I was thinking about the labor market congestion problem and came up with a solution that is often used in service marketplaces: pay to apply. Then I asked myself what this solution has that AI doesn't. That’s how I arrived at the analogy that prices act as model weights: they encode market information. An important difference: prices incorporate signals from dispersed, hard-to-observe data that an AI/ML model may not access.
P.S. Paying to apply may sound provocative and require thoughtful consideration and careful testing. Payments can be made with platform-issued virtual points, available in a limited supply. But here, I focus on why price signals may address this problem better than AI-based screening.
Prices as model weights: learning from dispersed data · HackerTrans
P.S. Paying to apply may sound provocative and require thoughtful consideration and careful testing. Payments can be made with platform-issued virtual points, available in a limited supply. But here, I focus on why price signals may address this problem better than AI-based screening.