Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
41 tokens/sec
GPT-4o
59 tokens/sec
Gemini 2.5 Pro Pro
41 tokens/sec
o3 Pro
7 tokens/sec
GPT-4.1 Pro
50 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Can AI Be as Creative as Humans? (2401.01623v4)

Published 3 Jan 2024 in cs.AI and cs.CL

Abstract: Creativity serves as a cornerstone for societal progress and innovation. With the rise of advanced generative AI models capable of tasks once reserved for human creativity, the study of AI's creative potential becomes imperative for its responsible development and application. In this paper, we prove in theory that AI can be as creative as humans under the condition that it can properly fit the data generated by human creators. Therefore, the debate on AI's creativity is reduced into the question of its ability to fit a sufficient amount of data. To arrive at this conclusion, this paper first addresses the complexities in defining creativity by introducing a new concept called Relative Creativity. Rather than attempting to define creativity universally, we shift the focus to whether AI can match the creative abilities of a hypothetical human. The methodological shift leads to a statistically quantifiable assessment of AI's creativity, term Statistical Creativity. This concept, statistically comparing the creative abilities of AI with those of specific human groups, facilitates theoretical exploration of AI's creative potential. Our analysis reveals that by fitting extensive conditional data without marginalizing out the generative conditions, AI can emerge as a hypothetical new creator. The creator possesses the same creative abilities on par with the human creators it was trained on. Building on theoretical findings, we discuss the application in prompt-conditioned autoregressive models, providing a practical means for evaluating creative abilities of generative AI models, such as LLMs. Additionally, this study provides an actionable training guideline, bridging the theoretical quantification of creativity with practical model training.

PDF HTML Abstract

Insights into AI's Potential for Creativity: A Study of Relative and Statistical Creativity

This paper rigorously investigates the complex question of whether AI can be as creative as humans. The authors approach the subject by introducing foundational concepts, such as Relative Creativity and Statistical Creativity, which establish a matrix for evaluating AI's creativity in comparison to human creativity.

The paper purposefully sidesteps the contentious issue of defining creativity in absolute terms, an endeavor that has historically led to numerous scholarly debates. Instead, it frames the notion of creativity from a relative standpoint, drawing parallels with the Turing Test for intelligence. The authors propose that AI can be deemed creative if it can generate outputs indistinguishable from those produced by humans when evaluated within a probabilistic framework. This fresh perspective offers a quantifiable means of assessing AI's creative abilities without being hindered by subjective interpretations of creativity.

In terms of methodology, the introduction of Statistical Creativity further facilitates theoretical exploration by juxtaposing AI's creative output with that of observable human creators. This approach leverages statistical comparison as a robust metric, bridging theoretical constructs with empirical validation, and enabling AI models to demonstrate creativity by accurately fitting human-generated data.

Moreover, the paper delineates practical implications, especially in the context of contemporary AI models like autoregressive and prompt-conditioned autoregressive models. In particular, the authors derive Prompt-Contextualized Autoregressive Statistical Creativity as a nuanced articulation of AI creativity within the confines of LLMs, such as GPT-4 and Llama. These models, which are contingent on generating text based on given prompts, are assessed for their ability to mimic the creative output of humans based on rigorous statistical metrics.

Beyond theoretical formulation, the paper offers actionable insights into the engineering and training of AI systems. Through an analysis of AI training processes, it underscores the importance of collecting substantial conditional data that encompasses the generative conditions of creative works. This approach challenges the prevalent trend of accumulating large datasets without considering the nuanced conditions under which data were created, and it suggests that successful assimilation of conditional generation data could enable AI to act as a hypothetical new human creator.

The implications of this research extend to both theoretical and practical domains. The paper illuminates the potential for AI to not only replicate but possibly exceed the creative capacities of human creators by aligning with rigorous mathematical principles rather than nebulous philosophical debates. The concept of reducing creativity to its statistical essence offers a promising avenue for future AI developments.

In conclusion, this paper propounds a structured theoretical framework that advances the discourse on AI creativity and delivers methodologies for assessing and cultivating creative capabilities in AI models. As AI's reach continues to expand, these insights could pave the way for novel creative applications, propelling AI closer to achieving human-like creativity. The paper's emphasis on data-centric training regimes and statistical creativity assessments serves as a blueprint for both researchers and practitioners aiming to push the boundaries of what AI models can achieve creatively.

File Document Download Save Streamline Icon: https://streamlinehq.com PDF File Document Download Save Streamline Icon: https://streamlinehq.com Markdown Bookmark Chat Bubble Oval Streamline Icon: https://streamlinehq.com Chat (Pro)
Definition Search Book Streamline Icon: https://streamlinehq.com
References (61)
  1. Creating Creativity: 101 Definitions (what Webster Never Told You). Alden B. Dow Creativity Center Press, Midland, MI.
  2. Amabile, T. M. (1996). Creativity and innovation in organizations, volume 5. Harvard Business School Boston.
  3. Rademacher and gaussian complexities: Risk bounds and structural results. Journal of Machine Learning Research, 3(Nov):463–482.
  4. Boden, M. A. (2003). The Creative Mind: Myths and Mechanisms. Routledge, London, UK.
  5. On the opportunities and risks of foundation models.
  6. Olá, bonjour, salve! XFORMAL: A benchmark for multilingual formality style transfer. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 3199–3216, Online. Association for Computational Linguistics.
  7. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901.
  8. Câmara Pereira, F. (2007). Creativity and artificial intelligence: a conceptual blending approach. Mouton de Gruyter.
  9. Expertise style transfer: A new task towards better communication between experts and laymen. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 1061–1071, Online. Association for Computational Linguistics.
  10. Evaluating prose style transfer with the bible. Royal Society open science, 5(10):171920.
  11. Unsupervised cross-lingual representation learning at scale. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 8440–8451, Online. Association for Computational Linguistics.
  12. Plug and play language models: A simple approach to controlled text generation. arXiv preprint arXiv:1912.02164.
  13. BERT: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), pages 4171–4186, Minneapolis, Minnesota. Association for Computational Linguistics.
  14. CAN: Creative Adversarial Networks, Generating ”Art” by Learning About Styles and Deviating from Style Norms. arXiv:1706.07068 [cs].
  15. Flek, L. (2020). Returning the N to NLP: Towards Contextually Personalized Classification Models. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 7828–7838, Online. Association for Computational Linguistics.
  16. Foster, D. (2022). Generative deep learning. ” O’Reilly Media, Inc.”.
  17. On the creativity of large language models. arXiv preprint arXiv:2304.00008.
  18. Creative sketch generation. In International Conference on Learning Representations.
  19. Size-independent sample complexity of neural networks. In Conference On Learning Theory, pages 297–299. PMLR.
  20. Generative adversarial nets. Advances in neural information processing systems, 27.
  21. Don’t stop pretraining: Adapt language models to domains and tasks. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 8342–8360.
  22. A probabilistic formulation of unsupervised text style transfer. arXiv preprint arXiv:2002.03912.
  23. Hertzmann, A. (2018). Can computers create art? In Arts, volume 7, page 18. MDPI.
  24. Hoeffding, W. (1963). Probability inequalities for sums of bounded random variables. Journal of the American Statistical Association, pages 13–30.
  25. Hovy, D. (2015). Demographic Factors Improve Classification Performance. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 752–762, Beijing, China. Association for Computational Linguistics.
  26. Deep Learning for Text Style Transfer: A Survey. Computational Linguistics, 48(1):155–205.
  27. Jordanous, A. (2016). Four PPPPerspectives on Computational Creativity in Theory and in Practice. Connection Science, 28(2):294–216.
  28. Highly accurate protein structure prediction with alphafold. Nature, 596(7873):583–589.
  29. How does information bottleneck help deep learning? In International Conference on Machine Learning (ICML).
  30. Robustness implies generalization via data-dependent generalization bounds. In International Conference on Machine Learning, pages 10866–10894. PMLR.
  31. Generalization in deep learning. Mathematical Aspects of Deep Learning, pages 112–148.
  32. Kirkpatrick, K. (2023). Can ai demonstrate creativity? Communications of the ACM, 66(2):21–23.
  33. Mitigating political bias in language models through reinforced calibration. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, pages 14857–14866.
  34. Politeness transfer: A tag and generate approach. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 1869–1881, Online. Association for Computational Linguistics.
  35. Foundations of machine learning. MIT press.
  36. OpenAI (2023). Gpt-4 technical report.
  37. Style transfer through back-translation. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 866–876, Melbourne, Australia. Association for Computational Linguistics.
  38. Ranaweera, W. L. (2016). Exquimo: An exquisite corpse tool for co-creative 3d shape modeling.
  39. Dear sir or madam, may I introduce the GYAFC dataset: Corpus, benchmarks and metrics for formality style transfer. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers), pages 129–140, New Orleans, Louisiana. Association for Computational Linguistics.
  40. High-resolution image synthesis with latent diffusion models.
  41. The standard definition of creativity. Creativity research journal, 24(1):92–96.
  42. Countergedi: A controllable approach to generate polite, detoxified and emotional counterspeech. In Raedt, L. D., editor, Proceedings of the Thirty-First International Joint Conference on Artificial Intelligence, IJCAI-22, pages 5157–5163. International Joint Conferences on Artificial Intelligence Organization. AI for Good.
  43. Sawyer, K. (2012). Extending sociocultural theory to group creativity. Vocations and Learning, 5(1):59–75.
  44. Design: Design inspiration from generative networks. In Proceedings of the European Conference on Computer Vision (ECCV) Workshops, pages 0–0.
  45. Personality, Gender, and Age in the Language of Social Media: The Open-Vocabulary Approach. PLOS ONE, 8(9):e73791. Publisher: Public Library of Science.
  46. Controlling politeness in neural machine translation via side constraints. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 35–40, San Diego, California. Association for Computational Linguistics.
  47. Understanding machine learning: From theory to algorithms. Cambridge university press.
  48. Diffusion art or digital forgery? investigating data replication in diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 6048–6058.
  49. M.b.a. students vs. chatgpt: Who comes up with more innovative ideas? The Wall Street Journal.
  50. Llama: Open and efficient foundation language models.
  51. Treffinger, D. J. (1996). Creativity, Creative Thinking, and Critical Thinking: In Search of Definitions. Center for Creative Learning, Sarasota, FL.
  52. Turing, A. M. (2009). Computing machinery and intelligence. Springer.
  53. Evaluating neural text simplification in the medical domain. In WWW’19 The World Wide Web Conference (WWW), pages 3286–3292, United States. Association for Computing Machinery (ACM). WWW 2019 : The Web Conference 2019, 30 years of the web, WWW’19 ; Conference date: 13-05-2019 Through 17-05-2019.
  54. Concept decomposition for visual exploration and inspiration. arXiv preprint arXiv:2305.18203.
  55. Unsupervised clinical language translation. In Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, KDD ’19, page 3121–3131, New York, NY, USA. Association for Computing Machinery.
  56. Using a knowledge base to automatically annotate speech corpora and to identify sociolinguistic variation. In Proceedings of the Thirteenth Language Resources and Evaluation Conference, Marseille, France. European Language Resources Association.
  57. Robustness and generalization. Machine learning, 86(3):391–423.
  58. Fit and diverse: Set evolution for inspiring 3d shape galleries. ACM Transactions on Graphics (TOG), 31(4):1–10.
  59. Paraphrasing for style. In Proceedings of COLING 2012, pages 2899–2914.
  60. A survey of large language models. ArXiv, abs/2303.18223.
  61. A monolingual tree-based translation model for sentence simplification. In Proceedings of the 23rd International Conference on Computational Linguistics (Coling 2010), pages 1353–1361, Beijing, China. Coling 2010 Organizing Committee.
User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (11)
  1. Haonan Wang (84 papers)
  2. James Zou (232 papers)
  3. Michael Mozer (17 papers)
  4. Linjun Zhang (70 papers)
  5. Anirudh Goyal (93 papers)
  6. Alex Lamb (45 papers)
  7. Zhun Deng (38 papers)
  8. Michael Qizhe Xie (3 papers)
  9. Hannah Brown (7 papers)
  10. Kenji Kawaguchi (147 papers)
  11. Weijie J Su (2 papers)
Citations (9)
View on Semantic Scholar
Youtube Logo Streamline Icon: https://streamlinehq.com

YouTube