On Data Integrity

Since the very first day that ChatGPT appeared, it was clear to me that AI would change the world of data.

I take data seriously. I think extensively about the validity of data, the ways we can use data, and, most frequently, the ways that AI is completely overhauling what data means.

I've been obsessed with data and statistics since I was a 12-year-old introduced to the world of baseball. What started as my hobby eventually became my career. First taking advantage of lucrative sign-up bonuses granted by gambling sportsbooks, a nice side-hustle for a university student, then as a proper professional sports gambler, then as a prediction markets trader, now as a market maker across prediction markets and crypto.

When data is so crucial to my well-being, it's hard for it to not matter. This project, Mapdis, represents one of my most deeply considered thoughts, that the use of AI opens up a new frontier of data, superior to the human-generated data that existed before it. Data that can be generated off of subjective AI assessments, at scale.

I understand that there is a perception that AI is somehow "lesser" than a skilled human, and that AI-generated data somehow implies "slop."

AI output is not intrinsically slop.

AI is a tool, a tool that has proven the ability to solve some of the hardest unsolved problems in Mathematics. It contains the cumulative insight of trillions of hours of human thought and research, expressed across trillions of weights, and it is the job of any competent AI practitioner to extract from those weights insights about the world. On verifiable domains, such as math, this is easy. In the world of social science, this is considerably more difficult.

Difficult does not mean impossible.

When used correctly, AI contains brilliant insights, above and beyond what a human can produce. It is my responsibility to use this exceptional tool, which can conduct a year's worth of advanced research in mere hours, to produce digestible data about the world of geopolitics.