生成式AI与通用人工智能（AGI）基础理论最新研究成果综述
模型链 (重庆) 科技有限公司, 重庆 401120
| 摘 要： | 本文聚焦于生成式AI，作为当代人工智能（AI）与通用人工智能（AGI）典型代表，旨在深入探讨其基础理论方面的最新研究进展。其方法是通过对比分析传统AI与当代AI在底层逻辑、形式化理解上的差异，并进一步探讨传统AI的三个核心观点（符号主义、联结主义、行为主义）与当代AI三大流派（基于大型语言模型（LLM）的生成式AI/AGI，如ChatGPT；以小模型为特征的新质生产力AGI，如I3DNA；兼容于大小两类模型的基于孪生图灵机双重形式化理解模型）之间的区别。其结果是揭示出了AI与AGI基础理论的核心构成：序位逻辑、联动函数以及基于数智文本的广义双语与广义翻译所遵循的三大基本定律。其意义在于：不仅增强了以ChatGPT为代表的基于LLM的生成式AI/AGI的可解释性，而且还为以小模型为特征的新质生产力AGI及其复杂的宇宙智能论和天下模型理论提供广义翻译，同时展示了孪生图灵机作为兼容并蓄的智能体在整合数据、知识、算力、算法及人机互助新时代认知范型的潜力，为构建超级智慧系统奠定了基础。 |
| 关 键 词： | 生成式AI; 通用人工智能; 符号主义; 联结主义; 行为主义; 大型语言模型; 孪生图灵机; 新质生产力AGI |
| DOI： | 10.57237/j.cst.2024.03.004 |
View:
Downloads:
收稿日期: 2024-07-31
接受日期: 2024-08-28
在线出版日期: 2024-09-09
A Review of the Latest Research Achievements in the Basic Theory of Generative AI and Artificial General Intelligence (AGI)
Model Chain (Chongqing) Technology Co., Ltd., Chongqing 401120, China
| Abstract: | This paper focuses on generative AI, a typical representative of contemporary artificial intelligence (AI) and artificial general intelligence (AGI), aiming to delve into the latest research progress in its basic theory. The research method involves a comparative analysis of the differences in underlying logic and formal understanding between traditional AI and Current AI, further exploring the distinctions between the three core viewpoints of traditional AI (symbolism, connectionism, behaviorism) and the three major schools of Current AI (generative AI/AGI based on large language models (LLMs) such as ChatGPT; new quality productive force AGI characterized by small models, such as I3DNA; and twin Turing machines based on dual formal understanding models that are compatible with both large and small models). The research reveals the core components of the basic theory of AI and AGI: bit-list logic, linkage functions, followed by generalized bilingualism or generalized translation based on digital and intelligent text with the three fundamental laws. The significance of this research lies in not only enhancing the interpretability of generative AI/AGI based on LLMs represented by ChatGPT but also providing generalized translations for the new quality productive force AGI characterized by small models and its complex theories of cosmic intelligence and the universal model series. At the same time, it demonstrates the potential of twin Turing machines as inclusive intelligent agents in integrating data, knowledge, computing power, algorithms, and human-computer mutual assistance in the new era of cognitive paradigms, laying the foundation for constructing super intelligent systems. |
| Keywords: | Generative AI; Artificial General Intelligence (AGI); Symbolism; Connectionism; Behaviorism; Large Language Models (LLMs); Twin Turing Machines; New Productive Force AGI |
参考文献
| [1] | Russell, S. J., & Norvig, P. (2020). Artificial Intelligence: A Modern Approach. Pearson. |
| [2] | LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444. |
| [3] | Chomsky, N. (1956). Three models for the description of language. IRE Transactions on Information Theory, 2(3), 113-124. |
| [4] | Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1986). Learning representations by back-propagating errors. Nature, 323(6088), 533-536. |
| [5] | Brooks, R. A. (1991). Intelligence without representation. Artificial Intelligence, 47(1-3), 139-159. |
| [6] | Bengio, Y. (2009). Learning deep architectures for AI. Foundations and Trends in Machine Learning, 2(1), 1-127. |
| [7] | Silver, D., Huang, A., Maddison, C. J., Guez, A., Sifre, L., van den Driessche, G.,... & Hassabis, D. (2016). Mastering the game of Go with deep neural networks and tree search. Nature, 529(7587), 484-489. |
| [8] | Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Wauth, S.,... & Liang, P. (2021). On the opportunities and risks of foundation models. arXiv preprint arXiv: 2108.07258. |
| [9] | Hinton, G. (2021). The forward-forward algorithm: Some preliminary investigations. arXiv preprint arXiv: 2112.02408. |
| [10] | Chollet, F. (2019). On the measure of intelligence. arXiv preprint arXiv: 1911.01547. |
| [11] | Kaplan, A., & Haenlein, M. (2019). Siri, Siri, in my hand, who’s the fairest in the land? On the interpretations, illustrations, and implications of artificial intelligence. Business Horizons, 62(1), 15-25. |
| [12] | Tegmark, M. (2017). Life 3.0: Being human in the age of artificial intelligence. Knopf. |
| [13] | Marcus, G. (2022). The next decade in AI: Four steps towards robust artificial intelligence. arXiv preprint arXiv: 2202.07218. |
| [14] | Shoham, Y., Perrault, C., Brynjolfsson, E., Clark, J., Etzioni, O., Grosz, B.,... & Slonim, A. (2022). Artificial intelligence index 2022 annual report. Stanford Institute for Human-Centered AI. |
| [15] | Zou X.(2023). New Opportunities for AI Innovation with Big Data: Indirect Docking between GLPS and LLM [C] 2023 6th International Conference on Artificial Intelligence and Big Data (ICAIBD). |
| [16] | Zou, X., Zou, S., & Wang, X.. (2019). The Strategy of Constructing an Interdisciplinary Knowledge Center. The International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery. Springer, Cham. |
| [17] | Zou, S., Zou, X., & Wang, X.. (2018). How to Understand the Fundamental Laws of Information. International Conference on Cognitive Systems and Signal Processing. Springer, Singapore. |
| [18] | Maimaiti, M., Zou, S., Wang, X., & Zou, X.. (2019). How to Understand: Three Types of Bilingual Information Processing?. International Conference on Cognitive Systems and Signal Processing. |
| [19] | Zou, X., Zou, S., & Wang, X.. (2019). New Approach of Big Data and Education: Any Term Must Be in the Characters Chessboard as a Super Matrix. International Conference on Big Data. ACM Press. |
| [20] | Lv Chenjun, & Zou Xiaohui. (2020). A Preliminary Study on the Mathematical Basis of General Artificial Intelligence. Research on Dialectics of Nature, 36(3), 7. |
