AI in science

Scientific progress fuels economic growth and prosperity, yet concerns exist regarding AI's impact on research. Three data sources provide early insights: 15 million Gemini interactions, an inventory of 2,600 specialized AI models, and a survey of over 600 scientists. These data are organized into a taxonomy of scientific tasks, revealing four key findings.

First, AI is extensively adopted across scientific occupations, with nearly half of surveyed scientists using AI daily. Specialized AI models cover multiple disciplines and are highly cited. Second, LLMs and specialized models complement each other—LLMs handle general analysis, coding, and manuscript preparation, while specialized models offer domain-specific predictions, data generation, and classification.

Third, scientists report substantial productivity gains from AI use, saving about 7 hours per week, which they reinvest in research. Lastly, AI alters the scientific process, shifting bottlenecks downstream. Scientists face an increased backlog of untested hypotheses and demand more output verification.

These findings suggest AI can significantly boost scientific productivity. However, its ultimate impact depends on task interdependencies and addressing emerging bottlenecks, as detailed in a new paper by Mihai Codreanu and colleagues. The post AI in science appeared first on Marginal REVOLUTION.