This paper studies the impact of artificial intelligence on innovation, using a randomized introduction of innovation technology to 1,018 scientists in the R&Dlab of a large US firm. AI-assisted researchers discover 44% more inventions, leading to a 39% increase in patent filings and a 17% increase in downstream product innovation. These compounds have novel chemical properties and lead to very powerful inventions. However, technology has strikingly different effects on the distribution of productivity: while the bottom third of scientists see little profit, the effect of top researchers is almost double. Investigating the mechanisms behind these results, I show that AI automates 57% of the “idea generation” tasks, and isolates researchers from the new task of evaluating model-generated applications. Top scientists use their domain knowledge to prioritize promising AI proposals, while others waste valuable resources testing false positives. Together, these findings demonstrate the potential of AI-augmented research and highlight the complementarity between algorithms and expertise in the innovation process. Research evidence suggests that these benefits come at a cost, however, as 82% of scientists report reduced job satisfaction due to reduced creativity and underutilization of skills.
That’s according to a new paper by Aidan Toner-Rodgers. By Kris Gulati.
Source link