AI scientist conducts autonomous experiments, raising safety concerns

Summary: Researchers at the University of British Columbia have developed an "AI scientist" that autonomously designs and conducts machine learning experiments. Collaborating with Oxford and Sakana AI, the project explores open-ended learning, allowing AI to generate novel ideas. Recent findings include improvements in image generation and deep learning. While results are modest, the potential for more capable AI systems is significant, raising concerns about reliability and safety.

Full article