Ingmar Schuster

@fpedregosa Nice word play

My favorite blog is bach!

@nathanbenaich So the business model for open source AI has to be around deployment/consulting of the model then I guess?

Wikipedia vs. Encarta is an analogy I’ve come back to many times re: competitive dynamics in AI. While neither Wikipedia nor Encarta were actually good money printing businesses, it’ll be different for AI, clearly.

@geoffreyhinton AI is not an existential threat, and way way way less dangerous than climate change.

Andrew Ng is claiming that the idea that AI could make us extinct is a big-tech conspiracy. A datapoint that does not fit this conspiracy theory is that I left Google so that I could speak freely about the existential threat.

Europa kann nicht zusehen wenn Antisemitismus wieder um sich greift.

@josipK Intellectually you're right. I think an "I stand with Israel" serves mostly emotional purposes and thus has its place. If somebody gets beaten up, their friends, kids, parents murdered and raped, they need emotional support. Then you can come back to principles of course

All this pro-Israel and pro-Palestine support gatherings (or anti-Israel or anti-Palestine protests) are wrong. It's *not* about taking sides. Taking side is *not* helping bring peace. It's about supporting principles: pro-peace, pro-prosperity, anti-terror and anti-destruction.

@SamuelAinsworth This is super great! So basically it's glue code connecting the torch API with a Jax backend?! I'm astounded it takes so few lines of code

Announcing torch2jax! Run PyTorch code natively in JAX. 🤝 Mix-and-match PyTorch and JAX code with seamless, end-to-end autodiff, use JAX classics like jit, grad, and vmap on PyTorch code, and run PyTorch models on TPUs.

@ApitiusHofmann @SwenRoschlau @wolfejosh @envedabio @viswacolluru @ReskiLab @ddomingof Yes it's interesting for sure. The "molecular sequencing" tech seems great and we also usher into molecules currently

@SwenRoschlau @wolfejosh @envedabio @viswacolluru @IngmarSchuster @ReskiLab The Director of Data Science of Enveda is @ddomingof , former PhD student of mine. He took a couple of really good colleagues to Enveda. So, easy to mediate contacts … 😎

The Chemical Code via Lux family co @envedabio + founder/CEO @viswacolluru linkedin.com/posts/envedabi…

@fpedregosa Almost all recent Belkin papers are worth a read

@IngmarSchuster Haven't read that one yet. Thanks for sharing!

Surprising property mentioned to me by @gowerrobert about the stochastic Polyak step-size variant below is that the iterate error is monotonically decreasing with just (star) convexity. Note decrease is _not_ in expectation! What other stochastic methods have this property?

@fpedregosa 🤔 this is interesting indeed. Conjecture that large mini-batches are required goes against the advice "as small as possible batches" in this other cool work: arxiv.org/abs/2306.04815 which is about SGD in particular for NN generalization

Ref: arxiv.org/abs/2307.14528

@levi_nz @sc_codeUM @tipiirai Rather they're offended by people who fight over the name because they have nothing better to do

@andrewgwils @m_finzi Do you plan to replace the linear operators from gpytorch with CoLA?

We're ecstatic to officially announce our new library, CoLA! CoLA is a framework for large-scale linear algebra in machine learning and beyond, supporting PyTorch and JAX. repo: github.com/wilson-labs/co… paper: arxiv.org/abs/2309.03060 w/ amazing @m_finzi, Andres Potap, Geoff Pleiss

📺 youtu.be/tWnoYYriJEk 🎧 open.spotify.com/episode/2p3cVi… Make sure to click the subscribe button! In the next episode, we feature Kristof Szalay of Turbine, a company that revolutionized the drug discovery process by simulating human cells.

We also delve into the question of AI biases and whether machines are conscious by now. 📺 Now streaming on YouTube and across all your podcast platforms.

🎙️ Machines, Molecules and …? We're back with another episode of "Machines and Molecules," where we explore the intersections of science, AI, and human experiences.

@ronawang And texted him?

saw a dating app profile that listed the guy’s SAT score

@syhw This is great Gabriel! Any plans to make this easily usable in IDEs/code editors?

I'm very thankful for everybody from the team who worked on this in the past few months, and thankful for all the partners who helped with the release. 7/7

Happy to be releasing Code Llama! We've built it on Llama 2 and improved it for code use cases. In particular it supports infilling out of the box, and was trained with sequences up to 16k tokens. Looking forward to what the community will build with it! 1/7

🔈 podcasters.spotify.com/pod/show/machi… 🎞️ youtube.com/watch?v=fv4JvS…

We nerd out about "all models are wrong, but some are useful," by talking about model misspecification, and discuss how FX has grown during his academic career.

🎙️ Machines, Molecules and …? We're back with another episode of "Machines and Molecules," where we delve deep into the intersections of science with machine learning and stats, learning from a prominent researcher in the field.