Latent Space spoke with Andy Beam and Rafa Gómez-Bombarelli of Lila Sciences about a vision of laboratories that operate more like data centers. The idea is to treat scientific experimentation as a scalable data-generation system.
That matters because AI for science is limited not only by models, but by the quality and volume of experimental data. Robotics and automation could help produce datasets that are more structured and useful for training scientific AI systems.
The conversation highlights a growing frontier for AI infrastructure: the next major training corpus may come from automated labs rather than the public internet.