Sources
Loading...
Additional media
Loading...
DeepNewz, mobile.
People-sourced. AI-powered. Unbiased News.
The Center for AI Safety (CAIS) has developed an AI forecasting bot named FiveThirtyNine, built on GPT-4o, which claims to perform superhumanly well in predicting various events. The bot achieved an 87.7% accuracy rate on 177 predictions. This development, highlighted by DanHendrycks, is part of a broader trend where AI systems are showing superhuman capabilities in forecasting and persuasion, on par with groups of human forecasters. Demonstrations and previously published papers indicate that AI forecasters can automate most prediction markets and outperform human forecasters when used in conjunction with them. However, individual large language models (LLMs) alone are not as effective as crowds of AI or human forecasters.
There is a set of careful papers examining the ability of AI to help forecast future events. They show that human forecasters using AI outperform humans alone & that “crowds” of AIs are as good as crowds of human forecasters, but also that individual LLMs are not good forecasters https://t.co/spo0kWrsvu
It's worth noting that previously published papers this year have also shown that AIs can approach crowd-level ("superhuman") performance. From a previous work's abstract: "the system nears the crowd aggregate of competitive forecasters, and in some settings surpasses it." The… https://t.co/VpNkBD6LbS https://t.co/IRk49888aw
AI can predict the future at a superhuman level (on par with groups of human forecasters). Check out the prompt below, from @DanHendrycks and Center for AI Safety. It's my favourite type of prompt: also useful for our own human thinking. Working demo where you can ask questions… https://t.co/V4sa5hltrD
