Verifiable Environments for AI in Biology — Kenny Workman, LatchBio

Summary

Kenny Workman, cofundador de LatchBio, explica cómo están aplicando agentes de IA al campo de la biología, un área que genera volúmenes masivos de datos experimentales. Al considerar el código y los flujos de trabajo de bioinformática como un sustrato ejecutable, LatchBio ha desarrollado benchmarks como 'Spatial Bench' para evaluar y mejorar la capacidad de los modelos de IA en tareas científicas complejas. El enfoque enfatiza la importancia de la verificabilidad y la necesidad de post-entrenamiento especializado para que las IAs realicen análisis biológicos fiables que superen las capacidades humanas básicas en el descubrimiento de fármacos.

Chapters

1. 0:00LatchBio's Vertical AI Lab
2. 3:17Evolving to Agent Prototypes
3. 5:57Benchmarking Spatial Biology Data
4. 7:48SpatialBench Design and Properties
5. 9:56Human Verification, Long Horizon Tasks
6. 13:20Rubrics and Future Expansion

Topics covered

• Crecimiento de datos biológicos
• Agentes de IA en biología

+5 more topics (full content when you follow the channel)

Transcript excerpt

Thank you to the organizers for having me. I'm one of the co-founders and CTO at Latch. We are basically a vertical AI lab for benchmark and agent engineering. I'm hoping to motivate and explain exactly what that means today. Starting directly with motivation for agents in in bio generally. Um Many people in my domain are familiar with this curve, but this is basically the log linear curve of data generated over the years in in biology. And the reason I'm bringing it up it will become directly important to the kinds of things we want to do in engineering. Um This curve is driven by a very small handful of experimental classes. One is called single cell biology. This is where we split up cells, break them apart, and measure their RNA. The second is spatial biology, which will become the focus of the next segment of the talk. Same thing as single cell, but you get spatial resolution. You c…