摘要
ChatGPT的成功把生成式人工智能逼近通用智能的重大问题推到了前台。生成式人工智能取得如此成果绝非偶然,当其以技术逻辑的方式使学习之于智能生成与发展的建构性意义展开为人工智能巨大能力的时候,这既是技术逻辑的重要发展,更是技术追求智能本质的重要进展。因为生成式人工智能逼近通用智能绝非仅仅是人工智能技术范式的重要进步与重大发展,而是因其对智能的哲学追问真正推进了人工智能本身。在我们看来,当生成式人工智能基于海量数据来发挥技术“大力出奇迹”的优势,使深度学习的提炼获得超越数据规律的原则之时,当生成式人工智能利用事态组合来展现系统“多样尽可能”的尝试,使自然学习的反思成为系统类主体的能力之时,当生成式人工智能通过引用价值偏好来实现“总结现智能”的可能,使强化学习的交互为系统引入经验、升华经验,实现了人工智能系统智能的升级之时,生成式人工智能就不仅获得了强大的力量,而且真正地逼近了通用智能。
The success of Chat GPT has pushed the major issue of Generative Artificial Intelligence approaching General Intelligence to the foreground.It is no accident that Generative Artificial Intelligence has achieved such results,when the constructive significance of learning to the generation and development of intelligence unfolds into the enormous capabilities of artificial intelligence in the form of technical logic,this is not only an important development of technical logic,but also an important progress in the pursuit of the essence of intelligence in technology.Because Generative Artificial Intelligence approaching general intelligence is by no means just an important advancement and major development of the artificial intelligence technology paradigm,it is because of its philosophical questioning of intelligence that it has really advanced artificial intelligence itself.In our opinion,when Generative Artificial Intelligence is based on massive data to give full play to the advantages of“powerful miracles”,the refinement of deep learning can obtain the principle of surpassing the laws of data;to use the combination of situations to show the system s“variety as much as possible”attempt,so that the reflection of natural learning becomes the ability of the subject of the system;to realize the possibility of“summarizing intelligence”by citing value preferences,so that the interaction of reinforcement learning can introduce and sublimate experience for the system,and the intelligence upgrade of the artificial intelligence system is realized,the Generative Artificial intelligence will not only gain a powerful force,but also truly approach the general intelligence.
作者
涂良川
TU Liang-chuan(School of Marxism,South China Normal University,Guangzhou 510631,China)
出处
《东北师大学报（哲学社会科学版）》
CSSCI
北大核心
2023年第4期40-47,93,共9页
Journal of Northeast Normal University(Philosophy and Social Science Edition)
基金
国家社会科学基金一般项目(21BZX002)
广东省教育科学规划项目(2021GXJK079)。
关键词
生成式人工智能
学习
哲学
通用智能
Generative Artificial Intelligence
Learning
Philosophy
General Intelligence
