background

演讲嘉宾

张钹院士

张钹

个人简介

张钹,国际著名计算机科学专家,中国科学院院士,俄罗斯自然科学院外籍院士, 清华大学计算机系教授,清华大学人工智能研究院名誉院长,德国汉堡大学自然科学名誉博士(2011 年授予), 获得 CCF (中国计算机学会)终身成就奖(2014 年),吴文俊人工智能科学技术奖最高成就奖(2019 年)。 张钹院士是智能技术与系统国家重点实验室创建者之一,并于 1990-1996 年担任智能技术与系统国家重点实验室主任。 张钹院士长期从事人工智能、人工神经网络和机器学习等理论研究以及模式识别、知识工程和机器人等应用技术研究, 在相关领域发表学术论文 200 余篇,编撰学术专著 5 部(章),科研成果曾获 ICL 欧洲人工智能奖等。

报告题目

生成式人工智能时代的机遇与挑战。

报告摘要

分析生成式人工智能时代的特点,从生成式模型具有的强大的生成能力、迁移能力和交互能力,阐述对社会可能造成的影响,以及为我们提供的机遇。并从生成式模型所存在的问题,分析我们所面临的挑战。

吴信东教授

吴信东

个人简介

吴信东,俄罗斯工程院外籍院士,美国科学促进协会会士,IEEE Fellow([美国]电气电子工程师学会会士),合肥工业大学“大数据知识工程”教育部重点实验室主任,“营销智能”国家新一代人工智能开放创新平台负责人。之江实验室高级研究专家。

报告题目

华谱系统:为华夏写史,助百姓寻根

报告摘要

华谱系统(https://www.zhonghuapu.com/)面向所有华人姓氏,接纳任何年代和任何地点的华夏人物,支持人人参与的信息录入,提供多方位的信息融合和检索,为用户提供电子化修谱、家谱打印、人物查询、人物关联、社区建设和隐私保护。我们介绍华谱系统的三级知识层次和知识图谱构建,并分析它们在社会媒体信息处理中的应用。

Professor Weizhu Chen

Weizhu Chen

Profile

Weizhu Chen is a Vice President at Microsoft, where he leads the science division for large models and Natural Language Processing in Azure AI. His work primarily focuses on the post-training and inference of Large Language Models (LLMs), their measurement and customization, and ensuring AI quality for NLP applications. In 2020, his team developed the Low-Rank Adaptation (LoRA) method as they worked on adapting the 175-billion parameter GPT-3 model. LoRA has since become a crucial component in LLMs for numerous Microsoft products and has significantly contributed to the broader community by increasing the efficiency of adapting various large models. In addition to pivotal product developments like GitHub Copilot, where his team was responsible for AI quality and LLM integration, his team has made a broad contribution to the open-source community with projects like LoRA, DeBERTa, MT-DNN, and R-Adam. Weizhu Chen joined Microsoft in 2005 and obtained his PhD degree from HKUST in 2012.

Report title

LoRA: Low-Rank Adaptation for Large Language Models

Report summary

Low-Rank Adaptation (LoRA) has established itself as a preferred method for fine-tuning Large Language Models (LLMs) with remarkable efficiency and simplicity. In this keynote, I will delve into the journey of LoRA, tracing its roots back to its inception in 2020. I will uncover the motivations behind its creation, the innovative strides it has taken, and why it stands out amidst the myriad of alternatives, especially in the challenging context of fine-tuning the 175B parameter GPT-3 model. The talk will also shed light on some unexpected revelations and novel insights gained when implementing LoRA in real-world applications.

As we pivot to the present, the talk will offer an examination of the contemporary best practices in the field. We will discuss the various enhancements and optimizations that have been made to LoRA for different use cases, aiming for better efficiency. Additionally, the wide-ranging applicability of LoRA across diverse domains will be highlighted, showcasing its versatility and effectiveness.

Looking ahead, we will navigate through the ongoing research endeavors, emerging trends, and envision the potential evolution of LoRA. This exploration will be contextualized within the backdrop of rapid advancements in quantization technology to reduce memory, the growing needs for efficient inference to reduce cost, and the continuous quest for maximized model efficiency in both training and serving LLMs.

张勇东教授

张勇东

个人简介

张勇东,教授,博士生导师,现任中国科学技术大学信息科学技术学院执行院长,人民日报社传播内容认知全国重点实验室首席科学家。国家自然基金委创新研究群体项目负责人(2021 年),“万人计划”科技创新领军人才 (2018 年),国家杰出青年科学基金获得者 (2015年)。曾获国家自然科学奖二等奖(排名第一,2019 年), 教育部技术发明奖一等奖(排名第一,2022年),中国电子学会科学技术奖(自然科学类)一等奖(排名第一,2018 年),国家科技进步奖二等奖(排名第五,2016 年),北京市科学技术奖一等奖 (排名第一,2014 年)。研究成果大规模应用于国家网络空间内容安全领域,取得了显著的应用效果。担任《中国通信》副主编,国家重点研发计划-“变革性技术关键科学问题”重点专项总体专家组成员,国家重点研发计划-“社会治理与智慧社会科技支撑”重点专项总体专家组成员。

报告题目

全媒体环境下智能传播技术体系

报告摘要

随着第五次传播革命不断发展,出现了全程媒体、全息媒体、全员媒体、全效媒体。信息无处不在、无所不及、无人不用的全媒体环境导致舆论生态、媒体格局、传播方式发生深刻变化,同时也给我国舆论安全和意识形态安全带来严峻挑战,急需发展以主流价值观为指导,社会传播理论和人工智能技术相结合的智能传播技术体系。本报告首先简述智能传播的当前背景,然后阐述智能传播难点挑战,最后深入探讨智能传播当前技术局限和下一步的解决路径。

喻国明教授

喻国明

个人简介

教育部长江学者特聘教授、第七届国务院学位委员会新闻传播学学科评议组成员、现为北京师范大学新闻传播学院学术委员会主任、北京师范大学传播创新与未来媒体实验平台主任,兼任北京市社会科学联合会副主席、中国新闻史学会传媒经济与管理专业委员会理事长、《中国社会舆情年度报告(蓝皮书)》主编、《中国互联网营销年度报告(蓝皮书)》主编等,是我国传播学实证研究的领军者、传媒经济学的奠基人及认知神经传播学的开创者之一。迄今为止,独著、合著出版的学术专著、教材、蓝皮书共46余本,论文1000余篇,迄今为止,在新闻传播学科的论文总被引文数居全国第一位。

报告题目

理解生成式AI:融通机器智能与人类智能的算法媒介

报告摘要

ChatGPT问世掀起生成式AI研究热潮,讲座从复杂性范式出发,依循“是什么-会怎样-应如何”的逻辑分析生成式AI的理解路径。生成式AI是实现传播理性与非理性要素交织的新媒介技术,将内容网络升维成更具开放性的复杂巨系统。从社会影响来看,生成式AI作为智能主体和智能工具,可通过“替代”与“增强”人类脑力的方式,促进人类非理性逻辑与机器理性逻辑交织并深入社会表达中,使普罗大众得以跨越传播“能力沟”并实现平均水平的智力增强,进一步打破精英宰制的社会并迈入“常人政治”的未来新社会。治理方面,应摒弃机械控制论思路,以其道德伦理和算法伦理为原则,形成抓大放小的复杂性治理思路。面对未来发展,为避免陷入科林格里奇困境,应沿智能试错方案的思路,允许通过技术小规模试错来厘清发展路径,研究也应更关注生成式AI具体应用层的议题。

Chi Wang

Chi Wang

个人简介

Chi Wang is a principal researcher in Microsoft Research AI Frontiers. He has worked on large language model and AI frameworks, automated machine learning, machine learning for systems, scalable solutions for data science and data analytics, and knowledge mining from text data and graph data (with a SIGKDD Data Science/Data Mining PhD Dissertation Award). Chi is the creator of AutoGen, a popular and rapidly growing open-source framework for enabling next-gen AI applications. Chi is the creator of FLAML, a fast open-source library for AutoML & tuning used widely inside and outside Microsoft. Chi has a PhD in Computer Science from University of Illinois at Urbana-Champaign, and a BS in Computer Science from Tsinghua University.

报告题目

AutoGen: Enabling Next-Gen AI Applications via Multi-Agent

报告摘要

Large AI Models demonstrate promising capabilities and open numerous possibilities for innovative applications. What are future AI applications like and how do we empower every developer to build them? AutoGen is a pioneering attempt to address this question as a generic multi-agent conversation framework. This talk will explore the core functionalities and key concepts of AutoGen, explain how it can be used to simplify and unify the implementation of complex AI workflows with integration of models, tools, and human inputs, and illustrate how it is applied across a broad spectrum of tasks and industries, paving the way for next-generation AI applications.

陈恩红教授

陈恩红

个人简介

陈恩红,中国科学技术大学 讲席教授、博士生导师,校学术委员会和学位委员会委员,大数据学院执行院长,认知智能全国重点实验室副主任。国家杰出青年基金获得者,国家级创新领军人才,科技部重点研发计划项目首席科学家,科技部重点领域创新团队“大数据分析及应用”团队负责人,大数据分析与应用安徽省重点实验室主任,安徽省计算机学会理事长。教育部应用伦理教指委副主任、计算机类专业教指委委员。主持了科技部重点研发计划项目、基金委重大仪器研制项目及区域联合基金重点项目。TKDE、 软件学报等多个国内外学术期刊编委,获KDD2008最佳应用论文奖、ICDM2011最佳研究论文奖、SDM2015最佳论文提名奖、KDD2018最佳学生论文奖等,作为第一完成人获得教育部自然科学一等奖、吴文俊人工智能科技进步一等奖等。

报告题目

面向富语义社交媒体的多模态认知智能

报告摘要

多模态认知智能已成为人工智能发展的主流趋势之一,旨在通过多模态语义知识的获取、表示与推理,有效支撑面向社交媒体场景的富语义下游应用。然而,现有的多模态分析技术难以感知与整合深度的语义线索,而知识工程技术路径则由于高质量图文对的稀缺而面临性能瓶颈。与此同时,方兴未艾的多模态大模型虽然掌握了丰富的语义线索,但存在可靠性弱、推理能力差等问题,且面临幻觉等因素的干扰,难以独挑大梁。针对上述问题,报告将系统性介绍团队近年来在多模态认知智能方面的研究成果,着重围绕多模态实体认知、多模态语义关联等方面的技术路径探索,以及将多模态知识与大模型相结合的应用尝试。

黄民烈教授

黄民烈

个人简介

黄民烈,清华大学长聘教授,博士生导师,国家杰青获得者,计算机系智能技术与系统实验室副主任,清华大学基础模型中心副主任,自然语言生成与智能写作专委会副主任、CCF学术工委秘书长。他的研究领域为大规模语言模型、对话系统、语言生成,著有《现代自然语言生成》一书。承担国家自然科学基金重点项目、面上项目、青年基金多项,多次参与国家重大研发计划项目。曾获得中国人工智能学会吴文俊人工智能科技进步奖一等奖(第一完成人),中文信息学会汉王青年创新奖,微软合作研究奖等。在国际顶级会议和期刊发表论文150多篇,谷歌学术引用16000多次,h-index 63,入选2022年Elsevier中国高被引学者,连续三年入选AI 2000全球最有影响力AI学者榜单;多次获得国际主流会议的最佳论文或提名(IJCAI、ACL、SIGDIAL等)。研发任务型对话系统平台ConvLab、ConvLab2,中文对话大模型EVA、OPD、CharacterGLM,智源中文大模型CPM的核心研发成员,国内大模型研究的主要力量之一,研发AI乌托邦拟人对话交互平台。担任顶级期刊TNNLS、TACL、CL、TBD编委,多次担任自然语言处理领域顶级会议ACL/EMNLP资深领域主席。

报告题目

类人智能对话系统(Humanlike AI Systems)

报告摘要

以chatGPT、GPT-4为代表的语言模型重点在解决生产力场景的问题:完成任务,获取信息,提升效率,解放生产力。它们强调AI的机器属性。但人和AI建立社会连接、信任,让AI满足人类的情感、陪伴、心理疏导等需求也非常重要,因此兼具情商和智商的拟人(humanlike)AI才是未来AGI智能体的理想形态,也是构建未来人机共融社会的基础。讲者将分享其在拟人AI方向的研究工作,包括现有大语言模型的情商测试,具有自然性、共情能力、有趣性、安全性的拟人大模型CharacterGLM,以及认知理论结合的深度对话框架。

Tat-Seng Chua

Tat-Seng Chua

个人简介

Dr. Chua is the KITHCT Chair Professor at the School of Computing, National University of Singapore (NUS). He is also the Distinguished Visiting Professor of Tsinghua University, the Visiting Pao Yue-Kong Chair Professor of Zhejiang University, and the Distinguished Visiting Professor of Sichuan University. Dr. Chua was the Founding Dean of the School of Computing from 1998-2000. His main research interests include unstructured data analytics, video analytics, conversational search and recommendation, and robust and trustable AI. He is the co-Director of NExT, a joint research Center between NUS and Tsinghua University.

Dr Chua is the recipient of the 2015 ACM SIGMM Achievements Award, and the winner of the 2022 NUS Research Recognition Award. He is the Chair of steering committee of Multimedia Modeling (MMM) conference series, and ACM International Conference on Multimedia Retrieval (ICMR) (2015-2018). He is the General Co-Chair of ACM Multimedia 2005, ACM SIGIR 2008, ACM Web Science 2015, ACM MM-Asia 2020, WSDM 2023, and the upcoming TheWebConf (or WWW) 2024. He serves in the editorial boards of several international journals. Dr. Chua is the co-Founder of two technology startup companies in Singapore.

报告题目

Towards a Safe and Trustable Framework for Generative AI Agents

报告摘要

The emergence of large language models (LLM’s) that offer significant capabilities in content comprehension, content generation, and flexible dialogues, has the potential to revolutionize the ways we seek and consume information. It also enables AI to emulate capabilities of humans, and permits autonomous AI agents to be developed to tackle a wide range of applications, such as the social media analytics, search and recommendation. However, before such systems can be widely used and accepted, we need to address several challenges, including that of trust and safety in using such systems. This talk will present the framework for generative AI agents and their applications, as well as the issues of trust, safety and accessibility.