Chemically clumsy AI and a fantastical catalyst killer

Artistic AI imagines atoms

Several Newscriptsters have noticed our social media feeds filling up with whimsical pictures produced by Dall-E Mini, which was recently renamed Craiyon. This artificial intelligence generates images from language prompts such as “sharks painted by Van Gogh” or “Queen Elizabeth II shopping at Target.” To the delight of Twitter users around the world, many of the results are passable—or at least entertaining. So science Twitter did as science Twitter does and tested the AI’s scientific savvy. “I was curious to see if it had some kind of chemical understanding,” says computational scientist Victor Fung. He tried seeing if Craiyon could generate images of crystal structures and aromatic carbons. “I wasn’t expecting very much,” he tells Newscripts. Rightly so. The AI crunched numbers in a minute of silent contemplation, but the best it could do was produce networks of painted orbs that bear a passing resemblance to atoms in ball-and-stick models, Fung says.

This Newscriptster got similar results. Craiyon also does not understand the concept of chemical safety posters or molecular orbitals. It seems to have learned about research laboratories exclusively through marketing stock images with colorful liquids in Erlenmeyer flasks. And while the AI has at least gleaned that the periodic table is composed of colorful squares, most attempts to replicate this iconic chemical image inexplicably involve oil paints.

Chemists could consider going easy on Craiyon. It is a simpler version of a much more powerful language-learning AI called Dall-E 2, created by the compay OpenAI. Though Craiyon is accessible at any time by anyone with an internet connection, there’s a long wait list for time with Dall-E 2. Newscripts will keep readers updated once we’ve had our turn.

Many heads make catalysts halt

Catalysts may help expedite reaction rates, but these capable compounds deactivate faster than most chemists might like. There are myriad ways for the catalytic compounds to sputter out, but Javier Pérez-Ramírez could fit only 10 of them on this 12-headed illustrated dragon. Pérez-Ramírez, a chemical engineer at the Swiss Federal Institute of Technology (ETH), Zurich, who has an affinity for catalysis, wanted to create an image that would make it easy for chemists to understand the many processes that deactivate catalysts, such as coking, sintering, and leaching. Catalysts can lose their mojo by multiple deactivation processes at the same time, he tells Newscripts. “That’s why the concept of a same body with different heads was a way to show that there is an interplay between mechanisms,” Pérez-Ramírez says. He worked with two artists over the course of several weeks to refine the deactivating dragon before sharing the final form in a Twitter thread.

Pérez-Ramírez is well known among his colleagues for blending art and science in educational illustrations and evocative journal covers. When he spoke to Newscripts via videoconference, Pérez-Ramírez didn’t need a virtual background because his office wall looks like a museum collection of scientific illustrations. He works with artists around the world to create digital graphics and epic paintings that he loves to include in eye-catching presentations to students. “We should bring to science more emotions,” Pérez-Ramírez says. “And the way to bring emotions is through art because art never leaves people indifferent.”

He hopes the cartoon brings attention to the fascinating fates of catalytic materials and shows young people that science can be both rigorous and fun.

Please send comments and suggestions to newscripts@acs.org.