AI’s challenge of understanding the world
Abstract
In thinking about the challenge of getting artificial intelligence (AI) to understand our complex world, I recalled a Twitter post from a user of Tesla's self-driving system. The user tweeted that his car kept stopping abruptly at a particular location for no apparent reason. Then he noticed a billboard advertisement on the side of the road, featuring a sheriff holding up a stop sign. The car's vision system had interpreted this as an actual stop sign, and slammed on the brakes.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Some factual and reasoning errors made by LLM-based systems occur because they lack human-like understanding of language.
“Some of the factual and reasoning errors made by LLM-based systems occur because they lack human-like understanding of language [29, 135, 152, 153].”
From The ethics of using artificial intelligence in scientific research: new guidance needed for a new tool (2025) · cites this paper as [153]
The advancement of LLMs has underlined the significance of fine-tuning, hallucination reduction, and alignment.
“The advancement of LLMs has underlined the significance of fine-tuning [60], [61], [62], [63], hallucination reduction [64], [65], [66], [67], and alignment [68], [69], [70], [71], [72].”
From From Google Gemini to OpenAI Q* (Q-Star): A Survey on Reshaping the Generative Artificial Intelligence (AI) Research Landscape (2025) · cites this paper as [68]
Cited by
Sorted by citations
Artificial intelligence and illusions of understanding in scientific research
Messeri, L., & Crockett, M. J. (2024) · Nature · doi:10.1038/s41586-024-07146-0 · Cited by 707 · see citing works
Resnik, D. B., & Hosseini, M. (2025) · AI and Ethics · doi:10.1007/s43681-024-00493-8 · Cited by 228 · see citing works
“Some of the factual and reasoning errors made by LLM-based systems occur because they lack human-like understanding of language [29, 135, 152, 153].”
McIntosh, T. R., Susnjak, T., Liu, T., Watters, P., Xu, D., Liu, D., & Halgamuge, M. N. (2025) · Technologies · doi:10.3390/technologies13020051 · Cited by 102 · see citing works
Conceptualizing generative AI as style engines: Application archetypes and implications
Riemer, K., & Peter, S. (2024) · International Journal of Information Management · doi:10.1016/j.ijinfomgt.2024.102824 · Cited by 63 · see citing works
What undergraduate students need to know and actually know about generative AI
Rismanchian, S., Babar, E. T. R., & Doroudi, S. (2026) · Computers and Education: Artificial Intelligence · doi:10.1016/j.caeai.2026.100554 · Cited by 5 · see citing works
- Authors
- Mitchell, M.
- Affiliations
- Santa Fe Institute (US)
- Corresponding author
- Mitchell, Melanie
- ORCID
- Mitchell, Melanie
- Venue
- Science
- Published
- 2023-11-10
- Volume
- 382
- Issue
- 6671
- Pages
- eadm8175-eadm8175
- ISSN
- 0036-8075 (ISSN-L)
- 1095-9203
- OpenAlex
- W4388520388
- Cited by
- 35
- References
- 0
- Citations by year
- 2026: 8 · 2025: 17 · 2024: 10
- Type
- journal article
- Language
- en
- Open access
- closed
- Locations
- Science (publishedVersion)
- PubMed (publishedVersion)
- Topics
- Physical Sciences › Engineering › Automotive Engineering › Autonomous Vehicle Technology and Safety
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Artificial Intelligence › Artificial Intelligence Applications
- Keywords
- artificial intelligence, autonomous vehicles, computer vision, object recognition, image misinterpretation, Tesla
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
