Ricardo Avila
ricardoavila.bsky.social
Ricardo Avila
@ricardoavila.bsky.social
Bioinformatics engineer
Writing about biology, code, and too many hobbies. 🌿📸💻🧬
Don’t give them ideas 😭
March 9, 2025 at 7:29 PM
Reposted by Ricardo Avila
Demo from the paper:
ProtChatGPT: Towards Understanding Proteins with Large Language Models https://arxiv.org/abs/2402.09649 🧬🖥️🧪
January 29, 2025 at 4:30 PM
Another thought is that it might be interesting to equip such a language model with a sandbox like PyMOL, and have it play around with the structure in order to answer questions!
January 30, 2025 at 11:49 PM
This paper looks interesting! But I feel like the example question/answers in the examples are so general that they could be answered with a simple Google search. I'd be curious to know if the model can answer things like "What happens if we mutate this specific residue from Glycine to Arginine?"
January 30, 2025 at 11:47 PM
To see the steps O1 is doing, you can click on the "Thought for X seconds" text. I believe it is not the full chain of thought, like R1's, but an abridged version.

I tried a few more, and it seems that coming up with the riddle takes longer than the answer for me: chatgpt.com/share/679be5...
January 30, 2025 at 8:50 PM
I tried it with O1 and it did pretty well after I explained and clarified the rules. I guess correctly, but I found later that if I had checked the “thought process” I would’ve seen the answer there.
January 30, 2025 at 6:34 PM