Turing-like tests for language and vision demonstrate a narrowing gap between humans and AI
As artificial intelligence (AI) becomes increasingly embedded in daily life, ascertaining whether an agent is human is critical. Here we systematically benchmark AI's ability to imitate humans in three language tasks (image captioning, word association and conversation) and three vision tasks (colour estimation, object detection and attention prediction), collecting data from n = 636 humans and 37 AI agents. We also conducted 72,191 Turing-like tests with 1,916 human judges and 10 AI judges. Current AIs are approaching the ability to convincingly impersonate humans and deceive human judges in both language and vision. Even simple AI judges outperformed humans in distinguishing AI from human responses. Imitation ability showed minimal correlation with conventional AI performance metrics, suggesting that passing as human is an important independent evaluation criterion. The large-scale Turing-like datasets and metrics introduced here offer valuable benchmarks for assessing human-likeness in AI and highlight the importance of rigorous, quantitative imitation tests for AI development.
Authors
- Shui’er Han (ORCID: https://orcid.org/0000-0002-2521-1416)
- Andrei Barbu (ORCID: https://orcid.org/0000-0001-7626-9266)
- Marcelo Armendáriz (ORCID: https://orcid.org/0000-0002-7905-1104)
- Ankur Sikarwar (ORCID: https://orcid.org/0009-0007-4946-5289)
- Caishun Chen (ORCID: https://orcid.org/0000-0003-1143-3138)
- Spandan Madan (ORCID: https://orcid.org/0000-0002-8957-0004)
- Mengmi Zhang (ORCID: https://orcid.org/0000-0002-2694-7097)
- Hanspeter Pfister (ORCID: https://orcid.org/0000-0002-3620-2582)
- Brandon Tang (ORCID: https://orcid.org/0000-0002-1969-7595)
- Michele Pizzochero (ORCID: https://orcid.org/0000-0003-3948-8202)
- Giorgia Dellaferrera (ORCID: https://orcid.org/0000-0003-1029-6637)
- Meghna Sadwani
- Yew-Soon Ong (ORCID: https://orcid.org/0000-0002-4480-169X)
- Noga Mudrik (ORCID: https://orcid.org/0000-0001-7086-9308)
- Stella Dellaferrera
- Gabriel Kreiman (ORCID: https://orcid.org/0000-0003-3505-8475)
- Haochen Yang (ORCID: https://orcid.org/0009-0004-9523-800X)
- Aman Singh (ORCID: https://orcid.org/0000-0003-4103-8761)
- Tanishq Kumar
- Elisa Pavarino
- Xiao Liu
- Prachi Agrawal
- Mranmay Shetty (ORCID: https://orcid.org/0000-0003-3114-9804)
Institutions
- Agency for Science, Technology and Research (SG)
- Boston Children's Hospital (US)
- Harvard University (US)
- Johns Hopkins University (US)
- Jawaharlal Nehru Medical College Hospital (IN)
- Nanyang Technological University (SG)
- University of Zurich (CH)
- Symbiosis International University (IN)
- ETH Zurich (CH)
- IBM Research - Zurich (CH)
- Institute of Neuroinformatics (CH)
- MIT Computer Science and Artificial Intelligence Laboratory (US)
- Institute of Advanced Intelligence and Computing (SG)
- University of Bath (GB)
- University of Turin (IT)
- Massachusetts Institute of Technology (US)
- Birla Institute of Technology and Science, Pilani (IN)
Publication Details
- Journal
- Nature Human Behaviour
- Published
- 2026-10-26
- DOI
- https://doi.org/10.1038/s41562-026-02590-6
- Primary Topic
- Artificial Intelligence Applications
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Science Foundation
- Harvard University
- Johns Hopkins University
- University of Bath
- Agency for Science, Technology and Research
- Nanyang Technological University
- Eidgenössische Technische Hochschule Zürich
- Fonds Wetenschappelijk Onderzoek
- Birla Institute of Technology and Science, Pilani
- Università degli Studi di Torino
- National Institutes of Health
- Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology