That’s the pragmatic advice from a top human forecaster after a British AI system secured a stunning eighth-place finish in a global prediction contest, surpassing many of its human counterparts. The quote, from third-place finisher Lubos Saloky, perfectly captures the shifting sentiment in a field grappling with the rapid ascent of artificial intelligence.
The competition, the Metaculus Cup, challenged participants to forecast the outcomes of 60 unpredictable events over the summer. The success of ManticAI, the AI system in question, has served as a powerful demonstration that machines are becoming adept at the kind of reasoning and analysis previously thought to be the exclusive domain of human experts.
ManticAI’s system is a testament to modern AI architecture. Co-founded by a former Google DeepMind scientist, it uses a collaborative team of AI agents, drawing on models from OpenAI, Google, and others. This allows it to tackle problems with a persistence and breadth that is superhuman, constantly updating its predictions as new information becomes available.
The AI’s ability to reason and even show a degree of “originality” by diverging from human consensus has impressed observers. Its performance is a milestone, suggesting that AI can be more than just a tool for data processing; it can be a partner in analysis, capable of challenging assumptions and preventing the emergence of echo chambers.
Saloky’s “merge with them” philosophy is increasingly seen as the most logical path forward. Experts like Warren Hatch of Good Judgment echo this, advocating for a “human and AI” approach. In this model, humans would leverage AI’s immense analytical power while providing the crucial oversight, judgment, and intuition that machines still lack, creating a partnership more powerful than either could be alone.