#预测市场 Looking at the data from this AI prediction market on Kalshi, I have to say, this wave of changes is quite painful. At the beginning of the year, Gemini's probability was only 30%, now it has skyrocketed to 86%. ChatGPT's probability dropped from 41% to 8%—isn't this just the market voting with real money? The trading volume has reached $14.08 million, indicating that someone is really placing bets here.

But I want to remind everyone not to be fooled by these numbers. Prediction markets may look scientific, but in reality, they are still a game of human strategy. Who are the market participants? Are retail investors following trends? Are big players manipulating the market? These are all hidden variables. I've seen too many things that seem like "objective data," but in the end, it's all about those with large capital having the final say.

Moreover, the settlement standard is the LM Arena leaderboard, which is inherently problematic. How the leaderboard is tested and what scenarios are tested will all influence the results. In other words, if someone can predict or influence the evaluation direction of the leaderboard in advance, then this prediction market is no longer a prediction but a gamble.

My advice is to look at the data, but don't use your own money to gamble on it. These types of prediction markets are essentially harvesting information asymmetry. The smart approach is to observe trends, not to participate in gambling.