Jyunmi Hatcher

So the honest caveat here is that these systems are still, in most cases, making incremental improvements rather than fundamental breakthroughs.

Asking the knowledges, the hypothesis of self-driving labs that they can currently test are relatively constrained.

Optimizing a compound in a drug assay for turning a material for a battery or solar panel.

They're exceptionally good exploring the defined search space, but they're not yet proposing the kind of conceptual leaps that reshape entire fields.

The technology is still in its iterative phase.

Lila Sciences, for all its scale, still relies on human input to validate AI predictions before moving to full autonomy.

Periodic Labs describes the process as deliberately incremental, automating pieces that make sure the AI's proposed experiments make sense before trusting it with the whole pipeline.

But the trajectory seems to be clear and extends well beyond any single lab.

The U.S.

Department of Energy has launched Forum AI, a four-year, $10 million project led by Berkeley Lab in collaboration with Oak Ridge, Argonne, MIT, and Ohio State.

to build the first full-stack, general-purpose AI platform for material science research.

At Argonne National Laboratory, a self-driving lab called PolyBot is already producing high-conductivity electronic polymer films using AI-driven decision-making.

Across the field, the infrastructure for autonomous science is being built at the national level.

Then there's the question that Nature posed in a separate piece in February.

Will self-driving labs replace biologists?

The consensus from researchers is no, but the rule changes.

The scientist becomes less of a bench worker and more of a strategist, defining the problems worth solving and interpreting results within a broader scientific context.

Ross King compares it to the shift from artisan workshops to industrial manufacturing.

The craft doesn't disappear, but the scale and spread of production transforms what's possible.

For institutions without massive budgets, the potential is significant.