Core progress in AI has stalled in some fields.

Artificial intelligence (AI) just seems to get smarter and smarter. The surge reflects faster chips, more data, and better algorithms. But some of the improvement comes from tweaks rather than the core innovations their inventors claim—and some of the gains may not exist at all. Researchers have evaluated 81 pruning algorithms, programs that make neural networks, a type of AI, more efficient by trimming unneeded connections. All claimed superiority in slightly different ways. But when the researchers tried to evaluate them side by side, there was no clear evidence of performance improvements over a 10-year period. There are other signs of shaky progress across AI. A 2019 meta-analysis of information retrieval algorithms used in search engines concluded the "high-water mark … was actually set in 2009." Another study in 2019 reproduced seven recommendation systems, of the kind used by media streaming services. It found that six failed to outperform much simpler algorithms developed years before, at least when the earlier techniques were fine-tuned, revealing "phantom progress."