Lucas Grosjean (@xo_lucas_13) 's Twitter Profile
Lucas Grosjean

@xo_lucas_13

Physics/Photonics @EPFLEngineering alum. prev @CNRS, @Google/@TheTeamAtX

ID: 40636970

linkhttps://github.com/lucasgrjn calendar_today17-05-2009 09:58:52

325 Tweet

141 Followers

1,1K Following

Pedro Cuenca (@pcuenq) 's Twitter Profile Photo

This is huge. Apple just released a Core ML conversion tool and inference pipeline for Stable Diffusion. Their code is inspired by diffusers and a pleasure to read. We converted the weights for you in the Hub. Go play with them! huggingface.co/blog/diffusers…

This is huge.

Apple just released a Core ML conversion tool and inference pipeline for Stable Diffusion. Their code is inspired by diffusers and a pleasure to read.

We converted the weights for you in the Hub. Go play with them!
huggingface.co/blog/diffusers…
Andrej Karpathy (@karpathy) 's Twitter Profile Photo

Great video on helion fusion. Few thoughts: - "no steam turbine" umm SOLD :) - triggers my hard tech envy for natural sciences, sometimes feel deep learning is not that deep - how can systems like chatgpt++ help accelerate this kind of work? how "intelligence constrained" is it?

elvis (@omarsar0) 's Twitter Profile Photo

How much can you get out of training a language model on a single consumer GPU in one day? Results attained in constrained setting: decent downstream performance on GLUE. Performance closely follows scaling laws observed in large-compute settings. arxiv.org/abs/2212.14034

How much can you get out of training a language model on a single consumer GPU in one day? 

Results attained in constrained setting: decent downstream performance on GLUE. Performance closely follows scaling laws observed in large-compute settings.

arxiv.org/abs/2212.14034
Bojan Tunguz (@tunguz) 's Twitter Profile Photo

If you think that ChatGPT is awesome, just wait to see ChatZero - a chatbot trained from scratch by having thousands of concurrent instances BS to itself for millions of GPU hours.

Tamay Besiroglu (@tamaybes) 's Twitter Profile Photo

Recent applications of deep learning in science and engineering, such as AlphaFold and Copilot, have been astonishing. What does standard economic growth theory say about the economic effects of its adoption in R&D? We sketch a simple picture: arxiv.org/abs/2212.08198

Jean de La Rochebrochard (@2lr) 's Twitter Profile Photo

Nos fondateurs reçoivent actuellement des emails frauduleux qui se font passer pour l'équipe de Kima. 1) Je ne vouvoie que mes grands-parents 2) Mes emails sont beaucoup plus laconiques 3) C'est à nous qu'on demande de l'argent, pas l'inverse

Nos fondateurs reçoivent actuellement des emails frauduleux qui se font passer pour l'équipe de Kima.

1) Je ne vouvoie que mes grands-parents

2) Mes emails sont beaucoup plus laconiques

3) C'est à nous qu'on demande de l'argent, pas l'inverse
Michael Black (@michael_j_black) 's Twitter Profile Photo

PhD students, don't worry. Technologies, trends, and even whole fields come and go. A PhD makes you an expert in a field but, more importantly, teaches you how to become an expert. Once you know that you can learn anything, you can adapt to major disruptions in your field.

Jean de La Rochebrochard (@2lr) 's Twitter Profile Photo

Once upon a time, there was a founder. Their vision was grand, but their bank account was empty. Their background was solid, so much so that they could design amazing slides. Their ambition was stellar and reflected beautifully in their pitch. Their fundraising was

Jim Fan (@drjimfan) 's Twitter Profile Photo

What if we set GPT-4 free in Minecraft? ⛏️ I’m excited to announce Voyager, the first lifelong learning agent that plays Minecraft purely in-context. Voyager continuously improves itself by writing, refining, committing, and retrieving *code* from a skill library. GPT-4 unlocks

Jenia Jitsev 🏳️‍🌈 🇺🇦 🇮🇱 (@jjitsev) 's Twitter Profile Photo

(Yet) another tale of Rise and Fall: DeepSeek R1 is claimed to match o1/o1-preview on olympiad level math & coding problems. Can it handle versions of AIW problems that reveal generalization & basic reasoning deficits in SOTA LLMs? (arxiv.org/abs/2406.02061) 🧵1/n

(Yet) another tale of Rise and Fall:  DeepSeek R1 is claimed to match o1/o1-preview on olympiad level math & coding problems. Can it handle versions of AIW problems that reveal generalization & basic reasoning deficits in SOTA LLMs? (arxiv.org/abs/2406.02061) 🧵1/n