AI-Khimik at center of scandal: breakthrough or failure?

Artificial intelligence (AI) controlled laboratory assistant, claiming to have created 41 new materials, has sparked controversy in the scientific community. The assistant, featured in Nature magazine, promises to accelerate the search for materials used in batteries and electronics. However, critics argue that the results are misleading.

Chemist Robert Palgrave from the University College of London is calling for the article to be retracted, stating that there is insufficient evidence to support the claim of creating new materials. Palgrave believes this work highlights the importance of maintaining high research standards when utilizing AI.

Leslie Schup from Princeton University also critiques the assistant’s analysis of the materials, although she acknowledges the positive aspects of the laboratory’s work, such as efficient recipe compilation and labor savings through robotics.

Herbrand Seder, the team leader from the Lawrence Berkeley National Laboratory, supports the assistant’s work. He acknowledges that the AI’s analysis may not be on par with human expertise but asserts that it provides a rapid way to demonstrate the potential for material creation.

The primary criticism revolves around the analysis of data using porch x-ray diffraction (PXRD), which the laboratory employs to determine material structure. Palgrave notes numerous remaining data points, rendering the algorithm unreliable in material identification. Furthermore, doubts arise regarding whether the added metal atoms truly result in new materials.

Seder acknowledges the novelty of some materials but agrees with Palgrave that the assistant mistakenly characterized certain “open” products, which had actually been previously created by chemists using traditional methods. Seder promises to update the article accordingly.

The dispute over the PXRD data highlights the challenges faced when employing AI in the discovery of new materials. Andy Cooper from the University of Liverpool asserts that this does not invalidate the concept of an AI-driven laboratory but highlights the substantial work that lies ahead.

Despite the disagreements, both Palgrave and Shup agree that the concept of autonomously discovering materials deserves attention. Palgrave states, “This could be a significant breakthrough in materials synthesis.”

/Reports, release notes, official announcements.