Mehdi Ataei

Physics AI @ NVIDIA

[Mehdi Ataei @AtaeiMe]

@tszzl GPUs are so cheap for what they deliver

[Mehdi Ataei @AtaeiMe]

@ajambrosino @TheRealAdamG I really wish in codex app we had more than 273k context. Compaction is great but not perfect.

[Mehdi Ataei @AtaeiMe]

New World Order meant recession I guess

[Mehdi Ataei @AtaeiMe]

@vasuman @thsottiaux we need a better monitoring system

[Mehdi Ataei @AtaeiMe]

@tszzl @garrytan Best explained by Welch Labs https://t.co/0d2HSxVxhk

[Mehdi Ataei @AtaeiMe]

GPT-5.5 is really something special. I just wish it was a bit more creative, even for science and coding stuff. It kind of feels like a model where pass@1 and pass@32 are basically the same, probably because it’s been heavily RL’d.

Opus is definitely ahead in creativity.

[Mehdi Ataei @AtaeiMe]

@SandraCobena_ So what’s the consequence here? Nothing? Like always.

[Mehdi Ataei @AtaeiMe]

I didn’t know if it was physically possible to spend this much on food tbh

[Mehdi Ataei @AtaeiMe]

@JustinBleuel @ChatGPTapp You still cannot select the reasoning level on chatgpt app for enterprise.

[Mehdi Ataei @AtaeiMe]

@SebastienBubeck Harness = asking the model to continue

[Mehdi Ataei @AtaeiMe]

@tszzl You’re doing more but you less productive

[Mehdi Ataei @AtaeiMe]

Heard a lot of good things about TMLR submitted two papers there and haven’t heard back for months. What’s going on

[Mehdi Ataei @AtaeiMe]

@jxnlco Launch in tmux when working over ssh. The job shouldn’t stop if you close your laptop

[Mehdi Ataei @AtaeiMe]

@thsottiaux We’re bored release the next model

[Mehdi Ataei @AtaeiMe]

@cloneofsimo The fact that he has 223k followers and gets like 8 likes

[Mehdi Ataei @AtaeiMe]

@scaling01 One of the biggest cons ever walked on earth. I also wonder how his NVIDIA’s shorts doing.

[Mehdi Ataei @AtaeiMe]

@VictorLevoso @FriesIlover49 😂

[jayce @jayce_edits]

@GaryMarcus @LotusAbhi @ASM65617010 very interested to see how you'll cope with this one gary, like genuinely im excited to see what you'll come up with

[Mehdi Ataei @AtaeiMe]

@SebastienBubeck Where the hell is Gary

[roon @tszzl]

on some level if you want civilization to ascend to a new level you need your AIs to do things that are not legible to you and maybe not even strictly obey you, in the same way that if you hire a great new ceo you give them a lot of autonomy to transform the company according to their own plan, even one which may not immediately read as a winning strategy (imagine the board of directors of Apple firing and rehiring Steve Jobs years later - except the board of directors are chimpanzees)

all else equal, companies and organizations that hand more of themselves over to machine intelligence will outcompete ones that demand the corrigibility and legibility tax of human oversight and human design. it is not a stable equilibrium and requires some sort of vast cooperation scheme if you’d like to enforce it

real asi alignment has to operate at a deeper level than oversight, control, or human corrigibility