Jyunmi Hatcher

A self-driving lab goes further.

It closes the loop.

The AI doesn't just execute experiments.

It decides which experiments to run in the first place based on what it's learned from the results of the previous experiments.

It forms hypotheses.

It designs tests.

It performs them, then analyzes the data, and then uses what it's learned to design the next round.

The human sets the goal, say, find a material that conducts electricity but bends like plastic, and the machine figures out how to get there.

The analogy that researchers use is the one in the name, self-driving cars.

A regular car with cruise control follows the speed you set.

A self-driving car perceives the road, makes decisions, adjusts routes, and responds to conditions in real time.

The jump from automated lab to self-driving lab is the same jump, from following instructions to making decisions.

Ross King, a computer scientist at the Chalmers University of Technology in Sweden, is one of the pioneers in this field.

And he puts it plainly.

A lot of biology and chemistry is still done like craft work.

A principal investigator with postdocs and students operating like an artisan and apprentices.

Self-driving labs are more like a production line.

Science, King says, will be done differently, like in a factory.

So what these labs are actually doing right now.

The Nature feature profiles several systems that are already operational and the range is fairly striking.