New “AI scientists” are making progress – but are revealing their fundamental limitations

Many of the most exciting discoveries in science require highly specialized knowledge and making connections between distant facts. Scientists must work together deep Analysis with broad Argument strategies.

As with many information-rich tasks, researchers are turning to artificial intelligence (AI) systems to speed up their work. AI tools can potentially support important steps such as generating ideas, reviewing existing work, and analyzing data.

The latest systems use large language models (LLMs) to allow scientists to interact naturally and directly with the vast body of knowledge captured in words in the scientific literature.

But as two new systems described in papers just published in Nature show, language alone can only go so far in science.

What AI does to science

A number of organizations, such as Sakana AI, are attempting to automate the entire scientific process. To date, these efforts have largely focused on computer science, where “experiments” are primarily…