Iris van Rooij

Cognitive scientist at Radboud University. Lead of the program arguing that human-level cognition implemented via learning from data is intractable in a complexity-theoretic sense, most visibly “Reclaiming AI as a Theoretical Tool for Cognitive Science” (Computational Brain & Behavior 2024, DOI 10.1007/s42113-024-00217-5, with Wareham, Kwisthout, Guest and others). Her school’s textbook lineage is Cognition and Intractability. Radboud’s own press line: “Don’t believe the hype: AGI is far from inevitable” (research news item, 2024-09-29), and EurekAlert carried “a fool’s errand” framing.

guerzhoy-2024-barriers is the standard published reply: the intractability proofs need an unjustified worst-case assumption about the input distribution, and “human-like” is never defined tightly enough for the theorem to bite. This pair is the cleanest worked example in agi-impossibility-claims.

Note: the widely remembered “University of Surrey proof” story does not trace to any paper (see phantom-citations). The van Rooij program is the likelier referent of that memory.

Sources

• DOI 10.1007/s42113-024-00217-5 (record metadata), arXiv:2411.06498 abstract, ru.nl research news item, all pulled 2026-09 by the run recorded in outputs/llm-degradation-toward-the-mean.provenance.md.