csgm
2.3K posts
csgm
@csgbwk
french husband, robotics, ai, computer graphics
France Katılım Nisan 2024
312 Takip Edilen69 Takipçiler
csgm retweetledi
guy who has been completely sedentary and hasn't left the house: why am I so sleepy and anxious?
English
The amount of LoD makes everything bland, in real life it would look a lot more lush and green.
Jackson@jayveeonYT
Yeah, I don't think I've ever played a game that does this
English
@VictorTaelin @_inception_ai @ArtificialAnlys I think Neogen can fix RL/synthetic data for diffusion LMs... Will be experimenting when bend2 comes out..
English
@_inception_ai @ArtificialAnlys this approach enormous potential, I suppose with a larger run this would approach the intelligence of the leading models?
English
Mercury 2 is in a league of its own.
1,200 tok/s at comparable quality to speed-optimized autoregressive models, per @ArtificialAnlys.
English
@IterIntellectus Even easier, terraform earth in places where it's annoying, we have most of the tech we just need political will and scaling
English
it’s terrifying that people still don’t fully realize you can terraform planets
Curiosity@CuriosityonX
It’s terrifying that people still don't fully realize there is no PLANET B. This is it.
English
@VictorTaelin Could it be that Anthropic and OpenAI trained on your own input for the past 3 years, which would be why they outperform so much?
Also impressive to see Gemini perform so well on all benchmarks even private and still being completely unusable in real life
English
Introducing LamBench . . .
You asked me to make a benchmark, so I made it. It is a simple, old style Q&A consisting of 120 fresh λ-calculus programming questions. Some are easy, like "implement add for λ-encoded nats". Some are harder, like "derive a generic fold for arbitrary λ-encodings".
It measures:
- intelligence (% tasks completed)
- elegance (BLC-length of solutions)
- speed (completion time)
Basically what I care about, other than long context.
I made it today because I was excited about GPT 5.5.
It didn't do too well ):
(My first-day impression is that I can't tell the difference between GPT 5.5 and GPT 5.4. I would be lying if I said otherwise. I'd not be able to distinguish in a blind test. I need more time. It is much faster though.)
This is a new, simple bench, so expect be bugs.
Specially on OpenRouter models. I'll retest soon.
Also, it was born saturated. V2 will be harder...
↓ Link and more charts below ↓
English
csgm retweetledi
Celui de l’X est à gauche sur la photo
Quantum optics enjoyer@shpostx
Mon stagiaire de polytechnique sympathise avec celui de polytech Nice
Français
csgm retweetledi
New Paper🙂
Nanite has shown that small triangles can be rendered fast in compute, we're exploring how fast for large meshes with up to 18.9 billion triangles, without the need to precompute LOD structures.
Paper: github.com/m-schuetz/CuRa…
Source: github.com/m-schuetz/CuRa…
English
"Well we can do an internship in gaussian splatting if you want, but i can't give you a subject yet since it will probably be solved by June"
PhD_Genie@PhD_Genie
Watching a high-impact paper getting published on your exact thesis topic notion.
English
Extremely disappointed with what I initially thought was a promising team. But this is just really tasteless.
Atelier Missor@AtelierMissor_
This 50 ft Prometheus will soon be shipped to the USA. It costs around 1 million to build. We want to build larger and larger Prometheus statues everywhere across the West.
English
@zuhaitz_dev I kind of got out of the compilation field sadly :/ I chose computer graphics and robotics, but could have gone either way tbh
English
I'm so glad I only wrote compilers in Ocaml and Rust ts looks hard in C 😭
Zuhaitz@zuhaitz_dev
For those who think that's bad, you should write a compiler ;p
English