What's this about?
Latent Space interviews Andy Beam and Rafa Gómez-Bombarelli of Lila Sciences (Jul 16), on the thesis that "the lab of the future should feel like a data center" — wet-lab science run the way we run compute: automated, parallelized, scheduled by software, with AI deciding which experiments to run next.
What exactly is an "AI-run wet lab"?
A wet lab is where physical experiments happen — pipettes, cells, reagents. The emerging model pairs robotic lab equipment with AI agents that design experiments, execute them via automation, read the results, and iterate — turning science from artisanal benchwork into a throughput problem. It directly attacks what DeepMind's essay this week called the validation bottleneck: hypothesis generation is getting cheap; real-world testing is now the constraint.
Why does it matter?
If experiments scale like compute, the bottleneck DeepMind identified starts to crack — and "self-driving laboratories" become the pick-and-shovel play of AI-for-science. (No deep transcript summary today — transcript tooling unavailable.)