The impact of generative artificial intelligence on innovation: Evidence from software products
After ChatGPT's release, new Firefox add-ons rose 34% but improvements to existing ones fell 20%, and new products were not more novel.
Summary
Studying the Firefox add-on ecosystem around the release of ChatGPT, the authors find that generative AI led to about 34% more new add-ons, from both existing and first-time developers, while developers cut improvements to existing add-ons by about 20%. The new add-ons drew on a broader range of knowledge but were not substantially more novel, which suggests generative AI helps recombine existing knowledge more than it produces breakthroughs. For platforms, the trade-off is more entry and variety against a risk to the quality and upkeep of what is already there, so platform managers may need strategies to maintain quality. The setting is one open software marketplace, so generalisation to other platforms is untested.
Editor's opinion
Generative AI lowers the cost of launching new products, and developers respond by shipping more new things and maintaining old ones less, without the new ones being more original. For platform owners, the risk is a flood of new entrants and decay in the upkeep of the existing catalogue, so curation and incentives for maintenance matter more than encouraging more entry.
Key concepts
generative AI · innovation · software ecosystem · product quality · developers