The debate over understanding in AI’s large language models
Open accessThe DOI matches its Crossref record; 78 of the 78 references it cites were found in a registry or an index.
Abstract
We survey a current, heated debate in the artificial intelligence (AI) research community on whether large pretrained language models can be said to understand language—and the physical and social situations language encodes—in any humanlike sense. We describe arguments that have been made for and against such understanding and key questions for the broader sciences of intelligence that have arisen in light of these arguments. We contend that an extended science of intelligence can be developed that will provide insight into distinct modes of understanding, their strengths and limitations, and the challenge of integrating diverse forms of cognition.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
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?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“While a growing number of studies have shown that LLMs exhibit an inherent brittleness (13) and unreliability (14, 15) in knowledge tasks, and that they struggle with real-world understanding (11, 16) and human-like reasoning tasks (9, 10, 14), others have shown that LLMs instead excel in mimicking human language and communicative abilities (17)”
From The benefits and dangers of anthropomorphic conversational agents (2025) · cites this paper as [11]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“In their work, an argument is built up towards the inability of a machine to attain human level recognition, with the important distinction between form and meaning being drawn.”
From Deanthropomorphising NLP: Can a language model be conscious? (2024) · cites this paper as [84]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“This concern is often raised under the guise of the ELIZA effect: a phenomenon in which participants attribute human characteristics to even very simple systems [27,28].”
From People cannot distinguish GPT-4 from a human in a Turing test (2025) · cites this paper as [28]
The cited source says what the claim attributes to it.
In the citing paper
“Instead of resorting to human metaphors, it may be more productive to consider LLMs as a new type of 'understanding' distinct from human cognition (Mitchell & Krakauer, 2023).”
From Using ChatGPT and Other Large Language Model (LLM) Applications for Academic Paper Assignments (2023) · cites this paper as Mitchell & Krakauer, 2023
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018) · arXiv · doi:10.48550/arxiv.1810.04805 · Cited by 73,173 · cited 4×
“An LLM called BERT [16] obtained near-human performance on this benchmark [38].”
BackgroundInfluentialHighly accurate protein structure prediction with AlphaFold
Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., Tunyasuvunakool, K., Bates, R., Žídek, A., Potapenko, A., et al. (2021) · Nature · doi:10.1038/s41586-021-03819-2 · Cited by 43,764 · cited 1×
“…“understanding”, one that enables extraordinary, superhuman predictive ability, such as in the case of the AlphaZero and AlphaFold systems from DeepMind [40, 72], which respectively seem to bring an “alien” form of intuition to the domains of chess playing and protein-structure prediction [39, 68].”
BackgroundLanguage Models are Few-Shot Learners
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.14165 · Cited by 23,372 · cited 5×
“Those on the other side of this debate argue that large pretrained models such as GPT-3 or LaMDA—however fluent their linguistic output—cannot possess understanding because they have no experience or mental models of the world; their training in predicting words in vast collections of text has taught them the form of l…”
BackgroundMethodInfluentialKaugeranna, R., Kaugeranna, E., & Aion, (. S. 4. 6. (2023) · DROPS (Schloss Dagstuhl – Leibniz Center for Informatics) · doi:10.4230/lipics.giscience.2023.43 · Cited by 13,690
On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜
Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021) · Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency · doi:10.1145/3442188.3445922 · Cited by 6,988
Barsalou, L. W. (2008) · Annual Review of Psychology · doi:10.1146/annurev.psych.59.103006.093639 · Cited by 5,936 · cited 1×
“Researchers disagree on the extent to which concepts are domain-specific and innate versus more general-purpose and learned [14, 30, 31, 53, 75, 85], the degree to which concepts are grounded via embodied metaphors [28, 46, 61], are represented in the brain via dynamic, situation-based simulations [5], and the conditi…”
BackgroundChain-Of-Thought Prompting Elicits Reasoning in Large Language Models
Wei, J., Wang, X., Schuurmans, D., Bosma, M., Ichter, B., Xia, F., Chi, E., Le, Q. V., & Zhou, D. (2022) · Advances in Neural Information Processing Systems 35 · doi:10.52202/068431-1800 · Cited by 4,505
Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
Wei, J., Wang, X., Schuurmans, D., Bosma, M., Ichter, B., Xia, F., Chi, E., Le, Q., & Zhou, D. (2022) · arXiv · doi:10.48550/arxiv.2201.11903 · Cited by 4,263 · cited 1×
“LLMs such as OpenAI’s well-known GPT-3 (5) and more recent ChatGPT (6) and Google’s PaLM (7) can produce astonishingly humanlike text, conversation, and, in some cases, what seems like human reasoning abilities (8), even though the models were not explicitly trained to reason.”
BackgroundGLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding
Wang, A., Singh, A., Michael, J., Hill, F., Levy, O., & Bowman, S. (2018) · Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP · doi:10.18653/v1/w18-5446 · Cited by 4,222 · cited 2×
“For example, two standard benchmarks for assessing LLMs are the General Language Understanding Evaluation (GLUE) [46], and its successor (SuperGLUE) [47], which include large-scale datasets with tasks such as “textual entailment” (given two sentences, can the meaning of the second be inferred from the first?), “words…”
BackgroundTraining language models to follow instructions with human feedback
Ouyang, L., Wu, J., Jiang, X., Almeida, D., Wainwright, C. L., Mishkin, P., Zhang, C., Agarwal, S., Slama, K., Ray, A., et al. (2022) · arXiv · doi:10.48550/arxiv.2203.02155 · Cited by 4,208 · cited 1×
“Such models can be used to generate natural language, be fine-tuned for specific language tasks [58], or be further trained to better match “user intent” [65].”
BackgroundOn the Opportunities and Risks of Foundation Models
Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Arx, S., Bernstein, M. S., Bohg, J., Bosselut, A., Brunskill, E., et al. (2021) · arXiv · doi:10.48550/arxiv.2108.07258 · Cited by 2,676 · cited 1×
“Variously called Large Language Models (LLMs), Large Pre-1 Trained Models, or Foundation Models [11], these systems are deep neural networks with billions to trillions of parameters (weights) that are “pre-trained” on enormous natural-language corpora, including large swathes of the Web, online…”
BackgroundPaLM: Scaling Language Modeling with Pathways
Chowdhery, A., Narang, S., Devlin, J., Bosma, M., Mishra, G., Roberts, A., Barham, P., Chung, H. W., Sutton, C., Gehrmann, S., et al. (2022) · arXiv · doi:10.48550/arxiv.2204.02311 · Cited by 2,590 · cited 4×
“OpenAI’s GPT-3, with 175 billion parameters, performed surprisingly well on these tasks (5), and Google’s PaLM, with 540 billion parameters, performed even better (7), often equaling or surpassing humans on the same tasks.”
BackgroundMethodInfluentialBuilding machines that learn and think like people
Lake, B. M., Ullman, T. D., Tenenbaum, J. B., & Gershman, S. J. (2017) · Behavioral and Brain Sciences · doi:10.1017/s0140525x16001837 · Cited by 2,550
Building Machines That Learn and Think Like People
Lake, B. M., Ullman, T. D., Tenenbaum, J. B., & Gershman, S. J. (2016) · arXiv · doi:10.48550/arxiv.1604.00289 · Cited by 2,550
Shortcut learning in deep neural networks
Geirhos, R., Jacobsen, J.-H., Michaelis, C., Zemel, R., Brendel, W., Bethge, M., & Wichmann, F. A. (2020) · Nature Machine Intelligence · doi:10.1038/s42256-020-00257-z · Cited by 2,224 · cited 1×
“This is a straightforward example of “shortcut learning”—a commonly cited phenomenon in machine learning in which a learning system relies on spurious correlations in the data, rather than humanlike understanding, in order to perform well on a particular benchmark [19, 23, 29, 34].”
BackgroundRecent Advances in Natural Language Processing via Large Pre-Trained Language Models: A Survey
Min, B., Ross, H., Sulem, E., Veyseh, A. P. B., Nguyen, T. H., Sainz, O., Agirre, E., Heinz, I., & Roth, D. (2021) · arXiv · doi:10.48550/arxiv.2111.01243 · Cited by 1,387 · cited 1×
“Such models can be used to generate natural language, be fine-tuned for specific language tasks (3), or be further trained to better match “user intent” (4).”
BackgroundRecent Advances in Natural Language Processing via Large Pre-trained Language Models: A Survey
Min, B., Ross, H., Sulem, E., Veyseh, A. P. B., Nguyen, T. H., Sainz, O., Agirre, E., Heintz, I., & Roth, D. (2024) · ACM Computing Surveys · doi:10.1145/3605943 · Cited by 1,387 · cited 1×
“Such models can be used to generate natural language, be fine-tuned for specific language tasks (3), or be further trained to better match “user intent” (4).”
BackgroundKeil, F. C. (2006) · Annual Review of Psychology · doi:10.1146/annurev.psych.57.102904.190100 · Cited by 1,350
The Metaphorical Structure of the Human Conceptual System
Lakoff, G. P., & JOHNSON, M. (1980) · Cognitive Science · doi:10.1016/s0364-0213(80)80017-6 · Cited by 1,284 · cited 1×
“Researchers disagree on the extent to which concepts are domain-specific and innate versus more general-purpose and learned [14, 30, 31, 53, 75, 85], the degree to which concepts are grounded via embodied metaphors [28, 46, 61], are represented in the brain via dynamic, situation-based simulations [5], and the conditi…”
BackgroundHow to build a baby: II. Conceptual primitives.
Mandler, J. M. (1992) · Psychological Review · doi:10.1037//0033-295x.99.4.587 · Cited by 1,265 · cited 1×
“Researchers disagree on the extent to which concepts are domain-specific and innate versus more general-purpose and learned [14, 30, 31, 53, 75, 85], the degree to which concepts are grounded via embodied metaphors [28, 46, 61], are represented in the brain via dynamic, situation-based simulations [5], and the conditi…”
BackgroundEmergent Abilities of Large Language Models
Wei, J., Tay, Y., Bommasani, R., Raffel, C., Zoph, B., Borgeaud, S., Yogatama, D., Bosma, M., Zhou, D., Metzler, D., et al. (2022) · arXiv · doi:10.48550/arxiv.2206.07682 · Cited by 1,264
In-context Learning and Induction Heads
Olsson, C., Elhage, N., Nanda, N., Joseph, N., DasSarma, N., Henighan, T., Mann, B., Askell, A., Bai, Y., Chen, A., et al. (2022) · arXiv · doi:10.48550/arxiv.2209.11895 · Cited by 1,170 · cited 1×
“…to develop new kinds of benchmarks and probing methods that can yield insight into the mechanisms of diverse types of intelligence and understanding, including the novel forms of ”exotic, mind-like entities” [71] we have created, perhaps along the lines of some promising initial efforts [49, 64].”
BackgroundMastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm
Silver, D., Hubert, T., Schrittwieser, J., Antonoglou, I., Lai, M., Guez, A., Lanctot, M., Sifre, L., Kumaran, D., Graepel, T., et al. (2017) · arXiv · doi:10.48550/arxiv.1712.01815 · Cited by 1,102 · cited 2×
“If LLMs and related models succeed by exploiting statistical correlations at a heretofore unthinkable scale, perhaps this could be considered a novel form of “understanding”, one that enables extraordinary, superhuman predictive ability, such as in the case of the AlphaZero and AlphaFold systems from DeepMind (82, 83)…”
BackgroundSpelke, E. S., & Kinzler, K. D. (2007) · Developmental Science · doi:10.1111/j.1467-7687.2007.00569.x · Cited by 1,089
Climbing towards NLU: On Meaning, Form, and Understanding in the Age of Data
Bender, E. M., & Koller, A. (2020) · Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/2020.acl-main.463 · Cited by 1,082 · cited 1×
“…models such as GPT-3 or LaMDA—however fluent their linguistic output—cannot possess understanding because they have no experience or mental models of the world; their training in predicting words in vast collections of text has taught them the form of language but not the meaning [8, 9, 55].”
SuperGLUE: A Stickier Benchmark for General-Purpose Language Understanding Systems
Wang, A., Pruksachatkun, Y., Nangia, N., Singh, A., Michael, J., Hill, F., Levy, O., & Bowman, S. R. (2019) · arXiv · doi:10.48550/arxiv.1905.00537 · Cited by 1,051 · cited 1×
“For example, two standard benchmarks for assessing LLMs are the General Language Understanding Evaluation (GLUE) [46], and its successor (SuperGLUE) [47], which include large-scale datasets with tasks such as “textual entailment” (given two sentences, can the meaning of the second be inferred from the first?), “words…”
BackgroundUnmasking Clever Hans predictors and assessing what machines really learn
Lapuschkin, S., Wäldchen, S., Binder, A., Montavon, G., Samek, W., & Müller, K.-R. (2019) · Nature Communications · doi:10.1038/s41467-019-08987-4 · Cited by 1,047 · cited 1×
“This is a straightforward example of “shortcut learning”—a commonly cited phenomenon in machine learning in which a learning system relies on spurious correlations in the data, rather than humanlike understanding, in order to perform well on a particular benchmark [25, 35, 47, 56].”
BackgroundRight for the Wrong Reasons: Diagnosing Syntactic Heuristics in Natural Language Inference
McCoy, R. T., Pavlick, E., & Linzen, T. (2019) · Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/p19-1334 · Cited by 944 · cited 1×
“This is a straightforward example of “shortcut learning”—a commonly cited phenomenon in machine learning in which a learning system relies on spurious correlations in the data, rather than humanlike understanding, in order to perform well on a particular benchmark [19, 23, 29, 34].”
BackgroundCognitive Development: Foundational Theories of Core Domains
Wellman, H. M., & Gelman, S. A. (1992) · Annual Review of Psychology · doi:10.1146/annurev.ps.43.020192.002005 · Cited by 931 · cited 1×
“Researchers disagree on the extent to which concepts are domain-specific and innate versus more general-purpose and learned [14, 30, 31, 53, 75, 85], the degree to which concepts are grounded via embodied metaphors [28, 46, 61], are represented in the brain via dynamic, situation-based simulations [5], and the conditi…”
BackgroundAnnotation Artifacts in Natural Language Inference Data
Gururangan, S., Swayamdipta, S., Levy, O., Schwartz, R., Bowman, S., & Smith, N. A. (2018) · Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers) · doi:10.18653/v1/n18-2017 · Cited by 907 · cited 1×
“This is a straightforward example of “shortcut learning”—a commonly cited phenomenon in machine learning in which a learning system relies on spurious correlations in the data, rather than humanlike understanding, in order to perform well on a particular benchmark [19, 23, 29, 34].”
BackgroundGopnik, A., & Wellman, H. M. (1994) · Mapping the Mind · doi:10.1017/cbo9780511752902.011 · Cited by 811 · cited 1×
“Humans, unlike machines, seem to have a strong innate drive for this form of understanding, both in science and in everyday life [34].”
BackgroundLaMDA: Language Models for Dialog Applications
Thoppilan, R., De Freitas, D., Hall, J., Shazeer, N., Kulshreshtha, A., Cheng, H.-T., Jin, A., Bos, T., Baker, L., Du, Y., et al. (2022) · arXiv · doi:10.48550/arxiv.2201.08239 · Cited by 778
Murphy, G. L. (1996) · Cognition · doi:10.1016/0010-0277(96)00711-1 · Cited by 602 · cited 1×
“Researchers disagree on the extent to which concepts are domain-specific and innate versus more general-purpose and learned [14, 30, 31, 53, 75, 85], the degree to which concepts are grounded via embodied metaphors [28, 46, 61], are represented in the brain via dynamic, situation-based simulations [5], and the conditi…”
BackgroundUnmasking Clever Hans predictors and assessing what machines really learn
Lapuschkin, S., Wäldchen, S., Binder, A., Montavon, G., Samek, W., & Müller, K. (2019) · RePEc: Research Papers in Economics · Cited by 580
Using cognitive psychology to understand GPT-3
Binz, M., & Schulz, E. (2023) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2218523120 · Cited by 559
Learning from Others: Children's Construction of Concepts
Gelman, S. A. (2009) · Annual Review of Psychology · doi:10.1146/annurev.psych.59.103006.093659 · Cited by 412 · cited 1×
“Researchers disagree on the extent to which concepts are domain-specific and innate versus more general-purpose and learned [14, 30, 31, 53, 75, 85], the degree to which concepts are grounded via embodied metaphors [28, 46, 61], are represented in the brain via dynamic, situation-based simulations [5], and the conditi…”
BackgroundProbing Neural Network Comprehension of Natural Language Arguments
Niven, T., & Kao, H.-Y. (2019) · Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/p19-1459 · Cited by 395 · cited 2×
“An LLM called BERT [16] obtained near-human performance on this benchmark [38].”
BackgroundNo understanding without explanation
Strevens, M. (2013) · Studies in History and Philosophy of Science Part A · doi:10.1016/j.shpsa.2012.12.005 · Cited by 355 · cited 1×
“It has been argued that understanding of this kind may enable abilities not possible for purely statistical models: learning new concepts from limited examples, being able to generalize that knowledge to novel instances of those concepts, being able to flexibly transfer one’s knowledge to new situations, robust reason…”
BackgroundDissociating language and thought in large language models
Mahowald, K., Ivanova, A. A., Blank, I. A., Kanwisher, N., Tenenbaum, J. B., & Fedorenko, E. (2024) · Trends in Cognitive Sciences · doi:10.1016/j.tics.2024.01.011 · Cited by 345
Theoretical Impediments to Machine Learning With Seven Sparks from the Causal Revolution
Pearl, J. (2018) · Proceedings of the Eleventh ACM International Conference on Web Search and Data Mining · doi:10.1145/3159652.3176182 · Cited by 338
Early word-learning entails reference, not merely associations
Waxman, S. R., & Gelman, S. A. (2009) · Trends in Cognitive Sciences · doi:10.1016/j.tics.2009.03.006 · Cited by 261 · cited 1×
“…85], the degree to which concepts are grounded via embodied metaphors [28, 46, 61], are represented in the brain via dynamic, situation-based simulations [5], and the conditions under which concepts are underpinned by language [20, 23, 51], by social learning [4, 81, 26] and by culture [7, 60, 63].”
BackgroundLearning a theory of causality.
Goodman, N. D., Ullman, T. D., & Tenenbaum, J. B. (2011) · Psychological Review · doi:10.1037/a0021336 · Cited by 247 · cited 1×
“Researchers disagree on the extent to which concepts are domain-specific and innate versus more general-purpose and learned [14, 30, 31, 53, 75, 85], the degree to which concepts are grounded via embodied metaphors [28, 46, 61], are represented in the brain via dynamic, situation-based simulations [5], and the conditi…”
BackgroundThe Role of Language in Theory of Mind Development
de Villiers, J. G., & de Villiers, P. A. (2014) · Topics in Language Disorders · doi:10.1097/tld.0000000000000037 · Cited by 240 · cited 1×
“Researchers disagree on the extent to which concepts are domain-specific and innate versus more general-purpose and learned [14, 30, 31, 53, 75, 85], the degree to which concepts are grounded via embodied metaphors [28, 46, 61], are represented in the brain via dynamic, situation-based simulations [5], and the conditi…”
BackgroundMorris, M. W., Menon, T., & Ames, D. R. (2001) · Personality and Social Psychology Review · doi:10.1207/s15327957pspr0502_7 · Cited by 215 · cited 1×
“Researchers disagree on the extent to which concepts are domain-specific and innate versus more general-purpose and learned [14, 30, 31, 53, 75, 85], the degree to which concepts are grounded via embodied metaphors [28, 46, 61], are represented in the brain via dynamic, situation-based simulations [5], and the conditi…”
BackgroundSloman, S. A., & Lagnado, D. (2015) · Annual Review of Psychology · doi:10.1146/annurev-psych-010814-015135 · Cited by 210 · cited 1×
“These models enable people to abstract their knowledge and experiences in order to make robust predictions, generalizations, and analogies, to reason compositionally and counterfactually, to actively intervene on the world in order to test hypotheses, and to explain one’s understanding to others [29, 32, 38, 41, 45, 7…”
BackgroundNorenzayan, A., & Nisbett, R. E. (2000) · Current Directions in Psychological Science · doi:10.1111/1467-8721.00077 · Cited by 203 · cited 1×
“Researchers disagree on the extent to which concepts are domain-specific and innate versus more general-purpose and learned [14, 30, 31, 53, 75, 85], the degree to which concepts are grounded via embodied metaphors [28, 46, 61], are represented in the brain via dynamic, situation-based simulations [5], and the conditi…”
BackgroundThe Social Nature of Words and Word Learning
Akhtar, N., & Tomasello, M. (2000) · Becoming a Word Learner · doi:10.1093/acprof:oso/9780195130324.003.005 · Cited by 188 · cited 1×
“Researchers disagree on the extent to which concepts are domain-specific and innate versus more general-purpose and learned [14, 30, 31, 53, 75, 85], the degree to which concepts are grounded via embodied metaphors [28, 46, 61], are represented in the brain via dynamic, situation-based simulations [5], and the conditi…”
BackgroundWhat Is Understanding? An Overview of Recent Debates in Epistemology and Philosophy of Science
Baumberger, C., Beisbart, C., & Brun, G. (2017) · BORIS (University Library Bern) · Cited by 178
How Language Programs the Mind
Lupyan, G., & Bergen, B. (2016) · Topics in Cognitive Science · doi:10.1111/tops.12155 · Cited by 170 · cited 1×
“Researchers disagree on the extent to which concepts are domain-specific and innate versus more general-purpose and learned [14, 30, 31, 53, 75, 85], the degree to which concepts are grounded via embodied metaphors [28, 46, 61], are represented in the brain via dynamic, situation-based simulations [5], and the conditi…”
BackgroundLarge Language Models and the Reverse Turing Test
Sejnowski, T. J. (2023) · Neural Computation · doi:10.1162/neco_a_01563 · Cited by 169
Human Language Understanding & Reasoning
Manning, C. D. (2022) · Daedalus · doi:10.1162/daed_a_01905 · Cited by 162
The impact of AlphaFold2 one year on
Jones, D. T., & Thornton, J. M. (2022) · Nature Methods · doi:10.1038/s41592-021-01365-3 · Cited by 159 · cited 1×
“If LLMs and related models succeed by exploiting statistical correlations at a heretofore unthinkable scale, perhaps this could be considered a novel form of “understanding”, one that enables extraordinary, superhuman predictive ability, such as in the case of the AlphaZero and AlphaFold systems from DeepMind (82, 83)…”
BackgroundOn the origin of causal understanding
Carey, S. (1996) · Causal Cognition · doi:10.1093/acprof:oso/9780198524021.003.0010 · Cited by 153 · cited 1×
“Researchers disagree on the extent to which concepts are domain-specific and innate versus more general-purpose and learned [14, 30, 31, 53, 75, 85], the degree to which concepts are grounded via embodied metaphors [28, 46, 61], are represented in the brain via dynamic, situation-based simulations [5], and the conditi…”
BackgroundHow Can We Accelerate Progress Towards Human-like Linguistic Generalization?
Linzen, T. (2020) · Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/2020.acl-main.465 · Cited by 148
Word meaning in minds and machines.
Lake, B. M., & Murphy, G. L. (2023) · Psychological Review · doi:10.1037/rev0000297 · Cited by 131
The Argument Reasoning Comprehension Task: Identification and Reconstruction of Implicit Warrants
Habernal, I., Wachsmuth, H., Gurevych, I., & Stein, B. (2018) · Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers) · doi:10.18653/v1/n18-1175 · Cited by 121 · cited 1×
“As an example, consider one such benchmark, the Argument Reasoning Comprehension Task [36].”
BackgroundGoldwater, M. B., & Gentner, D. (2015) · Cognition · doi:10.1016/j.cognition.2014.12.001 · Cited by 112 · cited 1×
“These models enable people to abstract their knowledge and experiences in order to make robust predictions, generalizations, and analogies, to reason compositionally and counterfactually, to actively intervene on the world in order to test hypotheses, and to explain one’s understanding to others [29, 32, 38, 41, 45, 7…”
BackgroundDo Large Language Models Know What Humans Know?
Trott, S., Jones, C., Chang, T., Michaelov, J., & Bergen, B. (2023) · Cognitive Science · doi:10.1111/cogs.13309 · Cited by 99
Discussion Note: Making Sense of Understanding
de Regt, H. W. (2004) · Philosophy of Science · doi:10.1086/381415 · Cited by 91 · cited 1×
“It has been argued that understanding of this kind may enable abilities not possible for purely statistical models: learning new concepts from limited examples, being able to generalize that knowledge to novel instances of those concepts, being able to flexibly transfer one’s knowledge to new situations, robust reason…”
BackgroundImplicit Representations of Meaning in Neural Language Models
Li, B. Z., Nye, M., & Andreas, J. (2021) · Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers) · doi:10.18653/v1/2021.acl-long.143 · Cited by 84 · cited 1×
“to develop new kinds of benchmarks and probing methods that can yield insight into the mechanisms of diverse types of intelligence and understanding, including the novel forms of “exotic, mind-like entities” (79) we have created, perhaps along the lines of some promising initial efforts (80, 81).”
BackgroundCompetency Problems: On Finding and Removing Artifacts in Language Data
Gardner, M., Merrill, W., Dodge, J., Peters, M., Ross, A., Singh, S., & Smith, N. A. (2021) · Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/2021.emnlp-main.135 · Cited by 81 · cited 1×
“Thus benchmarks or assessments that would be appropriate for measuring human understanding might not be appropriate for assessing such machines [15, 24, 50].”
BackgroundArtificial Intelligence Hits the Barrier of Meaning
Mitchell, M. (2019) · Information · doi:10.3390/info10020051 · Cited by 80 · cited 1×
“Indeed, the oft-noted brittleness of these AI systems—their unpredictable errors and lack of robust generalization abilities—are key indicators of their lack of understanding [59].”
BackgroundSelf-citationDissociating language and thought in large language models
Mahowald, K., Ivanova, A. A., Blank, I. A., Kanwisher, N., Tenenbaum, J. B., & Fedorenko, E. (2023) · arXiv · doi:10.48550/arxiv.2301.06627 · Cited by 80
More than a scaffold: Language is a neuroenhancement
Dove, G. (2020) · Cognitive Neuropsychology · doi:10.1080/02643294.2019.1637338 · Cited by 77 · cited 1×
“Researchers disagree on the extent to which concepts are domain-specific and innate versus more general-purpose and learned [14, 30, 31, 53, 75, 85], the degree to which concepts are grounded via embodied metaphors [28, 46, 61], are represented in the brain via dynamic, situation-based simulations [5], and the conditi…”
BackgroundLanguage models, like humans, show content effects on reasoning tasks
Lampinen, A. K., Dasgupta, I., Chan, S. C. Y., Sheahan, H. R., Creswell, A., Kumaran, D., McClelland, J. L., & Hill, F. (2024) · PNAS Nexus · doi:10.1093/pnasnexus/pgae233 · Cited by 69
Smolensky, P., McCoy, R. T., Fernandez, R., Goldrick, M., & Gao, J. (2022) · AI Magazine · doi:10.1002/aaai.12065 · Cited by 68
Talking About Large Language Models
Shanahan, M. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2212.03551 · Cited by 67
Meaning without reference in large language models
Piantadosi, S. T., & Hill, F. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2208.02957 · Cited by 65 · cited 1×
“Another group argues that LLMs “likely capture important aspects of meaning, and moreover work in a way that approximates a compelling account of human cognition in which meaning arises from conceptual role” [67].”
BackgroundUsing cognitive psychology to understand GPT-3Preprint
Binz, M., & Schulz, E. (2022) · doi:10.31234/osf.io/6dfgk · Cited by 62
Language models show human-like content effects on reasoning tasks
Dasgupta, I., Lampinen, A. K., Chan, S. C. Y., Sheahan, H. R., Creswell, A., Kumaran, D., McClelland, J. L., & Hill, F. (2022) · arXiv · doi:10.48550/arxiv.2207.07051 · Cited by 53
What Do NLP Researchers Believe? Results of the NLP Community MetasurveyPreprint
Michael, J., Holtzman, A., Parrish, A., Mueller, A., Wang, A., Chen, A., Madaan, D., Nangia, N., Pang, R. Y., Phang, J., et al. (2022) · Annual Meeting of the Association for Computational Linguistics · doi:10.48550/arxiv.2208.12852 · Cited by 50 · cited 1×
“%) disagreed [57].”
Bender, A., Beller, S., & Medin, D. L. (2017) · Oxford University Press eBooks · doi:10.1093/oxfordhb/9780199399550.013.34 · Cited by 28 · cited 1×
“Researchers disagree on the extent to which concepts are domain-specific and innate versus more general-purpose and learned [14, 30, 31, 53, 75, 85], the degree to which concepts are grounded via embodied metaphors [28, 46, 61], are represented in the brain via dynamic, situation-based simulations [5], and the conditi…”
BackgroundKnowledge, Understanding, and Reasons for Belief
Kvanvig, J. L. (2018) · Oxford University Press eBooks · doi:10.1093/oxfordhb/9780199657889.013.30 · Cited by 23 · cited 1×
“Indeed, much of the long history of research in cognitive science has been a quest to understand the nature of concepts and how understanding arises from coherent, hierarchical sets of relations among concepts that include underlying causal knowledge (39, 40).”
BackgroundFrom CAPTCHA to Commonsense: How Brain Can Teach Us About Artificial Intelligence
George, D., Lázaro-Gredilla, M., & Guntupalli, J. S. (2020) · Frontiers in Computational Neuroscience · doi:10.3389/fncom.2020.554097 · Cited by 17 · cited 1×
“It has been argued that understanding of this kind may enable abilities not possible for purely statistical models [19, 27, 44, 66, 76].”
BackgroundPredicting Human Psychometric Properties Using Computational Language Models
Laverghetta, A., Nighojkar, A., Mirzakhalov, J., & Licato, J. (2022) · Springer proceedings in mathematics & statistics · doi:10.1007/978-3-031-04572-1_12 · Cited by 16
Alison Gopnik, C. (2011) · Behavioral and Brain Sciences · doi:10.1017/s0140525x10002438 · Cited by 13
Machine Reading, Fast and Slow: When Do Models "Understand" Language?Preprint
Choudhury, S. R., Rogers, A., & Augenstein, I. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2209.07430 · Cited by 10
Causal Models and Cognitive Development
Gopnik, A. (2022) · ACM eBooks · doi:10.1145/3501714.3501746 · Cited by 4 · cited 1×
“These models enable people to abstract their knowledge and experiences in order to make robust predictions, generalizations, and analogies, to reason compositionally and counterfactually, to actively intervene on the world in order to test hypotheses, and to explain one’s understanding to others [29, 32, 38, 41, 45, 7…”
Background
Research works printed but not linked: 3
- 10.1037/0033-295X.99.4.587 · doi:10.1037/0033-295x.99.4.587 · Paper page
- R. W. Gibbs Metaphor Wars (Cambridge University Press 2017).
- 10.1207/s15516709cog0402_4 · doi:10.1207/s15516709cog0402_4 · Paper page
Cited by
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
“Although the question of what AI models can ‘understand’ remains controversial 14 , the AI visions (alongside the language of ‘deep learning’, ‘deep neural networks’ and marketing hype 95 ) portray AI as having a depth of understanding beyond what can be grasped by limited human minds.”
Can Generative AI improve social science?
Bail, C. A. (2024) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2314021121 · Cited by 369
Can Generative AI Improve Social Science?
Bail, C. A. (2023) · doi:10.31235/osf.io/rwtzs · Cited by 369
Dissociating language and thought in large language models
Mahowald, K., Ivanova, A. A., Blank, I. A., Kanwisher, N., Tenenbaum, J. B., & Fedorenko, E. (2024) · Trends in Cognitive Sciences · doi:10.1016/j.tics.2024.01.011 · Cited by 345
Susarla, A., Gopal, R., Thatcher, J. B., & Sarker, S. (2023) · Information Systems Research · doi:10.1287/isre.2023.ed.v34.n2 · Cited by 285
Metadata
- Authors
- Mitchell, M., & Krakauer, D. C.
- Affiliations
- Santa Fe Institute (US)
- Corresponding author
- Krakauer, David C.
- Venue
- Proceedings of the National Academy of Sciences
- Publisher
- National Academy of Sciences
- Published
- 2023-03-21
- Volume
- 120
- Issue
- 13
- Pages
- e2215907120-e2215907120
- ISSN
- 0027-8424 (ISSN-L)
- 1091-6490
- PMC
- PMC10068812
- OpenAlex
- W4328049044
- Cited by
- 340
- References
- 56
- Citations by year
- 2026: 107 · 2025: 105 · 2024: 84 · 2023: 42
- FWCI
- 31.82
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- Locations
- Proceedings of the National Academy of Sciences (publishedVersion, cc-by-nc-nd)
- PubMed Central (publishedVersion, cc-by-nc-nd)
- PubMed (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- National Science Foundation (2020103)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Social Sciences › Arts and Humanities › Philosophy › Philosophy and Theoretical Science
- Social Sciences › Arts and Humanities › History and Philosophy of Science › Cybernetics and Technology in Society
- Keywords
- large language models, language comprehension, artificial intelligence, cognition, pre-trained language models
- MeSH terms
- Artificial Intelligence, Cognition, Dissent and Disputes
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
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.