The chemists argue that the initial experiment didn't take into account 'compositional disorder', a feature of inorganic materials, and that the AI used to interpret the experimental data didn't perform well. They believe that the results don't necessarily cast doubt on the GNoMe algorithm itself, but rather on the process of determining the novelty of the compounds. The chemists suggest that the paper should be retracted as they believe none of GNoMe's new materials appear to have been produced yet. Google DeepMind declined to comment, while a lead author of the original paper disputed the chemists' findings and stated they will respond through peer-reviewed literature.
Key takeaways:
- A group of chemists from Princeton University and University College London are disputing the results of a study by Google DeepMind and UC Berkeley that used a robotic lab system, A-Lab, to synthesize new materials predicted by AI algorithms.
- The chemists argue that A-Lab did not create any novel inorganic material as claimed, and that most of the materials have been misclassified.
- The errors are believed to have stemmed from the use of AI to computationally determine whether a new material had been made, a process the chemists claim is unreliable.
- Despite the dispute, the chemists believe that if some of the inorganic crystal structures predicted by Google DeepMind's model GNoMe were successfully synthesized, it would result in a novel material.