{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,17]],"date-time":"2026-03-17T14:05:09Z","timestamp":1773756309861,"version":"3.50.1"},"reference-count":75,"publisher":"SAGE Publications","issue":"6","license":[{"start":{"date-parts":[[2024,5,24]],"date-time":"2024-05-24T00:00:00Z","timestamp":1716508800000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/journals.sagepub.com\/page\/policies\/text-and-data-mining-license"}],"funder":[{"DOI":"10.13039\/501100003407","name":"Ministero dell\u2019Istruzione, dell\u2019Universit\u00e0 e della Ricerca","doi-asserted-by":"crossref","award":["PNRR MUR project PE0000013-FAIR"],"award-info":[{"award-number":["PNRR MUR project PE0000013-FAIR"]}],"id":[{"id":"10.13039\/501100003407","id-type":"DOI","asserted-by":"crossref"}]}],"content-domain":{"domain":["journals.sagepub.com"],"crossmark-restriction":true},"short-container-title":["Adaptive Behavior"],"published-print":{"date-parts":[[2024,12]]},"abstract":"<jats:p> Large Language Models (LLMs) are capable of displaying a wide range of abilities that are not directly connected with the task for which they are trained: predicting the next words of human-written texts. In this article, I review recent research investigating the cognitive abilities developed by LLMs and their relation to human cognition. I discuss the nature of the indirect process that leads to the acquisition of these cognitive abilities, their relation to other indirect processes, and the implications for the acquisition of integrated abilities. Moreover, I propose the factors that enable the development of abilities that are related only very indirectly to the proximal objective of the training task. Finally, I discuss whether the full set of capabilities that LLMs could possibly develop is predictable. <\/jats:p>","DOI":"10.1177\/10597123241256754","type":"journal-article","created":{"date-parts":[[2024,5,24]],"date-time":"2024-05-24T22:43:27Z","timestamp":1716590607000},"page":"493-502","update-policy":"https:\/\/doi.org\/10.1177\/sage-journals-update-policy","source":"Crossref","is-referenced-by-count":7,"title":["On the Unexpected Abilities of Large Language Models"],"prefix":"10.1177","volume":"32","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-5035-442X","authenticated-orcid":false,"given":"Stefano","family":"Nolfi","sequence":"first","affiliation":[{"name":"Institute of Cognitive Sciences and Technologies, National Research Council(CNR-ISTC), Roma, Italy"}]}],"member":"179","published-online":{"date-parts":[[2024,5,24]]},"reference":[{"key":"bibr1-10597123241256754","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2021.conll-1.9"},{"key":"bibr2-10597123241256754","volume-title":"Solving rubik\u2019s cube with a robot hand","author":"Akkaya I.","year":"2019"},{"key":"bibr3-10597123241256754","volume-title":"Training a helpful and harmless assistant with reinforcement learning from human feedback","author":"Bai Y.","year":"2022"},{"key":"bibr4-10597123241256754","volume-title":"Constitutional ai: Harmlessness from ai feedback","author":"Bai Y.","year":"2022"},{"key":"bibr5-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1145\/3442188.3445922"},{"key":"bibr6-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.2218523120"},{"key":"bibr7-10597123241256754","volume-title":"Turning large language models into cognitive models","author":"Binz M.","year":"2023"},{"key":"bibr8-10597123241256754","volume-title":"On the opportunities and risks of foundation models","author":"Bommasani R.","year":"2021"},{"issue":"2","key":"bibr9-10597123241256754","first-page":"367","volume":"35","author":"Borghi A.","year":"2023","journal-title":"Sistemi Intelligenti"},{"key":"bibr10-10597123241256754","volume-title":"Eight things to know about large language models","author":"Bowman S. R.","year":"2023"},{"key":"bibr11-10597123241256754","first-page":"1877","volume":"33","author":"Brown T.","year":"2020","journal-title":"Advances in Neural Information Processing Systems"},{"key":"bibr12-10597123241256754","volume-title":"Sparks of artificial general intelligence: Early experiments with gpt-4","volume":"2303","author":"Bubeck S.","year":"2003"},{"key":"bibr13-10597123241256754","volume-title":"CoMPosT: Characterizing and evaluating caricature in LLM simulations","author":"Cheng M.","year":"2023"},{"key":"bibr14-10597123241256754","volume-title":"Palm: Scaling language modeling with pathways","author":"Chowdhery A.","year":"2022"},{"key":"bibr15-10597123241256754","volume":"30","author":"Christiano P. F.","year":"2017","journal-title":"Advances in Neural Information Processing Systems"},{"key":"bibr16-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1017\/S0140525X1500031X"},{"key":"bibr17-10597123241256754","volume-title":"Simulating opinion dynamics with networks of LLM-based agents","author":"Chuang Y. S.","year":"2023"},{"key":"bibr18-10597123241256754","volume-title":"Evaluating LLM agent group dynamics against human group dynamics: A case study on wisdom of partisan crowds","author":"Chuang Y. S.","year":"2023"},{"issue":"11","key":"bibr19-10597123241256754","first-page":"125","volume":"6","author":"Cisek P.","year":"1999","journal-title":"Journal of Consciousness Studies"},{"key":"bibr20-10597123241256754","volume-title":"Selection-inference: Exploiting large language models for interpretable logical reasoning","author":"Creswell A.","year":"2022"},{"key":"bibr21-10597123241256754","volume-title":"Why can gpt learn in-context? Language models secretly perform gradient descent as meta optimizers","author":"Dai D.","year":"2022"},{"key":"bibr22-10597123241256754","volume-title":"Build it break it fix it for dialogue safety: Robustness from adversarial human attack","author":"Dinan E.","year":"2019"},{"key":"bibr23-10597123241256754","volume-title":"Palm-e: An embodied multimodal language model","author":"Driess D.","year":"2023"},{"key":"bibr24-10597123241256754","volume-title":"Faith and fate: Limits of transformers on compositionality","author":"Dziri N.","year":"2023"},{"key":"bibr25-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1162\/tacl_a_00410"},{"key":"bibr26-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1016\/j.tics.2015.09.008"},{"key":"bibr27-10597123241256754","volume-title":"The capacity for moral self-correction in large language models","author":"Ganguli D.","year":"2023"},{"key":"bibr28-10597123241256754","volume-title":"Red teaming language models to reduce harms: Methods, scaling behaviors, and lessons learned","author":"Ganguli D.","year":"2022"},{"key":"bibr29-10597123241256754","volume-title":"Realtoxicityprompts: Evaluating neural toxic degeneration in language models","author":"Gehman S.","year":"2020"},{"key":"bibr30-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1080\/03057240.2023.2250570"},{"key":"bibr31-10597123241256754","first-page":"9118","volume-title":"International conference on machine learning","author":"Huang W.","year":"2022"},{"key":"bibr32-10597123241256754","first-page":"14","volume-title":"Proceedings of the annual meeting of the cognitive science society","volume":"44","author":"Jones C. R.","year":"2022"},{"key":"bibr33-10597123241256754","volume-title":"Scaling laws for neural language models","author":"Kaplan J.","year":"2020"},{"key":"bibr34-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1098\/rstb.2017.0052"},{"key":"bibr35-10597123241256754","volume-title":"Pretraining language models with human preferences","author":"Korbak T.","year":"2023"},{"key":"bibr36-10597123241256754","volume-title":"Theory of mind may have spontaneously emerged in large language models","author":"Kosinski M.","year":"2023"},{"key":"bibr37-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1145\/3638530.3654238"},{"key":"bibr38-10597123241256754","first-page":"27921","volume":"35","author":"Lee K. H.","year":"2022","journal-title":"Advances in Neural Information Processing Systems"},{"key":"bibr39-10597123241256754","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2021.acl-long.143"},{"key":"bibr40-10597123241256754","volume-title":"Dissociating language and thought in large language models: A cognitive perspective","author":"Mahowald K.","year":"2023"},{"key":"bibr41-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1016\/j.newideapsych.2009.07.001"},{"key":"bibr42-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.2215907120"},{"key":"bibr43-10597123241256754","volume-title":"The vector grounding problem","author":"Mollo D. C.","year":"2023"},{"key":"bibr44-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-642-01250-1"},{"key":"bibr45-10597123241256754","volume-title":"Shaking the foundations: Delusions in sequence models for interaction and control","author":"Ortega P. A.","year":"2021"},{"key":"bibr46-10597123241256754","first-page":"27730","volume":"35","author":"Ouyang L.","year":"2022","journal-title":"Advances in Neural Information Processing Systems"},{"key":"bibr47-10597123241256754","volume-title":"International conference on learning representations","author":"Patel R.","year":"2022"},{"key":"bibr48-10597123241256754","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D19-1250"},{"key":"bibr49-10597123241256754","volume-title":"Generating meaning: Active inference and the scope and limits of passive AI","author":"Pezzulo G.","year":"2023"},{"key":"bibr50-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1016\/0010-0277(88)90032-7"},{"key":"bibr51-10597123241256754","volume-title":"Scaling language models: Methods, analysis & insights from training gopher","author":"Rae J. W.","year":"2021"},{"key":"bibr52-10597123241256754","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2020.emnlp-main.437"},{"key":"bibr53-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1080\/13506285.2020.1866726"},{"key":"bibr54-10597123241256754","volume-title":"Are emergent abilities of Large Language Models a mirage?","author":"Schaeffer R.","year":"2023"},{"key":"bibr55-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-023-06647-8"},{"key":"bibr56-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1038\/nature24270"},{"key":"bibr57-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1007\/s11023-023-09622-4"},{"key":"bibr58-10597123241256754","volume-title":"Beyond the imitation game: Quantifying and extrapolating the capabilities of language models","author":"Srivastava A.","year":"2022"},{"key":"bibr59-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1162\/tacl_a_00342"},{"key":"bibr60-10597123241256754","volume-title":"Understanding the capabilities, limitations, and societal impact of large language models","author":"Tamkin A.","year":"2021"},{"key":"bibr61-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1162\/artl_a_00290"},{"key":"bibr62-10597123241256754","volume-title":"Lamda: Language models for dialog applications","author":"Thoppilan R.","year":"2022"},{"key":"bibr63-10597123241256754","volume-title":"Llama: Open and efficient foundation language models","author":"Touvron H.","year":"2023"},{"key":"bibr64-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1111\/cogs.13309"},{"key":"bibr65-10597123241256754","volume-title":"Large language models fail on trivial alterations to theory-of-mind tasks","author":"Ullman T.","year":"2023"},{"key":"bibr66-10597123241256754","volume":"30","author":"Vaswani A.","year":"2017","journal-title":"Advances in Neural Information Processing Systems"},{"key":"bibr67-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1201\/9781003205388-2"},{"key":"bibr68-10597123241256754","volume-title":"Call for papers--the BabyLM challenge: Sample-efficient pretraining on a developmentally plausible corpus","author":"Warstadt A.","year":"2023"},{"key":"bibr69-10597123241256754","doi-asserted-by":"publisher","DOI":"10.1162\/tacl_a_00321"},{"key":"bibr70-10597123241256754","volume-title":"Emergent abilities of Large Language Models","author":"Wei J.","year":"2022"},{"key":"bibr71-10597123241256754","volume-title":"Ethical and social risks of harm from language models","author":"Weidinger L.","year":"2021"},{"key":"bibr72-10597123241256754","volume-title":"Fundamental limitations of alignment in large language models","author":"Wolf Y.","year":"2023"},{"key":"bibr73-10597123241256754","volume-title":"From language modeling to instruction following: Understanding the behavior shift in LLMs after instruction tuning","author":"Wu X.","year":"2023"},{"key":"bibr74-10597123241256754","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2021.findings-emnlp.370"},{"key":"bibr75-10597123241256754","volume-title":"Instruction tuning for large language models: A survey","author":"Zhang S.","year":"2023"}],"container-title":["Adaptive Behavior"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/journals.sagepub.com\/doi\/pdf\/10.1177\/10597123241256754","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/journals.sagepub.com\/doi\/full-xml\/10.1177\/10597123241256754","content-type":"application\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/journals.sagepub.com\/doi\/pdf\/10.1177\/10597123241256754","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,3,1]],"date-time":"2025-03-01T04:49:28Z","timestamp":1740804568000},"score":1,"resource":{"primary":{"URL":"https:\/\/journals.sagepub.com\/doi\/10.1177\/10597123241256754"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,5,24]]},"references-count":75,"journal-issue":{"issue":"6","published-print":{"date-parts":[[2024,12]]}},"alternative-id":["10.1177\/10597123241256754"],"URL":"https:\/\/doi.org\/10.1177\/10597123241256754","relation":{},"ISSN":["1059-7123","1741-2633"],"issn-type":[{"value":"1059-7123","type":"print"},{"value":"1741-2633","type":"electronic"}],"subject":[],"published":{"date-parts":[[2024,5,24]]}}}