Ok the real proof GPT-4 is an improvement over previous versions: it can generate correct PyMol commands on the first go

How to RLHF #LLAMA if you don't have hundreds of GPUS? Do it in a parameter-efficient way. I'm happy to finally share our parameter-efficient fine-tuning #PEFT survey! It took quite a bit more time to make than I expected, but I feel good about the result arxiv.org/abs/2303.15647

If you're in NYC, come hang out with techbio crowd tonight! Entertainment includes yours truly talking about protein design for like 5 minutes

Had a lot of fun helping Casper Goverde and Martin Pacesa with designing novel proteins! Check out the explainer thread by Martin Pacesa

Not a competition.

Бостон - это Академгородок, который смог.

My experience going from tech to biology Industry: cool, you can do ML and software, but dude, you don’t have any biology background, sorry Academia: bro you wanna do ML and work with us for the money we can pay? we will mentor and give you tons of opportunities to do cool stuff

DALL·E 3 prompt "protein structure, surrounded by small molecules resembling amino acid sidechains"🧑‍🎨🖼️ (1/10)

1/ We’ve submitted a letter to President Biden regarding the AI Executive Order and its potential for restricting open source AI. We believe strongly that open source is the only way to keep software safe and free from monopoly. Please help amplify.

Excited to share the work with Zhidian Zhang Hannah Wayment-Steele Garyk Brixi haobo wang Matteo Dal Peraro Dorothee Kern biorxiv.org/content/10.110… (1/7)

Use ML to design soluble versions of integral membrane proteins. Martin Pacesa Casper Goverde Nico Goldbach Justas Dauparas Sergey Ovchinnikov Simon Kozlov Bruno Correia

The true rest is giving up ambition, even for a bit.

Nature speedrun Any%: completed. Hard carry by Casper Goverde Martin Pacesa Bruno Correia Sergey Ovchinnikov and others!

Someone please ask for applying code snippets from chat to work in Jupyter Notebooks. I’m begging you.

Very counterintuitive: Walking or just doing nothing is hard because it feels like a waste of time, but unblocks both creativity and productivity. Scrolling through social media on the other hand is mostly waste of time, but is easy because it doesn’t feel so.

I needed to use R library from Python and had to use r2py for the first time (what a mess). Had to fix 10 subtle conversion errors, and most of them got fixed just by asking Sonnet to fix them. My main contribution except copy-pasting errors: ask to run on a toy dataset first.

Salvador Dali, "Using Roblox Studio for the first time"

What could Alphafold 4 look like? (Sergey Ovchinnikov, Ep #3) 2 hours listening time (links below) To those in the (machine-learning for protein design) space, Dr. Sergey Ovchinnikov (Sergey Ovchinnikov) is a very, very well-recognized name. A recent MIT professor (circa early