OpenAI Co-Founder Predicts “Full Lights Out Operations” for Future Wet Labs

As biological data moves from the petri dish to the petabyte scale, how will the role of the researcher change? In a recent fireside chat at the Precision Medicine World Conference 2026, Vinod Khosla, founder of Khosla Ventures, and Greg Brockman, president and cofounder of OpenAI, discuss a future where AI handles the hypothesis-to-result loop, potentially redefining what it means to be a "scientist."

What does the rise of AI tools mean for the future of the wet lab?

Greg Brockman, President and Co-Founder of OpenAI

With AI, we are starting to see the leading edges of new knowledge creation. For example, recently, a group of physicists applied our unreleased model to a physics problem that they have been working on for a year. And the lead here was someone who was previously an AI skeptic. He was amazed that 12 hours later they actually had a solution to this quantum physics problem.

In the future, I think we are gearing towards fully robotic wet labs and being able to go for full lights out operation. Biology is like a foreign language to us. We didn't evolve to grow up to understand ATGC or to understand the language of proteins. But to AI models, it's just another language. I think we have started to see these kinds of results in different ways. There's systems like AlphaFold that are able to predict the folding of proteins. I actually spent some time training DNA foundation models, and the results there feel kind of like GPT-1 level results, where you can basically take this model and apply it to essentially any downstream biological task, maybe with a little bit of fine tuning, and it gives you some interesting predictions. I think that we're going to see these models that can deeply understand biology in a way that humans just cannot, like we just are limited in terms of our ability to go deep. And the other thing that's very interesting is that humans are limited in our ability to go wide. If you are doing a PhD in biology, you're going to be studying like one specific reaction, one specific pathway, for years. And AI can actually know about all these pathways, all of these reactions, and look for insights across them.

Suddenly, a 13-year-old kid can be a scientist and do novel work that today you'd have to go through a whole PhD program to even have the tools to start on. I think it's going to be this massively democratizing force where we're going to have breakthroughs coming from all over the world in ways that we would be very surprised by today.

But what can humans uniquely do that is going to be very hard for us to ever get a model there? It's very hard for a model to actually bear responsibility for some action. It kind of feels like at the end of the day, you do need a human there. The point is that you can build anything on top of these technologies you can imagine. The true potential of AI is to unlock human imagination and improve the quality of life for everyone.

Vinod Khosla, Founder, Khosla Ventures

I envision a more blowout kind of change where AI is doing most of the hypothesis, then running a lab and testing the hypothesis, and then doing the next iteration of the hypothesis. Only the goal is specified by human beings, but the AI is doing all the invention and discovery. So, I don't imagine a lab that has humans in it.

Just recently, I've seen researchers who had data for three years that their graduate students were pouring over to try and get a paper out of. And I have seen an unreleased model take that data and, in a few hours, almost completely replicate what the grad students had done over three years. If you had a million people playing around, or 100,000 people playing around with hundreds of thousands of chemical syntheses or whatever, you generate a lot of data in the real world. And that real-world data being distilled back into the models creates this vicious positive loop of really understanding reality.