#scigen

Scigen MIT

MIT Unveils SCIGEN, an AI Tool to Speed Up Quantum Material Discovery

MIT SCIGEN and Google DeepMind developed a new tool that uses geometric and physical principles in generative AI models to build dependable and useful materials for quantum computing and other cutting-edge technologies.

SCIGEN, developed by MIT researchers, improves generative artificial intelligence to design new, stable quantum materials, advancing materials science. One problem with AI-driven discovery is that models often yield “hallucinations” of chemically unstable or physically impossible structures. This innovation addresses it. SCIGEN ensures that its materials are promising for crucial applications like enhanced electronics, quantum computing, and clean energy by directly integrating physics-based design concepts into the AI's creative process.

Combine AI Creativity with Physical Reality

Traditional generative AI for materials design often makes irrational and power-hungry recommendations. Structural Constraint Integration in a Generative model (SCIGEN) software layers common AI diffusion models. It protects structures from violating user-defined guidelines like atomic distances and lattice symmetries throughout production.

“The models from these large companies generate materials optimised for stability,” says research senior author Mingda Li, a Career Development Professor at MIT's Class of 1947. Materials science rarely progresses that way. One great item can change the world, not ten million. This strategy emphasises choosing materials with desired attributes over quantity. MIT, Google DeepMind, Emory, Michigan State, Oak Ridge National Laboratory, and Princeton University participated on the Nature Materials study.

Quantum Leap in Material Creation

SCIGEN was tested by generating Archimedean lattices. These structures produce exotic quantum phenomena like quantum spin liquids and “flat bands,” which are needed to make error-resistant quantum computers.

The results were impressive:

The system generated almost 10 million pattern-matching candidate materials.

About one million candidates passed the initial stability tests.

41% of 26,000 structures selected for high-fidelity simulations on Oak Ridge National Laboratory supercomputers had the predicted magnetic behaviours.

Most importantly, the group created two novel compounds in the lab, TiPdBi and TiPbSb, whose properties matched the AI's predictions.

By creating several materials like that, experimentalists have hundreds or thousands more candidates to expedite quantum computing materials research, said Princeton University scientist Robert Cava.

broader implications and future directions

Although quantum technology is the focus, SCIGEN has many applications. Materials with certain properties may affect renewable energy technologies like carbon capture and energy storage and biomedicine, including novel antibiotics. It represents a shift from serendipity to precise engineering in material discovery.

There are still issues. Smaller labs cannot use the technology because it demands enormous datasets and processing power. Researchers also stress that AI-generated materials must be experimentally verified to prove their feasibility and predictability.

100+ articles in peer-reviewed science journals are gibberish

Scholarly journals are peer-reviewed, i.e., submitted articles are sent out to members of the author’s professional peer group — experts in the same field who are similarly trained and credentialed — for their evaluation. For example, an article on a China-related subject would be sent to sinologists for review; while an article on quantum physics would be sent to quantum physicists for review.

Publishers withdraw more than 120 gibberish papers : Nature News & Comment

The publishers Springer and IEEE are removing more than 120 papers from their subscription services after a French researcher discovered that the works were computer-generated nonsense.

Over the past two years, computer scientist Cyril Labbé of Joseph Fourier University in Grenoble, France, has catalogued computer-generated papers that made it into more than 30 published conference proceedings…

The publishers Springer and IEEE are removing more than 120 papers from their subscription services after a French researcher discovered that the works were computer-generated nonsense. Over the past two years, computer scientist Cyril Labbé of Joseph Fourier University in Grenoble, France, has catalogued computer-generated papers that made it into more than 30 published conference proceedings between 2008 and 2013. Sixteen appeared in publications by Springer, which is headquartered in Heidelberg, Germany, and more than 100 were published by the Institute of Electrical and Electronic Engineers (IEEE), based in New York. Both publishers, which were privately informed by Labbé, say that they are now removing the papers. Among the works were, for example, a paper published as a proceeding from the 2013 International Conference on Quality, Reliability, Risk, Maintenance, and Safety Engineering, held in Chengdu, China. (The conference website says that all manuscripts are “reviewed for merits and contents”.) The authors of the paper, entitled ‘TIC: a methodology for the construction of e-commerce’, write in the abstract that they “concentrate our efforts on disproving that spreadsheets can be made knowledge-based, empathic, and compact”. (Nature News has attempted to contact the conference organizers and named authors of the paper but received no reply; however at least some of the names belong to real people. The IEEE has now removed the paper).