AI Generative Models Drive Rapid Progress in Drug Discovery

@Latent_Labs Of course, they have an agent Latent-Y that can do all sorts of tasks: antibody design and interaction with Latent-X2, fetching info from literature. It gives up to 50x speedups #cvpr2026

However progress in the past 5 years has accelerated. Simon gives a historical review #cvpr2026

The AlphaFold2 architecture in a nutshell: a beautiful mix of AI and domain experience poured there, eg. the evoformer block uses a tweaked attention to take into consideration the structure and interactions in proteins #cvpr2026

Protein mutation analogy to computer vision: as if a single wrong pixel could break an image #cvpr2026

@Latent_Labs Utimately these predictions are tested in wet labs and Latent-X1 did well #cvpr2026

Now a clear and quick intro to proteins (what are they, their structure, how they work). Checking badge … yes, we’re still at #cvpr2026

The decades-long protein folding problem and how crazy the AlphaFold series of models performed #cvpr2026

Behold the prediction of Latent-X1, the first foundation model of @Latent_Labs #cvpr2026

An analogy between how Imagenet ultimately enabled generative models and how the CASP challenge and then Alphafold let to generative protein design #cvpr2026

@Latent_Labs Much effort is invested in making producing drug-like generation but also improving screening efficiency with genAI #cvpr2026

For the generation, the protein binders are key and to model that they treat atoms as point clouds (it’s still worth studying 3d and point clouds, folks) #cvpr2026

@Latent_Labs A highly optimistic view for the next 10 years. Go go @Latent_Labs! #cvpr2026

@abursuc Scaling laws go woooo