Jyunmi Hatcher

They just gave the AI systems carefully written prompts, natural language instructions and said, hey, analyze this data and build us a prediction model.

So they paired the UCSF master student named Ruben Sarwal with a high school student, Victor Tarka, from Huron High School in Ann Arbor, Michigan.

These two working with AI tools generated functioning computer code in minutes.

The same code that would normally take experienced programmers hours or even days to write.

So, you know, this is what we've been seeing in the AI space.

Like, oh, AI, I didn't know how to code before.

Now I can, you know, I can make my own software.

Hmm.

So...

So they stacked the AI results versus the 300 plus expert teams.

Four of the eight AI chatbots produced usable prediction models.

And of those four, some of them matched and in certain cases actually outperformed the models built by human expert teams who had spent months on the same data.

The entire AI powered project from the very first prompt to submitting the finished paper to a journal took six months.

And that's in comparison to the nearly two years for the human-only version.

So six months versus two years.

Two-person team with AI versus hundreds of specialist teams without it.

Now, there is a few caveats, right?

Half of the AI tools failed.

Four out of the eight did not produce usable code.

So that's a coin flip.