AI Systems Highlight Fundamental Limits in Scientific Research
Recent advancements in AI systems, specifically Robin and Co-Scientist, have been developed to assist in scientific research by automating parts of the scientific process. These systems, described in papers published in Nature, aim to accelerate scientific discovery by collaborating with human scientists. Robin, developed by Future House, focuses on drug repurposing, while Co-Scientist, from Google DeepMind, uses multi-agent systems to simulate scientific reasoning and hypothesis testing. Despite their potential, these AI systems reveal fundamental limitations, such as the inability to conduct physical experiments and the reliance on human input for defining scientific questions and validating hypotheses. 
