Listen to full episode :
What if the messiest part of your data is trying to tell you something? Spreadsheets and databases rarely make headlines, but they run our hospitals, banks and offices. In this episode we sit down with Gaël Varoquaux (Inria, Probabl), co-founder of scikit-learn, to talk about what happens when AI learns to read them.
Gaël argues that cleaning data is not as neutral as it sounds: every time we squeeze a complex world into a simple model, we distort it. Along the way we explore table foundation models, why synthetic data made them possible, and why teaching people to frame a problem may matter more than teaching them tools.
We also talk about why we should still learn to code, why AI safety looks to Gaël more like industrial safety than science fiction, and why he and his colleagues refuse to sell hype – while still being proud of what Europe does well.
Plus, our pub quiz: how did an experimental quantum physicist end up in AI? The answer might surprise you. Listen all the way to the end!
USEFUL LINKS
More about Gaël Varoquaux: https://gael-varoquaux.info/about.html
scikit-learn: https://scikit-learn.org
Probabl: https://probabl.ai
jingle track (get it) provided by mobygratis.