We are less than a week from the virtual hackathon for #LLM applications in materials science and chemistry. Here is some inspiration from other's applications to drive you!
@nc_frey investigated neural scaling in large chemical models with 1B+ parameters & 10M datapoints. Explored physical priors and scale interplay, plus potential for prompt engineering. #ChemGPT encoder
ChemNLP is a new open community effort collecting chemistry datasets and exploring LLM applications in chemistry. You can join the Discord today, and start collaborating with the community.
Ethan Mollick shows an example using Bing to design some trendy new watches. Really, you could just scroll through @emollick's entire timeline for LLM application inspiration. 😄
Great story from @Mechanophore and colleagues @UofIllinois. From skeptic to believer: the power of models. Let's take a look at some of these journeys and what we can learn.
@Mechanophore's story - seeing models make intriguing predictions into reaction behavior. "A key turning point for me was seeing how FEA of Frontal Ring-Opening Metathesis Polymerizations simulates autonomous curing processes by balancing reaction enthalpy with heat transport"
@MrBSussels Story - modeled complex reaction mechanisms using a simplified graphical abstraction to create a system of coupled linear equations and analytically solved them to provide an empirical rate law in agreement with the experimental observations. Skeptic → Believer
It's been a wild week already in #ML/#AI for science. Advancements in using diffusion models for protein folding, using learned potentials to discover new catalyst materials, a proposed battery data genome to speed energy storage material discovery, and so much more! 🧵 (1/8)
@KevinKaichuang et al use diffusion models to generate novel foldable protein structures. (2/8)
@_akhaliq presents NeuralPLexer to predict protein-ligand structures. This will help understand the interaction between small molecules and proteins. Also with diffusion models ✨ (3/8)