CBN Friday Special丨WAIC 2023 gets underway in Shanghai, spotlighting homegrown AIGC

The World Artificial Intelligence Conference (WAIC) 2023 kicked off on Thursday at the Shanghai Expo Center.

The three-day event is being attended by top scientists, entrepreneurs, government officials, and industry experts.

Themed "Intelligent Connectivity, Generating Future", the conference comprises various forums and discussions about core technologies, intelligent terminals, application empowerment, and cutting-edge technology.  

(相关资料图)

A record number of companies are attending the event, which is in its sixth edition this year. Over 400 exhibitors are showcasing a wide variety of products related to scientific intelligence, robotics, meta-universe, smart chips and autonomous driving in a 50,000-square-meter exhibition area. Both figures are at their highest levels since 2018, when the annual event was inaugurated.

Considered the most high-profile artificial intelligence event in China, the WAIC attracted over 600,000 offline visitors and witnessed the signing of 200 major industrial projects with a combined investment of 70 billion yuan over the last five editions.

"In the process of promoting high-quality development, creating high-quality life, and achieving efficient governance, we need to leverage the power of AI to empower all industries more than ever before. Shanghai will make greater efforts to promote the healthy development of AI in the future," said Chen Jining, Shanghai's Party chief, during the opening ceremony on Thursday morning.

China's AI industry is in the midst of a development boom, with the scale of the core industry at 500 billion yuan and the number of AI enterprises exceeding 4,300, Vice Minister of Industry and Information Technology Xu Xiaolan said at the event's opening ceremony.

China has been accelerating its construction of AI-related infrastructure, with its total computing power ranking second globally and its number of 5G stations exceeding 2.8 million, Xu said.

According to 2022 Global AI Innovation Index Report released at the WAIC, China ranks second in Global AI Innovation Index for the third straight year in 2022, as it has been accelerating efforts in expanding its talent pool, building data infrastructure and making technological breakthroughs.

While adhering to self-reliant innovation, China has been actively promoting scientific and technological sharing and exchanges, aiming to contribute to global sci-tech development for the benefit of all mankind, as a number of high-profile US multinationals including Google and Tesla also showcased their AI-powered products and services at the WAIC.

In a video speech delivered to the opening ceremony of the event, Tesla CEO Elon Musk lauded China's advance in the field. Musk said artificial intelligence will play a profound role in the future of human evolution and civilization and predicted that fully autonomous driving could be achieved later this year. He also expects China to have very strong AI capabilities in the future.

AIGC being the centerpiece  

Large-language models and generative AI products are the undisputed highlights of this year’s WAIC, which includes a specially allocated area for the booths for more than 30 LLMs. 

Chinese internet giants Alibaba Group Holding and Baidu, as well as AI startups such as Sense Time and iFlytek are just some of the LLM developers showcasing their latest products in the exhibition.

NetEase Youdao unveiled an interactive English language learning chatbot at the event, which allows speakers to practice spoken English with the AI model. 

Chinese tech firm Meituan also launched the 4.0 version of its unmanned aircraft system. Compared with the previous editions, it has longer delivery range and better performance operating in inclement weather environment.

In terms of semiconductors, a number of domestic chip companies including Enflame, Vastai Technologies and Hexaflake unveiled their cutting-edge AI chips, which enable LLMs to process vast amounts of data at lightning speed.

Fourier Intelligence, a Chinese tech startup specializing in rehabilitation robotics and artificial intelligence, unveiled its first-generation humanoid robot at the event.

Standing 1.65 meters tall and weighing 55 kg, the robot, named GR-1, is able to walk at 5 km per hour and carry a load of up to 50 kg, the company said. With human-like motion control, it can walk bipedally, avoid obstacles, ascend and descend a slope with stability, and withstand shocks.

With the help of cognitive intelligence provided by ChatGPT-like tools, the robot is capable of interacting with humans while performing a variety of movements and tasks. General-purpose humanoid robots can be used in an array of scenarios including industrial production, rehabilitative care, scientific research and household service.

Meanwhile, several AIGC industry alliances were founded in Shanghai during the opening session, such as the Corpus Data Alliance for Foundation Model and the Large Model Assessment and Collaborative Innovation Center.

AIGC features like ChatGPT are initially being used in office work and collaboration and would be used further in next-generation robots, which would be more powerful and intelligent, Yao Qizhi, winner of the Turing Award and dean of the Shanghai Qi Zhi Institute, told the WAIC's opening forum.

China’s AI race intensifies

ChatGPT is currently making waves around the world with its user base crossing 100 million just two months after its launch.

Over the past two months alone, several Chinese tech companies have announced significant progress in the development of LLMs. 

On June 28, ByteDance's Volcano Engine has announced the launch of Volcano Ark, an AI model service platform.

On June 9, iFlytek released version 1.5 new AI-driven cognition model “Spark” as the company aims to exceed ChatGPT's capabilities in Chinese language, reaching the same level as the English chatbot possibly in the fourth quarter this year.

During the Huawei Developer Conference Cloud 2023 today, the tech giant launched the much-anticipated Pangu Large Scale model 3.0. Huawei Executive Director and Huawei Cloud CEO Zhang Pingan said the new model focuses on value scenarios and is committed to being robust for government affairs, finance, manufacturing, coal mines, railways, pharmaceuticals, meteorology, and other industries.

Chinese organizations have launched over 80 LLMs as of July, according to Zhang.

The leading LLMs in China have shown impressive performance, comparable to GPT3.5, industry insiders said. Among them, Baidu's ERNIE bot has surpassed GPT3.5 and ranks first among domestic large models, according to a test made by China Academy of Industrial Internet (CAII).

In the automotive industry, Changan Automobile is leveraging Baidu's ERNIE bot to develop generative AI products, in order to enhance user experience for mass-produced vehicles. 

360 Security Group is also adapting its LLM into various scenarios. In smart city application, a LLM developed by 360 surpassed ChatGPT in the national certified tax agents examination, the company said on Wednesday.

"Large language models will become standard in every digital system, deployable in smartphones and cars, and ubiquitous,"Zhou Hongyi, the founder of 360 Security Group said at Global Digital Economy Conference in Beijing on July 2.

According to a report by iResearch, the market size of China's AI industry is estimated to reach 612.2 billion yuan by 2027, with a compound annual growth rate of 25.6 percent from 2022 to 2027.

While the US has consistently held the position of global leader in terms of the number of large models, China has entered a phase of rapid development since 2020 and is currently experiencing synchronized growth with the US.

However, there are still some gaps between domestic and foreign models in terms of precision and universality, particularly when compared to advanced models like GPT4.

To promote implementation of LLMs and empower digital innovation and transformation across industries, a number of Chinese cities have joined forces in this endeavor.

Shanghai will continue to focus on embodied intelligence research and applications, with a particular emphasis on machine vision, natural language processing, and training LLMs for robots.

Beijing city has also taken note of the evolving situation and plans to ramp up policy support and leadership in setting standards. Efforts will be made to support Beijing's AI enterprises and various industry organizations in playing a leading role in the formulation of broad industry standards, including model performance, data security and privacy protection, according to a recent document.

在AI热潮席卷全球之际，2023世界人工智能大会于7月6-8日在上海举办。

以“智联世界 生成未来”为主题本届大会聚焦科学前沿和产业发展，围绕技术、产业、人文三大话题，重点关注大模型、智能芯片、科学智能、机器人、类脑智能、元宇宙、自动驾驶、数据论坛、法治与安全、区块链等十大前沿风向。

值得注意的是，今年大会重磅回归线下，参展企业数量、展览面积均创历届之最。5万平方米世博主展览涵盖核心技术、智能终端、应用赋能、前沿技术四大板块，包括大模型、芯片、机器人、智能驾驶等领域，参展企业超400家，优秀初创企业超50家，首发首展新品达30余款。 

上海市委书记陈吉宁在开幕式上指出，放眼全球，人类社会已进入创新密集活跃的大科学时代。以大模型为代表的人工智能突飞猛进，正推动新一轮科技革命和产业变革加速演进。新征程上，上海要努力推动高质量发展、创造高品质生活、实现高效能治理，比以往任何时候都更加需要发挥人工智能赋能百业的“头雁效应”、拉动发展的“乘数效应”。

工业和信息化部副部长徐晓兰在开幕致辞时表示，在各界的共同努力下，我国人工智能产业蓬勃发展。徐晓兰表示，我国人工智能核心产业规模达到5000亿元，企业数量超过4300家，智能芯片开发框架、通用大模型等创新成果不断涌现。

开幕式上，特斯拉创始人兼首席执行官埃隆·马斯克以视频方式“现身”并发表讲话。马斯克说，中国一旦下定决定要做一件事情，就一定能做好，各个产业都是这样，包括人工智能，相信中国会有很强的人工智能能力。

马斯克表示，生成式人工智能的出现，对人类文明正产生着非常深刻的作用和影响，随着数字计算能力的爆炸式增长，机器计算能力和生物算力的比例正在不断扩大，这意味着，机器和生物的算力差距在进一步扩大，经过一段时间之后，人工的智能所占的比例会越来越低，这将是一个根本性的深度的变化。

中国科学院院士、清华大学交叉信息研究院院长姚期智表示，在ChatGPT之后，人工智能研究的下一个重要目标，是要让有多重感知能力的机器人，在不同的环境下能够自主学习新技能。

谈及大模型在垂直领域的发展，姚期智认为，最容易想到的场景就是文书工作，基于大模型的语言，更多的工作可以交由机器来做。他同时指出，当下业内的强化学习方法速度太慢，常常需要几个月的时间才能学好新技术，如果有强化学习的方法让机器人在几个小时内就能完成自主学习，将是重大的突破。

图灵奖得主、被誉为“深度学习三剑客”之一的Meta AI基础人工智能研究(FAIR)团队首席人工智能科学家杨立昆(Yann LeCun)直言，不喜欢AGI(通用人工智能)这个名词。他解释称，每一个智能都是专项的，人类智能也是非常专业、专项的，但LLM(大语言模型)只基于文本训练，不是利用视频训练，因此它们不理解物理世界，没办法规划、推理，就没办法实现目标。

对于如何实现AI的治理和规范，杨立昆认为可以通过可控的方法、设置安全护栏等使它们不会出现欺骗和支配人类的行为。他认为，严格监管人工智能发展并不能保证AI安全，从长远看，开源才能让AI保持安全、良善。未来每个人都能通过人工智能助手和数字世界沟通。如果人工智能只被少数人控制，这是不可以的。

“大模型”由虚向实

若论及此次世界人工智能大会最为“吸睛”的存在，大模型当仁不让。

这次展会上共有20余款机器人参展，并集中展示30余款通用或行业用大模型。

其中，商汤日日新、百度文心一言、科大讯飞星火认知大模型、华为云盘古大模型、阿里云通义千问、京东云言犀大模型、清华大学ChatGLM、复旦大学MOSS以及出门问问序列猴子、360智脑大模型、第四范式的式说大模型等10余家厂商研发的基础应用大模型将集体亮相。

在WAIC展台上，蚂蚁集团首次系统展示蚂蚁可信AI的技术架构，以及全新升级的可信AI工业检测平台——蚁鉴2.0。该平台可以通过智能对抗技术，生成海量测试集，对AIGC生成式模型进行交互诱导。可以理解成，它像一个24小时不眠不休的“安全黑客”在找茬大模型，从而找到大模型存在的弱点和安全问题所在。

商汤科技此次带来了“商汤如影SenseAvatar”AI数字人视频生成平台。该公司相关负责人对21世纪经济报道记者称，如影数字人生成方案只需一次真人素材录入，就能生成超写实的数字分身，实现视频制作效率的指数级提升。

目前，这一平台主要场景是帮助企业和个人轻松开展数字人短视频、直播等内容创作，同时也可为教育培训、企业宣传、娱乐文化等领域提供视频解决方案。 

ChatGPT主导的对话式AI，令普通人能轻易感知和理解人工智能，人和数据结合、交互比以往任何时候都更顺畅。但是，大模型的发展和落地也需要时间探索。任何技术，归根结底需要为产业和场景服务。

例如，科大讯飞、网易有道等科技公司的大模型落地，就率先瞄准了教育领域的场景。网易有道自研的大模型“子曰”最新应用成果也首次亮相。虚拟人口语教练跟用户可以互动，它不仅能够针对练习者的回答提供情感反馈，还可以主动抛出问题。

昨日，华为轮值董事长胡厚崑在WAIC称，2022年底ChatGPT的出现，将人工智能推向新风口，人工智能将帮助人类改写身边的一切。“走深向实”则是华为的重要战略。

胡厚崑透露，目前华为云生态已孵化了20多个基础的大模型，同时还适配了10多个业界主流的大模型。另外，华为云盘古大模型已经深入金融、制造、政务、电力、煤矿、医疗、铁路等10多个行业，支撑400多个业务场景的AI应用落地。

WAIC上，机器人同样十分亮眼。被誉为“制造业皇冠顶端的明珠”，机器人的研发、制造、应用是衡量一个国家科技创新和高端制造业水平的重要标志。在本届世界人工智能大会上，机器人展品琳琅满目。

比如，特斯拉就展示了明星机器人Optimus人形机器人。据了解，Optimus搭载了与特斯拉汽车同源的最新科技，包括完全自动驾驶能力（FSD）电脑以及Tesla Vision视觉神经网路。特斯拉人形机器人身高172公分，体重56.6公斤，与一个正常成年人的身形无异。机器人全身拥有28个自由度，就如同人体的关节。手部有11个自由度，可以更加灵活的抓握，且力大无比，可以单手举起一台钢琴。

本土科创公司傅利叶智能带来的人形机器人GR-1身高1.65米，体重55公斤，全身自由度达40个，最大关节模组峰值扭矩可达300NM，步行速度可达到5KM/h，负重50公斤。这款机器人有高度仿生的躯干构型、拟人的运动控制，具备快速行走、敏捷避障、稳健上下坡、抗冲击干扰等运动功能，结合认知智能，能够与人协同完成动作，在工业、康复、居家、科研等多应用场景潜能巨大。

“就像从台式电脑到个人电脑，再到手机的发展趋势一样，通用人形机器人未来会逐渐成为行业的主流。”傅利叶智能负责人表示，目前通用人形机器人硬件基础基本奠定，大模型推动的应用浪潮即将来临。“我们感受到通用人形机器人的商业化未来可期，因此其也抢抓先机，将发展战略逐渐切入到万亿级通用机器人领域。”

在业内人士看来，随着AI的快速演进，机器人变得更加“可支配”，未来落地生活场景的方方面面也具备了更多可能性。

“百模大战”风起云涌

由Open AI掀起的“百模大战”正在上演。

今天，在华为开发者大会2023（Cloud）上华为常务董事、华为云CEO张平安宣布，华为云盘古大模型3.0正式发布。张平安表示，盘古大模型3.0是一个完全面向行业的大模型系列，包括5+N+X三层架构，围绕“行业重塑，技术扎根，开放同飞”三大方向，持续打造自己的核心竞争力，将重塑千行百业。

自今年3月，基于文心大模型3.0打造的百度新一代知识增强大语言模型“文心一言”公开邀测之后，阿里通义千问、讯飞星火大模型、智谱AI的ChatGLM 等纷纷发布。此后，美团、百川智能、云知声、美图、腾讯……新加入大模型赛道的国内科技公司此起彼伏。

目前，中国紧跟潮流，已进入大模型快速发展期。从全球已经发布的大模型分布来看，中美两国数量合计占全球总数的超80%，美国在大模型数量方面居全球之首，中国目前则已进入大模型快速发展期，与美国保持同步增长态势。

截至7月3日，我国10亿级参数规模以上大模型已超80个，大模型产业将进入“百模大战”时代。

“我们把大模型及相关产品分为三类。第一类是通用大模型，如OpenAI的ChatGPT、百度的‘文心一言’、科大讯飞的‘星火’等都属于通用大模型。第二类是行业大模型。第三类是基于通用大模型或行业大模型的应用型服务。目前已经发布的产品大多集中在第一类和第二类。”有业内人士表示，“从参数量来看，‘文心一言’等通用大模型的参数量在1000亿级别，其他企业或初创公司大模型参数量通常在100亿、10亿级别。”

全国多地正在以大模型作为人工智能发展突破口，抢占行业和市场发展新高点。7月2日，北京发布全国首个政务服务大模型场景需求。共涉及市政务服务管理局、市中小企业服务中心等32个项目，涵盖智慧政务、智慧办公等14类场景。北京市政务服务大模型在国内外探索大模型助力政务管理实践方面具有极强的示范带动意义，在推动大模型赋能千行百业、加快推进大模型商业化落地应用等方面也极具吸引力。

Executive Editor: Sonia YU

Editor: LI Yanxia

Host: Stephanie LI

Writer: Stephanie LI 

Sound Editor: Stephanie LI

Graphic Designer: ZHENG Wenjing, LIAO Yuanni

Produced by 21st Century Business Herald Dept. of Overseas News.

Presented by SFC

编委:  于晓娜

策划、编辑：李艳霞 

播音：李莹亮

撰稿：李莹亮

音频制作：李莹亮

设计：郑文静、廖苑妮

21世纪经济报道海外部 制作

南方财经全媒体集团  出品

关键词：