Will Hua

Ex @mcgillu @Mila_Quebec. PhD dropout @Yale.

Building for post-AGI.

[Sebastien Bubeck @SebastienBubeck]

yes, nonsofic groups exist: this statement is one of many new beautiful results proved by Astra, our next major model.

We're releasing 10 such Astra proofs, complete with lean certificates and CoT walkthroughs for each of them. The results are wide-ranging, from von Neumann algebras (disproof of Connes' Rigidity Conjecture) to better bounds for high dimensional sphere packing, for circuit complexity, for monochromatic triangles in multicolored graphs, and more.

More thoughts here: https://t.co/8SjXONeh38

[Will Hua @itswillhua]

@jasondeanlee Is it the official name of gpt6?

[Will Hua @itswillhua]

@anshulkundaje But you believe at end of the day, at some point, we are able to build the first virtual cell? Do you think LLMs can accelerate the process from decades to years?

[Will Hua @itswillhua]

@DdelAlamo Yeah bro is so funny

[Will Hua @itswillhua]

难绷😂

[Jürgen Schmidhuber @SchmidhuberAI]

@FT Unfortunately, the Aplhafold paper was based on earlier work that was not cited: https://t.co/p8naxHZtsr

[Will Hua @itswillhua]

This is amazing!

[Bowen Jing @bjing2016]

It’s getting hard to keep track of all the new open-source cofolding models coming out. To help everyone stay up to date, I'm excited to share https://t.co/Gc9PaF6GoY, a PDB-synced leaderboard of open models, updated weekly. All predictions are available to view & download! (1/4)

[Will Hua @itswillhua]

@nanli @RecursionChris Unbelievable!

[Bohan @loubohan]

Everyone in the Valley has a take on whether China is winning AI, robotics, and biotech. Almost nobody understands how the capital works

Notes from meeting China's tier-1 investors and founders on how their venture ecosystem actually works. https://t.co/Dii6Oeqn1A

[ali @waterloo_intern]

https://t.co/aivc7B4A7Z

[Will Hua @itswillhua]

I thought LVMH also signed the open letter for open-weight models 😂

[Bernard Arnault @BernardArnault]

J’ai été très sensible aux nombreux messages que vous m’avez adressés à la suite de ma lettre ouverte au Monde.

Je tenais à vous en remercier toutes et tous, directement et personnellement, ici, sur X.

Merci!
Bernard Arnault

[Will Hua @itswillhua]

@sokrypton It’s hard…from professional perspective, you should respond. But as a respected researcher, I don’t think you need to respond to AI generated reviews.

[Will Hua @itswillhua]

@RegenbaumShaun Great idea!

[Will Hua @itswillhua]

We made a really good protein design model, validated by wet-lab already.

Now, we want to run in-silico tests, what kind of computational validation you want to see?

Leave a comment, thanks!

[Will Hua @itswillhua]

@wizardofid It has to be tuned in a certain way, from data construction to post-training methodology.

[Will Hua @itswillhua]

I have three questions:

1. Can LLMs do protein folding?

2. If LLM is the form of AGI, is protein folding a necessary intelligence?

3. If biology or science is the next topic after coding, do we need LLMs to understand protein folding?

Should scientific AGI understand folding?

[Andrej Karpathy @karpathy]

@lakshyaag weird misinformation to find circling on twitter, no.

[Will Hua @itswillhua]

Haters will hate, but why hate? I don’t understand.

[Julian Schrittwieser @Mononofu]

I’m so excited that @JensenHuang is a believer in open source now, looking forward to the CUDA and GPU driver open source release!

[Will Hua @itswillhua]

@DrSusanNasif @SeqeraLabs Well, it’s easy to set independent, leakage-controlled test set for co-folding. But for de novo designs, how?

[Will Hua @itswillhua]

@NielsRogge Exactly.
We are just saying open-source is good; the community needs some high-quality open-source weights. Not asking Anthropic to open-source their models…but they just hate.

[Will Hua @itswillhua]

@CalvinMccarter Good idea, will do!

[Will Hua @itswillhua]

@biocheMichael Now it’s mAbs, but since we will open-source the model: a new OpenDDE checkpoint, the community can test its capabilities to design enzymes, mini binders, and more.

[Will Hua @itswillhua]

@lhmartin1 @adaptyvbio Awesome. Will do!