Abstract
Given the impressive performances of LLM-derived tools across a range of tasks considered all but impossible for computers until recently, the capabilities of LLMs seem limitless. However, there are some fundamental limitations to what they can or cannot do inherent to the current architecture of LLMs. I will attempt to review the most notable of them to give the reader an understanding of what architectural modifications will need to take place before a given problem is solved. Specifically, I discuss counterfactual generation, private information leakage, reasoning, limited attention span, dependence on the training dataset, bias, and non-normative language.
Chapter PDF
References
Ross Taylor et al. Galactica: A large language model for science. CoRR, abs/2211.09085, 2022.
Benj Edwards. New meta ai demo writes racist and inaccurate scientific literature, gets pulled. Ars Technica, 2022.
Amelia Glaese et al. Improving alignment of dialogue agents via targeted human judgements. CoRR, abs/2209.14375, 2022.
Kurt Shuster et al. Language models that seek for knowledge: Modular search & generation for dialogue and prompt completion. CoRR, abs/2203.13224, 2022.
OpenAI. Gpt-4 technical report. CoRR, abs/2303.08774, 2023.
Nicholas Carlini et al. Extracting training data from large language models. In Michael Bailey and Rachel Greenstadt, editors, 30th USENIX Security Symposium, USENIX Security 2021, August 11–13, 2021, pages 2633–2650. USENIX Association, 2021.
Ethan Perez et al. Red teaming language models with language models. CoRR, abs/2202.03286, 2022.
Tom B. Brown et al. Language models are few-shot learners. In Hugo Larochelle, Marc’Aurelio Ranzato, Raia Hadsell, Maria-Florina Balcan, and Hsuan-Tien Lin, editors, Advances in Neural Information Processing Systems 33: Annual Conference on Neural Information Processing Systems 2020, NeurIPS 2020, December 6–12, 2020, virtual, 2020.
Jason Wei et al. Chain of thought prompting elicits reasoning in large language models. CoRR, abs/2201.11903, 2022.
Takeshi Kojima et al. Large language models are zero-shot reasoners. In NeurIPS, 2022.
Hyung Won Chung et al. Scaling instruction-finetuned language models. CoRR, abs/2210.11416, 2022.
Or Honovich, Thomas Scialom, Omer Levy, and Timo Schick. Unnatural instructions: Tuning language models with (almost) no human labor. CoRR, abs/2212.09689, 2022.
Jason Wei et al. Finetuned language models are zero-shot learners. In The Tenth International Conference on Learning Representations, ICLR 2022, Virtual Event, April 25–29, 2022. OpenReview.net, 2022.
Simon Frieder et al. Mathematical capabilities of chatgpt. CoRR, abs/2301.13867, 2023.
Romal Thoppilan et al. Lamda: Language models for dialog applications. CoRR, abs/2201.08239, 2022.
Kurt Shuster et al. Blenderbot 3: a deployed conversational agent that continually learns to responsibly engage. CoRR, abs/2208.03188, 2022.
Stephen Wolfram. Chatgpt gets its “wolfram superpowers”!, 2023.
Shouyuan Chen, Sherman Wong, Liangjian Chen, and Yuandong Tian. Extending context window of large language models via positional interpolation. CoRR, abs/2306.15595, 2023.
Jianlin Su et al. Roformer: Enhanced transformer with rotary position embedding. CoRR, abs/2104.09864, 2021.
Ofir Press, Noah A. Smith, and Mike Lewis. Train short, test long: Attention with linear biases enables input length extrapolation. In The Tenth International Conference on Learning Representations, ICLR 2022, Virtual Event, April 25–29, 2022. OpenReview.net, 2022.
Nelson F. Liu et al. Lost in the middle: How language models use long contexts. CoRR, abs/2307.03172, 2023.
Xiangyu Peng, Siyan Li, Spencer Frazier, and Mark O. Riedl. Reducing non-normative text generation from language models. In Brian Davis, Yvette Graham, John D. Kelleher, and Yaji Sripada, editors, Proceedings of the 13th International Conference on Natural Language Generation, INLG 2020, Dublin, Ireland, December 15–18, 2020, pages 374–383. Association for Computational Linguistics, 2020.
Ben Krause et al. Gedi: Generative discriminator guided sequence generation. In Marie-Francine Moens, Xuanjing Huang, Lucia Specia, and Scott Wen-tau Yih, editors, Findings of the Association for Computational Linguistics: EMNLP 2021, Virtual Event / Punta Cana, Dominican Republic, 16–20 November, 2021, pages 4929–4952. Association for Computational Linguistics, 2021.
Long Ouyang et al. Training language models to follow instructions with human feedback. CoRR, abs/2203.02155, 2022.
Zvi Mowshowitz. Jailbreaking chatgpt on release day, 2022. Accessed on Feb 07, 2023.
James Vincent. Microsoft’s bing is an emotionally manipulative liar, and people love it. The Verge, February 2023.
Yixin Wan et al. “kelly is a warm person, joseph is a role model”: Gender biases in llm-generated reference letters. CoRR, abs/2310.09219, 2023.
Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. On the dangers of stochastic parrots: Can language models be too big? In Madeleine Clare Elish, William Isaac, and Richard S. Zemel, editors, FAccT ’21: 2021 ACM Conference on Fairness, Accountability, and Transparency, Virtual Event / Toronto, Canada, March 3–10, 2021, pages 610–623. ACM, 2021.
Jieyu Zhao a et al. Gender bias in contextualized word embeddings. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2019, Minneapolis, MN, USA, June 2–7, 2019, Volume 1 (Long and Short Papers), pages 629–634. Association for Computational Linguistics, 2019.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Copyright information
© 2024 The Author(s)
About this chapter
Cite this chapter
Kucharavy, A. (2024). Fundamental Limitations of Generative LLMs. In: Kucharavy, A., Plancherel, O., Mulder, V., Mermoud, A., Lenders, V. (eds) Large Language Models in Cybersecurity. Springer, Cham. https://doi.org/10.1007/978-3-031-54827-7_5
Download citation
DOI: https://doi.org/10.1007/978-3-031-54827-7_5
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-54826-0
Online ISBN: 978-3-031-54827-7
eBook Packages: Computer ScienceComputer Science (R0)