#Polymarket预测市场 Seeing the probability shifts of AI large models in the Kalshi prediction market, I feel a bit emotional. At the beginning of the year, ChatGPT still held a 41% chance of winning, while Gemini only had 30%. Now, in December, the situation has completely reversed—Gemini skyrocketed to 86%, and ChatGPT dropped to 8%. The $14 million trading volume indicates this is no small matter; real money is voting.

This reminds me of the ICO boom in 2017. Back then, the market sentiment shifted just as quickly, and the flow of funds was equally decisive. The difference is that back then, we were betting on the imagination of the future; now, bettors are wagering on tangible model performance—using the LM Arena leaderboard as the final settlement basis, which means the game rules are relatively transparent.

Historically, the most interesting aspect of prediction markets is that they often reflect turning points in market perception in advance. The rise in Gemini’s probability should be a comprehensive reflection of recent months’ technological progress, user feedback, and market testing. ChatGPT’s fall from an absolute leader to being suppressed is a stark contrast, worth deep reflection—how a once-dominant player loses its voice. This is not the first time in tech history, nor will it be the last.

What we should truly be cautious about is not the accuracy of individual predictions, but what kind of information these market participants are basing their judgments on. If most people are making decisions from the same information source, then prediction markets are essentially amplifiers of emotion rather than machines of truth.