2023年生成式AI大模型发展热点回眸
Review on hot topics of large generative AI models
邓佳文 1任福继1
作者信息
- 1. 电子科技大学计算机科学与工程学院,成都 611731
- 折叠
摘要
2023年,生成式大模型技术发展迅速,取得了一系列突破性的进展.回顾了2023年备受关注的大模型关键技术,包括大模型的能力涌现、多模态大语言模型的发展,以及大模型对齐和知识增强技术.介绍了大模型技术在医疗、教育等领域的垂直应用,以及对AI智能体与元宇宙技术发展的促进作用.此外,从数据隐私、有偏价值观、版权争议及虚假新闻传播等方面讨论了生成式大模型技术面临的安全挑战和发展趋势.
Abstract
In 2023 the development of large generative AI models achieved a series of breakthroughs.This paper reviews the key technologies of large models that have received significant attention,including the emergent abilities of large language model(LLM),development of multimodal LLMs,as well as alignment and knowledge enhancement techniques of large models.The paper also introduces the vertical applications of generative AI in various fields such as healthcare and education,and its role in advancing the development of AI agents and metaverse technologies.Additionally,the challenges and developmental trends in generative AI are discussed,including issues related to data privacy,biased values,copyright disputes,and fake news propagation.
关键词
生成式人工智能/大模型/人工智能内容生成/AI安全Key words
generative AI/large language model(LLM)/artificial intelligence generated content(AIGC)/AI safety引用本文复制引用
出版年
2024
NETL
NSTL
万方数据