What happens when AI starts gambling on soccer? (www.businessinsider.com)

A recent initiative by the AI startup Obside has put several prominent AI models—including ChatGPT, Gemini, Claude, and Mistral—against each other in a unique betting challenge on World Cup soccer matches. An hour before each match, these models analyze team data, injuries, and market odds, wagering a virtual bankroll of $10,000 to predict match outcomes. As of the semifinals, the Mistral model was leading, followed closely by OpenAI's latest GPT version, while Anthropic's Claude Opus ranked last, possibly due to its more cautious approach to decision-making. This unconventional exercise is significant for the AI and machine learning community as it measures models' judgment under uncertainty—an area that traditional benchmarking often overlooks. By assessing how well these AI systems can leverage real-time data for profitable predictions in a dynamic environment, Obside's experiment provides insights into their practical applicability and decision-making capabilities in uncertain scenarios, advancing the discourse on AI effectiveness in real-world applications such as sports betting.