Today at Answer.AI we've got something new for you: FSDP/QDoRA. We've tested it with AI at Meta Llama3 and the results blow away anything we've seen before. I believe that this combination is likely to create better task-specific models than anything else at any cost. 🧵

Huge news for anyone working in tech in the US. Noncompetes are now banned: not just in California (like before), but nationwide. Very, very relevant for anyone at Amazon (which is the Big Tech that has enforced noncompetes even for low-level engineering positions).

Excited to share our new Nature paper! In this work, we propose a new display design that pairs inverse-designed metasurface waveguides with AI-driven holographic displays to enable full-color 3D augmented reality from a compact eyeglasses-like form factor. 1/8

Introducing Genie 2, our latest protein design model! Genie 2 sets a new state of the art on key design metrics and supports motif scaffolding, including that of multiple motifs with undefined geometric relationships, a feature new to protein diffusion models. (1/5)

Discovering tens of thousands of new antibiotics from #AI of proteins of extinct animals! nature.com/articles/s4155… @NatBME @delafuenteupenn Penn Medicine Penn Bioengineering Marcelo Torres Jackie Peng

Introducing CHARM: a new epigenome editor to methylate DNA at the promoter of a targeted gene. Our lab's collaboration with Jonathan Weissman's Lab's Edwin Neumann & Tessa Bertozzi shows deep silencing of brain PrP Paper: science.org/doi/10.1126/sc… Blog: cureffi.org/2024/06/27/int…

Woohoo gpu.cpp has now been officially released! This is a really exciting project which provides a device-independent way to use GPU compute. No more writing separate code for CUDA, AMD, Mac, and Intel GPUs! We can't wait to see what you all are able to create with this.

AlphaFold3 reproduced and params/code released. 🤩

We used PaperQA2 to extract claims from papers and then see if they're contradicted anywhere in literature. This task is time consuming for humans, but we were able to use this for hundreds of papers to look for trends in disagreement in fields, decades, and journals.

Go to Settings → Privacy and Safety → Data Sharing → Grok to turn it off.

A new optimizer that's better than Adam. I think I've heard that before. But, I tried this one, I was doing regression testing on some recent optimizer cleanup w/ real scenarios, I threw this in the mix. It did beat Adam, every time (so far). This one appears worth a closer look.

The Evo 1 paper is now published!

Excited to report that we have discovered a new, endogenous molecular clock in unmodified human cells and tissues. It is ticking away right now in almost every cell in your body. Specifically, in a new paper on BioRxiv, we show that RNA editing by the near-ubiquitous ADAR1 in

20+ labs just worked together to open source Genesis - a physics engine with a VLM agent that turns text prompts into interactive 4D worlds Think instant physics-accurate environments, camera paths, and character animations - all from natural language 🤯

Today, we release QwQ-32B, our new reasoning model with only 32 billion parameters that rivals cutting-edge reasoning model, e.g., DeepSeek-R1. Blog: qwenlm.github.io/blog/qwq-32b HF: huggingface.co/Qwen/QwQ-32B ModelScope: modelscope.cn/models/Qwen/Qw… Demo: huggingface.co/spaces/Qwen/Qw… Qwen Chat:

I like that framing - ULMFiT is the "last common ancestor". :D That's not a coincidence BTW. It was ULMFiT that first showed how fine-tuning a general purpose language model can achieve SoTA results. (All today's frontier models are fine-tuned general purpose language models.)

My colleague Thomas's story highlights the most remarkable aspects of the human spirit - perseverance; of not giving up. We still have only a limited understanding of Nature. This is why we Google DeepMind are working on advancing Science- it is the most human thing to do.

It is insanely optimistic that we have drugs that can reduce Lp(a) levels by 98% in trials right now. If they succeed, they could help to crush heart disease and stroke.

The plan at FutureHouse has been to build scientific agents and use them to make novel discoveries. We’ve spent the last year researching the best way to make agents. We’ve made a ton of progress and now we’ve engineered them to be used at scale, by anyone. Today, we’re launching

How Sebastian Ruder @ ACL made the ULMFiT paper happen 😂 (from 2018 fast.ai student notes: cedrickchee.gitbook.io/knowledge/cour…)