You don't need to be in the room to run a real lab experiment anymore. Jonathan Green talks with Keith Robison, a computational biologist of 35 years who now works on lab automation at Ginkgo, about the company's autonomous lab: reconfigurable carts you assemble like Lego, an agent that turns scientific intent into a protocol, and a run you can launch, leave, and check from your phone. They talk about what that means for students and amateurs, and what still gets in the way, from slow, expensive DNA synthesis to instruments that were never built for robots.
Key Takeaways:
- Keith has run an experiment on the autonomous lab from 30 miles away. A scientist builds a protocol with an agent, loads the reagents, and pulls all the execution data afterward.
- Ginkgo's Cloud Lab offers free design pipelines for protein engineering. You send in the design, Ginkgo runs it on the autonomous lab, and the data comes back.
- The real barriers are cost and access, not lab skill. DNA synthesis is still slow and expensive, and equipment like surface plasmon resonance is rare outside pharma labs.
- Instruments were built for humans. Protocols often live on the machine under a name with no version control, so every change needs review to be sure you get exactly what you asked for.
- Vendor APIs are the bottleneck. Agents reviewing public APIs keep finding bugs and documentation gaps, such as a sleep function where the examples say milliseconds but the underlying code counts seconds.
Notable Quotes:
"I've run an experiment on the system from 30 miles away." — Keith Robison
"You no longer need to be skilled in the lab, and you can switch places very quickly." — Keith Robison
Connect with Keith Robison:
LinkedIn: https://www.linkedin.com/in/keithrobison
Ginkgo Cloud Lab: https://cloud.ginkgo.bio
Company: Ginkgo Bioworks — https://www.ginkgo.bio
Enjoyed this? Follow The Artificial Intelligence Podcast and send it to anyone who still thinks real experiments are out of reach.