AI robots take over lab benches, rewriting science

AI robots – In an autonomous lab overlooking Boston harbor, Ginkgo Bioworks uses robotics and AI to translate experimental designs into instructions for machine systems—then, in a newer twist, has an AI model draft the experimental “recipe” itself. Founders say the approa

The first time Reshma Shetty saw a lab notebook entry written by a model, it landed like something out of reach that suddenly became real.

She’s a co-founder and COO of Ginkgo Bioworks. a company that has built an autonomous laboratory where AI and robotics increasingly replace the classic bench-and-pipette workflow. On a tour. her voice still carries the surprise of it: “The really. really wild moment was the first time I saw a lab notebook entry written by the model. ” she said.

Inside the lab, robots line the room like one-armed machines displayed behind glass. A large screen at the front shows a color-coded schedule for the day. Equipment moves along a track that runs through the space—designed to look. at least from a distance. like an oversized toy train set—delivering materials from one robot to another.

Ginkgo Bioworks uses these systems on a range of work across pharmaceuticals, agricultural projects, and government contracts. Current assignments include engineering microbes for better fertilizer and creating proteins that will make snow or ice. Much of the lab’s effort is aimed at pharmaceuticals. and Kelly points to the evidence of living biology in the workflow—gesturing at a petri dish being ferried from one robot to another and saying. “that has actual live cells in it.”.

The company’s promise rests on a specific shift: scientists use AI to translate experimental designs into instructions for robots. It’s the difference between drafting plans and watching machines carry them out.

Jason Kelly. another Ginkgo co-founder. frames that trajectory as a long bet that started with four graduate students from MIT nearly two decades ago. “We believed that programming cells would ultimately be more important than programming computers. ” Kelly said. recalling the early days when the idea was treated as far-fetched.

Gene editing and testing new molecules could take many hours in the laboratory, he explained—mixing hundreds of chemical cocktails by hand and pipetting them into petri dishes, with a huge amount of human labor behind every attempt.

To change that, Kelly said the group built a company to replace those human lab workers with robots. But early investors weren’t convinced. Kelly remembers running their startup on “ramen. ” buying equipment on eBay. and “we could not raise venture capital. ” a reminder that the tech revolution didn’t arrive in their orbit on schedule.

Everything changed after the AI boom. In 2014. Kelly said he read a blog post from Sam Altman—roughly a year before Altman went on to found OpenAI—about the potential to automate biotechnology the same way he imagined automating other technologies. The two began talking. Kelly recalled that he felt grateful and practical at the same time: “I was like. man. thanks for this blog post. We’ve been around for five years. It is impossible to raise money.”.

Eventually, Silicon Valley money started flowing.

Today, Kelly runs Ginkgo Bioworks with his former classmates. He describes the lab as a place where human scientists oversee robotic versions of themselves—robots executing the experimental work while people focus on design and decision-making.

Now the lab is experimenting with another step: not just using AI to convert plans into instructions, but asking the model to participate in the thinking that usually belongs to human scientists.

Shetty described a recent collaboration connected to OpenAI. Working through ChatGPT, she said they challenged the bot to create a certain protein. The typical setup. she explained. is to write the mental and experimental groundwork yourself—like writing a recipe—and then hand it to a robot to execute. In this experiment, they asked the bot to write the recipe.

“We had no idea if it would even be able to make protein,” Shetty said.

Their results were strong enough to measure. Shetty said they concluded the protein synthesis showed a 40 percent reduction in costs compared with human work. She also said the system ran more than 30,000 experiments in six months. They have published the results, though the paper has not been peer reviewed.

Both Shetty and Kelly emphasized that humans still need to provide the right questions and constraints for experiments. Even in the most automated workflow. the point is not that robots replace scientific judgment—it’s that machines can take more of the procedural burden. making discovery faster and potentially cheaper.

That shift is already changing Shetty’s own practice. “Normally. I rush through designing my experiment because I need to get it done so that I can actually do all the pipetting in the lab and set it all up. ” she said. “Now, she says, she spends more time designing her experiments so that the robot can do them for her overnight.”.

If that sounds like a trade—more time upfront, less time trapped in repetitive lab work—it also explains why some researchers see urgency in the other side of the equation.

Drew Endy. a bioengineering researcher at Stanford. warned that these new freedoms could open the door to people with little to no training in science running experiments with questionable goals. Endy and colleagues recently wrote a report illustrating ways artificial intelligence could be used to mass-produce viruses or create other biosecurity threats.

He described the split in his own feelings clearly: “I’m thrilled about AI and science right now as a researcher,” he said, but he is worried about risks including potential bioweapons programs in other countries. “I’m not excited about that.”

Endy said regulations and policy to mitigate these dangers are within reach—but they need to be prioritized well in advance of a biotechnological disaster or warfare. He pointed out that biotechnology has often been insulated from these risks through intellectual gatekeeping. saying. “Biology has traditionally been hard for people to really gain control over.” In his view. AI could “nudge it a little bit more towards concentration of power.”.

Kelly, for his part, thinks the technology is pushing toward democratized access—whether society is ready for it or not. “I do think you’ll have a culture clash,” he said, “coming of what happens when everyday people can ask scientific questions.”

For now. the lab’s promise is visible in the choreography: petri dishes moving from robot to robot. schedules driving experiments. and AI models writing parts of the scientific workflow that were once locked behind human hands. The question running beneath it all is whether the same automation that accelerates discovery will also accelerate the wrong kind of capability—faster than safety can keep up.

AI-powered robots autonomous laboratory Ginkgo Bioworks robotics in biotech ChatGPT protein synthesis bioweapons risk biosecurity Drew Endy Jason Kelly Reshma Shetty MIT graduate students