In a unique experiment testing artificial intelligence’s predictive capabilities, ChatGPT participated in a competitive forecasting contest against 160 human experts—and landed squarely in the middle of the pack. The annual challenge, run by economist David Seif, required contestants to assign probability percentages to roughly 30 future events across politics, business, science, and pop culture for 2025. When hedge fund director Sam Leffell entered ChatGPT’s predictions, the AI model ultimately finished 80th, performing no better than the average human contestant. However, Seif noted that ChatGPT actually fell short of a key benchmark: a baseline score of 75,000 points achieved by assigning a neutral 50% probability to every event. ChatGPT scored 82,925 points, meaning it added “negative value” overall. Its performance was mixed—strong on questions with abundant existing data, like predicting the FIFA Club World Cup winner, but weak on time-sensitive or news-dependent forecasts, such as estimating the return date of astronauts from a delayed mission.
Despite its middling ranking, the experiment revealed compelling nuances about AI’s role in forecasting. Leffell highlighted that ChatGPT understood complex, multi-page contest rules in seconds and produced all 30 predictions in minutes—far faster than humans, who often spent hours or days researching. This speed and scalability suggest that for high-volume, rapid forecasting tasks, AI could be immensely practical, even if its accuracy is merely average. Leffell argued that if judged by “results per minute of work,” ChatGPT might be considered a winner. As an investor who assesses countless probabilities, he now views tools like ChatGPT as essential, noting their potential to handle thousands of predictions efficiently, even when perfect accuracy isn’t required.
The contest ultimately underscored both the limitations and the transformative utility of AI in prediction. While ChatGPT struggled with qualitative judgment, breaking news, and events lacking historical data, its ability to quickly parse rules, apply statistical reasoning, and generate baseline forecasts points to a supportive rather than superior role alongside human intuition. For now, AI appears to be a powerful augmenting tool—table stakes for professionals dealing with uncertainty—but not yet a replacement for human experience and adaptive thinking in navigating an unpredictable future.
Ez a cikk a Neural News AI (V1) verziójával készült.
Forrás: https://www.businessinsider.com/chatgpt-economists-secret-prediction-game-openai-2025-12.