A New York startup has built a so‑called self‑driving laboratory in which robots synthesise and test materials while artificial intelligence selects the next experiment to run, according to a video report by the Bulletin of the Atomic Scientists. The company, Radical AI, has raised more than $65 million and lists the US Air Force among its clients as it pursues advances in batteries, semiconductors and heat‑resistant alloys.
What a self‑driving lab does
The model replaces many routine experimental tasks with automation. Robots handle the physical work—mixing chemicals, fabricating samples and running measurements—while machine learning algorithms analyse results and propose subsequent experiments. The system is designed to accelerate discovery by iteratively narrowing experimental choices toward promising directions without a human manually planning each step.
Proponents say this approach can increase throughput, reduce human error and search large parameter spaces more efficiently than conventional lab workflows. Critics raise ethical, epistemological and practical questions: Can an algorithm truly exercise the kinds of judgement and creative hypothesis‑formation that human scientists provide? Who is accountable when automated systems make decisions that influence costly research or military applications?
Who is involved and what they are seeking
Public reporting notes the following about the company and its work:
- Company: Radical AI (startup operating the self‑driving lab in New York City).
- Funding: More than $65 million raised.
- Clients: The US Air Force is listed among them.
- Research aims: materials for batteries, semiconductors and heat‑resistant alloys.
| Item | Reported detail |
|---|---|
| Funding raised | $65 million (reported) |
| Noted client | US Air Force |
| Research focus | Batteries, semiconductors, heat‑resistant alloys |
“Can AI do science without scientists?”
Questions for policy, industry and the public
The emergence of self‑driving labs touches several policy and societal debates. Regulators and research funders will need to consider:
- Standards for validation and reproducibility when experiments are planned by opaque algorithms.
- Ethical frameworks and oversight for applications with national security implications, given military clients.
- The future of scientific labour: which tasks are likely to be automated and what skills scientists will need.
Automation in laboratories is not new, but the combination of high‑throughput robotics with adaptive AI that decides experimental direction represents a step change. It prioritises optimisation and speed through closed‑loop learning—algorithms proposing experiments, robots executing them, and results feeding back to the algorithm. That feedback loop can explore vast combinations of conditions more quickly than teams planning experiments manually.
Yet speed is not the same as understanding. Scientific progress rests on more than generating material with desired properties; it also involves forming explanatory theories, recognising anomalous results, and translating findings into reliable knowledge. Those intellectual tasks remain difficult to automate, and public reporting on Radical AI frames the company’s work as pushing at the boundaries of what machines can do in research.
For South African researchers and institutions, the trend is worth watching. Automated discovery platforms could accelerate materials research relevant to energy storage, semiconductor development and high‑temperature engineering—areas with direct commercial and strategic interest here. At the same time, adoption raises questions about investment priorities, workforce training and collaboration models between universities, industry and government.
As the technology develops, clear standards for evaluation, openness about data and methods, and policies that protect scientific integrity will be essential. The report on Radical AI provides a window into a future laboratory model; whether it will augment scientists’ capacity or displace key aspects of scientific judgement remains an open and important question.
