works? we tested it on a setup where a 753B model reads the problem, and a 4B edge-class model writes the answer. with this approach, we get results 80% as accurate as the frontier model, but at 20x faster performance.
we're committed to preventing frontier model lock-in and are already partnering with inference providers to accelerate open-weight adoption. we've done this between 15 of us, in four months, 12 PhDs and a Fields medalist, backed by @generalcatalyst
WIRED has the first external account of the company and the work: https://t.co/tP8nItCsDl
full writeup, the setup, and all the numbers: https://t.co/C9NZ5vtV1V
FYI if you’re like PhD level good at math, you’re criminally under leveraging it in a software company if you don’t flex on people with it regularly.
Being able to confidently whiteboard simple calculus even makes 90% of genpop terrified of you.
Next time someone says “stochastic” in a meeting write up a stochastic equation on the board. You’ll never have to convince people of an idea again they’ll be so afraid of getting embarrassed by you.
92 Johnson Solids
In 1966, Norman Johnson found 92 convex polyhedra made only of regular polygons that didn't belong to any known family. He guessed that was all of them.
In 1969, Victor Zalgaller proved that there are exactly 92.
Is anyone in the market for a 700-year-old Templar Knights watchtower?
Because there's one for sale in the Lot valley, France. 7 bedrooms across 6 floors, a frescoed dining room, spiral staircase and views over a peaceful medieval village in the Lot region.
There's also a 2-bedroom stone cottage in the walled courtyard, an enclosed garden and a heated swimming pool.
Asking price: €550k ($634k).
The Templar Knights who built this probably didn't picture the pool. But here we are.
for years, society was limited to only 16 syrup squares per waffle but with recent combinatorial optimization breakthroughs our research department has achieved previously unheard of densities of waffle syrup
Я сегодня закоммитил правки в документацию CatBoost. Теперь могу считать себя ML-разработчиком? Какие у вас критерии в профессии?
Там в формулах вместо \log было log, невозможно такое читать.
@alisameowmeowmi Возможно, перекошенные выборки:
– в среднем ты засыпаешь на диване, когда совсем устала и капец надо выспаться, откуда сон лучше качеством;
– в среднем засыпаешь на кровати, когда просто пора спать, и качество сна обычное.
@sokobath@xamgore Предыдущие спонсоры (в частности, JB) не имеют отношения к новому проекту, и с нашей стороны было бы неправильно присваивать названия их проектов: CSClub и CSCenter.
Мы решили придумать своё название, новое и одновременно очевидным образом отсылающее к существовавшим.