If AI could interact and learn from the physical world, could it make more scientific advances?
We had GPT-5 optimize molecular cloning protocols in the wet lab. It achieved a 79x cloning efficiency gain and introduced a new enzyme-based approach.
Cloning protocols are important for protein engineering, organism engineering, and genetic screens. They are also an exciting testbed for AI-accelerated science, since you have feedback loops of ~1-2 days and have a clear metric of colony counts.
We partnered with Red Queen Bio to introduce an evolutionary framework where GPT-5 proposes a batch of changes to the Gibson Assembly protocol, gets the results of each change, and proposes anew. It did surprisingly well.
While humans acted as GPT-5’s hands for carrying out the protocols, we also piloted an autonomous robot. It was built to execute arbitrary Gibson cloning protocols from natural language, with human supervision for safety.
Notably, GPT-5 proposed a new enzymatic procedure that adds two proteins: RecA and gp32. While they have been studied together biochemically, to our knowledge, this is the first time they've been functionally co-used in a cloning method.
To be clear: this is not a bio breakthrough. But it is a novel optimization, and perhaps at the level of a competent PhD student for this task. And it was surprising to us: when we first set out, we thought a 10x gain would be an impressive achievement.
Read our early write-up here:
We hope to continue accelerating scientific advances.openai.com/index/accelera…
Share this Scrolly Tale with your friends.
A Scrolly Tale is a new way to read Twitter threads with a more visually immersive experience.
Discover more beautiful Scrolly Tales like this.