A British artificial intelligence company founded by a former Google DeepMind scientist earned eighth place in an international prediction contest. Mantic AI competed against human forecasters in the Metaculus Cup, which challenged participants to assess 60 potential future events during summer months. The competition featured predictions about political developments such as conflicts between Donald Trump and Elon Musk, plus Conservative Party leadership changes involving Kemi Badenoch.
The startup exceeded expectations by achieving over 80 percent of top human forecaster performance levels. Experts had previously estimated that leading artificial intelligence systems would reach only 40 percent of human capability. Mantic employs multiple machine learning models from OpenAI, Google, and DeepSeek, assigning each system tasks based on their particular strengths. Company leadership argues their approach demonstrates genuine reasoning abilities rather than simple data repetition.
This achievement signals potential progress toward Artificial General Intelligence, with Google DeepMind researchers projecting AGI arrival by 2030. Unlike current task-specific systems, AGI would possess human-like intelligence applicable across diverse domains and challenges.
The startup exceeded expectations by achieving over 80 percent of top human forecaster performance levels. Experts had previously estimated that leading artificial intelligence systems would reach only 40 percent of human capability. Mantic employs multiple machine learning models from OpenAI, Google, and DeepSeek, assigning each system tasks based on their particular strengths. Company leadership argues their approach demonstrates genuine reasoning abilities rather than simple data repetition.
This achievement signals potential progress toward Artificial General Intelligence, with Google DeepMind researchers projecting AGI arrival by 2030. Unlike current task-specific systems, AGI would possess human-like intelligence applicable across diverse domains and challenges.
