The research team at UC Berkeley developed a language model system that approaches human-level prediction capabilities. The system automates key prediction processes using large-scale data and the rapid processing abilities of language models. The study shows an average Brier score of 0.179, indicating potential for increased prediction accuracy and efficiency. This research provides strong evidence for integrating language models into the prediction field.

["Researchers at the University of California, Berkeley have developed robotic vision tracking technology,", "Through privileged learning, robots can predict the target's movements and intentions,", "This breakthrough takes robotic technology into the real world."]

The UC Berkeley team successfully leveraged ChatGPT to produce a large-scale chemistry research dataset. Through prompt engineering, ChatGPT extracted data from papers with an accuracy rate of 95%. The study shows that ChatGPT can significantly accelerate the research process.