Jyunmi Hatcher

This is, of course, where AI comes in, right?

So normally, without the assistive AI, you'd have to have a whole team of experienced data scientists to just build the code pipeline before you can even look for patterns or notice anything that's significant in the data.

And then you have to build the analysis tools.

And so there's lots of bottlenecks through that traditional process.

So a few years ago, UCSF, they tried to solve these bottlenecks by crowdsourcing.

So they launched something called the DREAM Challenge.

And that stands for Dialogue on Reverse Engineering Assessment and Methods.

So pretty much they come up with the acronym first and then fit what words are going to make the acronym.

It's basically a global competition where data science teams from around the world compete to build the best prediction models.

Over 300 teams participated and then more than 100 groups submitted models trying to predict preterm birth from the data.

Most teams finished their analysis within about three months.

So here's the thing.

It took nearly two years to pull all those results together, validate everything and actually publish the findings.

Two years from competition to publish paper.

Now, here's the AI experiment part.

So, of course, research asked what the obvious next question is.

Can AI do this faster?

So this is what they did.

They took eight different generative AI chatbots, think chat GPT kind of things, and gave them the exact same data sets from those dream challenges.

No human coding.