Big names gathered to inspire the future│The third phase of the Artificial Intelligence Innovation Technology Workshop concluded perfectly

On August 19-20, 2023, the China Artificial Intelligence Society hosted the 3rd Artificial Intelligence Innovation Technology Workshop "Multimodal Large Model Technology Frontiers and Applications", hosted by the China Artificial Intelligence Society Fellows' Home (Hangzhou Station) and co-organized by Zhejiang Hangzhou Future Science and Technology City Management Committee. This workshop was successfully held using a combination of online and offline methods, with a cumulative number of viewers exceeding 160,000

This workshop focuses on the hot topic "Multimodal Large Models", aiming to promote academic exchanges and promote technological innovation. Professor Qiu Xipeng from Fudan University and Associate Professor Xu Feng from Tsinghua University jointly serve as the academic directors of this workshop. Six experts and scholars with important influence in multi-modal large models and related fields, including Associate Professor Song Ruihua, Associate Professor Dai Jifeng, Dr. Sun Tianxiang, Mr. Wang Bingning, Researcher Zhao Junbo, and Mr. Yan Ming, were also invited to join the lecturer group to teach the audience. It was a wonderful academic feast. During the opening of the workshop, the audience was enthusiastic and actively participated in the interaction. The strong academic atmosphere also continued after class. After the lectures, the lecturers continued to answer questions for the online and offline audiences one by one. The atmosphere was very lively.

Associate Professor Xu Feng

In the two days from August 19th to 20th, the lecturers used a rigorous and interesting way to bring 6 wonderful knowledge sessions to online and offline students and audiences, from theoretical research to practical application. Feast

The content that needs to be rewritten is: Lecture 1: "Multimodal Understanding and Artificial Intelligence Creation" Song Ruihua

On the morning of August 19, Associate Professor Song Ruihua gave the first lecture - "Multimodal Understanding and Artificial Intelligence Creation". Before giving formal lectures, Associate Professor Song Ruihua first explained the concept of "multimodality" in a simple and easy-to-understand manner, its relationship with multimedia, and why it has suddenly become so popular. After introducing the basic information related to "multimodality", Associate Professor Song Ruihua began to introduce Wenlan Model 1.0-3.0, and shared the entire process and results of "Research on Text Embedding Differences between Multimodality and Text Pre-training Models". At the end of the lively lecture, Associate Professor Song Ruihua summarized the trend of multi-modal understanding and creation, and encouraged the audience to move forward to tackle more challenging problems in the future

Associate Professor Song Ruihua

Lecture 2: "Visual Basics Large Model" Dai Jifeng

The second class of the workshop, "Visual Basics Large Model", was taught by Associate Professor Dai Jifeng. He pointed out in class that general large models are leading the development of artificial intelligence. Focusing on the "very large-scale visual model" and related technical routes, Associate Professor Dai Jifeng gave a wonderful lecture in the following lectures. He not only described the powerful functions of LLM (large-scale language model), but also praised it as A milestone on the road to the development of general artificial intelligence.

Associate Professor Dai Jifeng

The content that needs to be rewritten is: Lecture Three: "Conversational Large-scale Language Model" Sun Tianxiang

On the afternoon of August 19, Dr. Sun Tianxiang held the third course entitled "Conversational Large-scale Language Model". The course content is mainly divided into four parts, including pre-training data, natural language base model, conversation language model and OpenLMLab

Dr. Sun Tianxiang

Lecture 4: "Training and Optimization Methods of Large Language Models" Wang Bingning

On the afternoon of August 19th, Teacher Wang Bingning, a very excellent teacher, brought the fourth lesson of this workshop-"Training and Optimization Methods of Large Language Models". He first systematically reviewed the development history of language models, and pointed out that the development of language models in recent years has been super linear, and the frequency of releases has been accelerating. In the following lectures, he elaborated on the basic framework and optimization methods of large language models, and at the end he optimistically expressed his vision for large language models and his determination to contribute to the Chinese Internet large language model.

Teacher Wang Bingning is a very good teacher

The content that needs to be rewritten is: Lecture 5: "TableGPT - A Preliminary Study on the Long Road to the Implementation of Large Models" Zhao Junbo

On the morning of August 20, researcher Zhao Junbo brought the fifth lesson - "TableGPT - A preliminary study on the long road to the implementation of large models". Before giving the lecture, he humorously explained why the word "overcoming tribulation" was added to the lecture title, because the relevant research process is very long and full of various setbacks and challenges. He hopes that through this topic, he can inspire the audience to also face the same research challenges. Bottleneck researchers. Afterwards, researcher Zhao Junbo briefly introduced his team members and research background, and explained why they chose to do rigid LLM and finally chose "TableGPT" and a series of other issues. During the following teaching process, Researcher Zhao introduced the functions, implementation process and solutions of "TableGPT", and concluded humorously at the end that going from a general LLM to a domain LLM is like the transition from a liberal arts student to a science student. , comparable to "overcoming the tribulation"

Researcher Zhao Junbo

The content that needs to be rewritten is: Lecture 6: "Dharma Academy Tongyi mPLUG Multi-modal Large Model Technology and Application" Yan Ming

On the afternoon of August 20, Teacher Yan Ming brought the last lesson of this workshop, "Dharma Academy Tongyi mPLUG Multi-modal Large Model Technology and Application". He first made a brief summary of the development history of multi-modal large model technology, and then focused on the two paths of his team's multi-modal research. One is mPLUG: the modular multi-modal pre-training large model technology series, and the other One is the Agent system for collaboration between large and small models. Finally, Teacher Yan Ming gave a detailed explanation and elaboration on topics related to the practical application of "ModelScope".

Teacher Yan Ming

At this point, the 6 courses of the 3rd Artificial Intelligence Innovation Technology Workshop "Multimodal Large Model Technology Frontiers and Applications" have ended. Professor Qiu Xipeng gave an online summary of the workshop. He reviewed the wonderful teaching content of six experts and scholars, and pointed out that from academic introduction to application, the cutting-edge results of these studies were refreshing to the audience. In addition, Professor Qiu Xipeng also said that the development of multi-modal large models is in its infancy and will be full of opportunities and challenges in the future development process. This workshop has built an interactive platform for researchers. A platform that helps scholars explore, learn and make progress together, both in academic research and practical applications. (China Artificial Intelligence Society)

Professor Qiu Xipeng

The above is the detailed content of Big names gathered to inspire the future│The third phase of the Artificial Intelligence Innovation Technology Workshop concluded perfectly. For more information, please follow other related articles on the PHP Chinese website!