e
@yihyunCS
Followers
2K
Following
14K
Media
906
Statuses
5K
schizo-shitposter the child asked me if god exists. i said 'not yet.' prev @nasa & @nvidia
Seoul
Joined February 2024
“we’re compute bound” they say thinking they’re correct. WRONG try again bitch we’re bound by memory bandwidth keep pretending you know what you’re talking about
1
0
7
I actually think this is relatively solvable from a single dataset curation tbh
New SWE/ML Leaderboard just dropped, like WeirdML but with a human baseline Turns out, all LLMs slow you down compared to human experts in ML/HPC optimization tasks (measured by runtime)
0
0
2
the first sign of baby agi is emotional emergence, this is me on record saying this
0
0
1
stop falling for rust psyops rewriting in rust is a waste of time, rust doesnt make it better than your c++ because its rust, its better because you rewrote it.
0
0
0
longest TLDR i’ve ever seen, diabolical stuff really
Yesterday we did a livestream. TL;DR: We have set internal goals of having an automated AI research intern by September of 2026 running on hundreds of thousands of GPUs, and a true automated AI researcher by March of 2028. We may totally fail at this goal, but given the
0
0
5
L larp tbh
We're doing all kinds of stuff with these models that the public isn't even thinking of yet.
1
0
7
rag with extra steps, you can scroll now
Holy shit...Google just built an AI that learns from its own mistakes in real time. New paper dropped on ReasoningBank. The idea is pretty simple but nobody's done it this way before. Instead of just saving chat history or raw logs, it pulls out the actual reasoning patterns,
3
0
14
no one deserves to die over things they say and opinions they have, that’s all i will speak on the matter
0
0
21
Well sorry to insist, but the catch is that the market is still small — that's less than chewing gum industry. Too small for a bubble.
12
4
62
gemini is the best model purely due to its ability to be critical and evaluate things with lower bias, some models (OAIs in particular) are vulnerable to agreeing with the user even when incorrect about an issue, gemini seems to not fall into this as much (deepseek too)
1
0
3