期刊文献+

从AIGC到AIGA,智能新赛道:决策大模型 被引量:1

From AIGC to AIGA,the New Frontier of Artificial Intelligence:Large Decision-Making Models
原文传递
导出
摘要 随着生成式人工智能技术的不断发展,及其与深度学习的结合,内容生成式智能技术AIGC应运而生,孕育了一系列内容生成式大模型,如GPT、DALL-E等。这些大模型能够生成高质量、逼真的内容,在自然语言处理、计算机视觉、语音识别等领域展现出了强大性能。随着大模型技术的不断进步,出现了大模型加大数据的科研范式以及预训练模型加提示的学习框架,为科研人员提供了强大的技术支持。在内容生成式大模型的基础上构建智能体,可以生成决策,实现数字世界和真实世界的交互,从而走向更为通用的决策大模型,帮助人们快速、准确地分析和理解大量复杂数据,支持各种物理域与社会域的决策过程。因此,决策生成式智能技术AIGA将成为智能领域的新赛道,目前已在棋类竞技、交通拥堵、医疗诊断等领域涌现出了令人振奋的创新。该文主要介绍了AIGC与AIGA的发展历程,特别是其与大模型的结合情况,并对其技术特点和所面临的挑战进行了概括。 With the continuous development of generative artificial intelligence(AI)technology,along with its integration with deep learning,the technology of content generative artificial intelligence,AIGC,has emerged.It has given rise to a series of large-scale content-generative models,such as GPT and DALL-E.These large models are capable of generating high-quality,realistic content and have demonstrated powerful performance in fields like natural language processing,computer vision,and speech recognition.With the ongoing advancement of large model technology,research paradigms such as scaling up models with more data and pre-training models with prompts have emerged,providing strong technical support for researchers.Building intelligent agents on the basis of content-generative models enables decision-making,facilitating interaction between the digital world and the real world,thereby moving towards more generalized decision-making models.This helps people analyze and understand large amounts of complex data quickly and accurately,supporting decisionmaking processes in various physical and social domains.Therefore,decision generative artificial intelligence technology,AIGA,will become a new track in the field of artificial intelligence.It has already shown exciting innovations in areas such as chess competition,traffic congestion,and medical diagnosis.This article mainly introduces the development of AIGC and AIGA,especially their integration with large models,and summarizes their technical characteristics and the challenges they face.
作者 谢正 李浩 宋伊萍 梁栋 陈颖 Xie Zheng;Li Hao;Song Yiping;Liang Dong;Chen Ying(College of Science,National University of Defense Technology,Changsha 430000,China;Department of CryptologyScience&Technology,Beijing Electronic Science&Technology Institute,Beijing 100070,China)
出处 《科学观察》 2024年第2期14-33,共20页 Science Focus
关键词 生成式智能 深度学习 智能决策 统计与优化 大模型 generative artificial intelligence deep learning decision-making artificial intelligence statistics and optimization large models
  • 相关文献

参考文献3

二级参考文献11

共引文献21

引证文献1

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部