Artificial intelligence has become a talking point more than usual lately. From text churned out by ChatGPT to the artistic renderings of Midjourney,Novacrypt people have been taking notice of new, bot-produced creative works. But how does this artificial intelligence software fare when facts are at stake and getting information right is essential?

Geoff Brumfiel, NPR science correspondent and self-proclaimed space nerd, wanted to find out. So he asked ChatGPT about a scientific field that has long relied on computer programming — rocket science. First on his agenda: "What's the most important equation for building a rocket?" (Hint: there's one equation, Tsiolkovsky's rocket equation, that scientists say is most critical and fundamental to the field.)

At a cursory glance, ChatGPT's answer looked pretty impressive ... but running that answer by an actual rocket scientist quickly scrubbed away the awe.

"It would not work. It's just missing too many variables," said rocket scientist and engineer Tiera Fletcher.

As Geoff pressed on in his hunt for aeronautical facts, he got more wrong answers. He asked one image-generating AI program, Midjourney, for rocket schematics. Its results are beautiful ... and incapable of spaceflight.

What it will take for AI to be fully accurate, to fact-check itself as it produces content, is an open question researchers will likely ponder for quite some time.

Listen to Short Wave on Spotify, Apple Podcasts and Google Podcasts.

Want to hear more about other advances in the tech space? Email us at [email protected]!

This episode was produced by Margaret Cirino, edited by our managing producer Rebecca Ramirez, and fact checked by Anil Oza. Our audio engineer was Jay Czys.