ARTICLE AD BOX
Alibaba’s DAMO Academy has developed an AI research tool called Elements Claw, and it’s already making waves in materials science. With Elements Claw, scientists recently found four new superconducting materials that real-world lab tests later confirmed. This is another sign of how AI is speeding up the discovery of new materials.
Their research paper explains that the AI sifted through about 2.4 million crystal structures—yep, that’s million—in just 28 GPU-hours. It narrowed things down to around 68,000 possible superconductors. Out of those, researchers managed to make and confirm four new compounds in the lab: Hf₂₁Re₂₅, Zr₄VRe₇, HfZrRe₄, and Zr₃ScRe₈. The best one of the bunch superconducts at 6.5 Kelvin.
So, how does Elements Claw actually work? Alibaba says it blends large atomic foundation models with big language models to judge crystal structures, sort the candidates, and recommend which materials scientists should test in the lab. Along the way, it also re-found some superconductors we already knew about, which reassured the team that their method works.
They’ve made their materials database open for academic researchers, so now anyone can dig deeper or build on what they’ve started.
Here’s the thing, though: don’t get too excited about practical superconductors just yet. These new materials only work at very low temperatures—not even close to what we need for, say, lossless power grids or room-temperature quantum gadgets. The real breakthrough here is speed. In the past, searching for new materials could take years of simulations and experiments. With AI, scientists can zero in on the most promising options way faster.
This effort is part of a bigger trend—AI is now a major force across materials science, from batteries to semiconductors and beyond. High-temperature superconductors are still out of reach, but as AI gets better, it keeps pushing the field forward, cutting down both the time and money needed for early-stage research.
English (US) ·