← All research spotlights

Research spotlight · Unit 8

Melanie Mitchell

Santa Fe Institute

Research field
Complex systems, conceptual abstraction and analogy, the evaluation of artificial intelligence

Melanie Mitchell studies how machines and people form concepts and make analogies, and she has written widely read assessments of how far artificial intelligence has come. In a 2021 paper she asked why the field has repeatedly predicted human-level intelligence within a decade and repeatedly been wrong, and identified four assumptions behind the predictions: that narrow abilities lie on a continuum with general intelligence, that what is easy for people is easy for machines, that programs and benchmarks named after human abilities measure those abilities, and that intelligence can be separated from a body. The unit presents the paper after its account of the two AI winters, as an explanation of why the springs and winters recur, and returns to her argument in the open questions.

In the masterclass

Section 2, after the account of the two AI winters, presents her four fallacies as an explanation of why the field keeps overestimating its progress.

The research is presented in Unit 8: The Big Questions of Artificial Intelligence, in a research spotlight box that explains the result at the level of the unit and links it to what you have just learned.

The paper

Mitchell, M. (2021). Why AI is Harder Than We Think. arXiv preprint, arXiv:2104.12871. https://doi.org/10.48550/arXiv.2104.12871

The scientists named here are not contributors to the AI Masterclass, are not affiliated with it, and have not endorsed it. The program presents and explains their published research, with references to the original papers.