AI Found 100 Insights. How Do I Pick?

Hack Session

About the session

My last talk ("RIP, Data Scientists") argued that AI was taking over the data scientist's tasks. This year we hit a different problem - abundance.
When I pointed agents at India's messy government data, they found over a hundred insights in an hour. Some were wrong, many boring, but some were front-page material. How do I pick?
Now that analysis is cheap, choosing is the bottleneck.
In this session I run the pipeline live. Agents hunt for hypotheses in raw public data, write code to test them, discard duplicates and statistical trivia, attack the survivors, and cite evidence for what remains. Then (the part AI can't do yet) we humans will pick one and put our name on it. I'll show the insights that looked good but died, and how we caught them.
You'll leave with the methods and SKILL.mds to go from raw data to one publishable insight, and a workflow you can run on your own / corporate data.

Speaker

Download Brochure