Let's go to the era of AI creation.
On May 15, 2024, the "AI Creation Era - 2024 Jiazi Gravity X Technology Industry New Trend" conference, hosted by China's science and technology industry think tank "Jiazi Lightyear" and co-organized by Zhongguancun Dongsheng Science City, was held in Beijing.
Zhang Yijia, founder and CEO of Jiazi Lightyear, released the theme report "The Era of AI Creation, 30 Judgments on the New Trend of AI in China in 2024".
In the summit forum session of "New Paradigm and New Inflection Point of Chinese Artificial Intelligence", Zhang Yijia discussed the development trend of artificial intelligence with Zhu Xiaohu, managing partner of GSR Venture Capital Fund, Fu Sheng, chairman and CEO of Cheetah Mobile and chairman of Orion Star, Li Zhifei, founder and CEO of Mobvoi, professor and head of the Department of Electronic Engineering of Tsinghua University, outstanding youth of the National Natural Science Foundation of China, IEEE Fellow and initiator of Wuwen Core Dome, and Zhang Jianzhong, founder and CEO of Moore Threads.
2024 will be a year of rapid development of generative AI, with 7 keynote speeches and 4 roundtable forums to discuss the current status and future of AI from different perspectives such as technology paradigm, infrastructure, and application implementation.
Finally, in the "Highlight Moment" session of the "Jiazi Gravity X" conference, "Jiazi Lightyear" released the [Star 100] 2024 Innovative Enterprise List on the spot, and these innovative enterprises and innovative practices in the field of AI will be the backbone of China's AI development.
According to statistics, more than 1,000 viewers attended the conference, and the conference was broadcast live through Jiazi Lightyear's WeChat video account, Baijia account, NetEase account and other channels, and was simultaneously broadcast by more than 20 media such as Sina News, Xinbang, City Boundary, and Investor.com, with more than 600,000 viewers on the whole network.
1.AI the era of creation, the new trend of AI in China
The theme report of this conference is "30 Judgments on the New Trend of AI in China in 2024 in the Era of AI Creation".
Zhang Yijia said that in the past year, we have experienced a compressed version of AI hype, and AI has entered the most intensive progress release period in history.
According to Zhang Yijia's analysis, there are two dimensions for AI to change the world. First, AI subjectivity: from human-led to AI-led, people's dominance has gradually been ceded, and AI has become subjective; Second, AI mapping power: AI's ability to map the physical world has gradually increased, and it has gradually achieved the catch-up and surpassing of human capabilities, from the brain, cerebellum to physical strength.
In March this year, Jiazi Lightyear Think Tank put forward the concept of "AI Creation Era" and proposed four stages for AI to change the world: AI production era, AI native era, AI creation era, and AI civilization era.
Zhang Yijia, founder and CEO of Jiazi Lightyear
At the scene, Zhang Yijia also interpreted these four stages. In the era of L1 AI production, AI means secondary productivity, and the key is to reduce costs and increase efficiency, and promote digital transformation, which is essentially a matter of efficiency and cost. In the L2 AI native era, AI means a second language, a new form of interaction and a content carrier. In the era of L3 AI, AI means the second subject other than humans, promoting the implementation of end-to-end intelligence, the combination of software and hardware, and the implementation of world models. In the era of L4 AI civilization, AI means the second civilization system.
In April last year, Jiazi Light-Year Think Tank proposed an evaluation index to evaluate the new generation of intelligence: "Letter-to-Energy Ratio", which reflects the amount of information that can be controlled by a unit of energy. Recently, Jiazi Light-Year Think Tank has further improved the evaluation system on the basis of the Xin-Energy Ratio, adding a new dimension to information and energy: behavior.
Energy, information and behaviour are three basic and interrelated concepts in modern society and the natural world – the progress of science and technology is a reflection of the increased transformative capacity between the three.
Zhang Yijia also summarized the wide application and impact of AI in all walks of life, including the transformation of the labor market, technological innovation, the use of data, the flow of capital and the management of land resources, discussed the commercialization path of AI, the controversy between open source and closed source, the development demand of computing power, the evolution of human-computer interaction and other key issues, and predicted the far-reaching impact of AI at the scientific, cultural, economic and social levels, including promoting the scientific revolution, triggering a new round of renaissance, and changing human cognition and psychological state.
In the summit forum session of "New Paradigm and New Inflection Point of Chinese Artificial Intelligence", Zhang Yijia and five guests, Zhu Xiaohu, Fu Sheng, Li Zhifei, Wang Yu and Zhang Jianzhong, had a wonderful discussion on topics such as technical beliefs and market beliefs, product PMF, Killer App, computing power ecology, and AI's first-mover advantage and late-mover advantage.
Zhu Xiaohu said that entrepreneurs are often full of romantic feelings and like to talk about infinite possibilities and imagination. Investors, on the other hand, are more pragmatic and focus more on the prospect of commercialization, "I have always emphasized that it is very important to achieve a commercial experience. ”
Fu Sheng said that he is a long-term technologist and a pragmatic short-term market. He also mentioned that innovation in science and technology can still unleash great potential, even with limited resources, and that excessive abundance of resources can sometimes lead to blind superstition about the potential of technology.
Li Zhifei believes that technical belief or market belief depends entirely on the position, providing more of an emotional value than absolute right or wrong. He said that it may be more important for technology entrepreneurs to be more pragmatic, pay more attention to business and market competition, and truly understand whether the technology meets the needs of users.
Wang Yu said that technology development and commercialization are closely linked, and pointed out that there are deficiencies in the closed loop of scientific research achievements from universities to industry, and it is necessary to strengthen contact and interaction. He believes that professors and researchers should understand market needs earlier and translate theoretical research into practical technical solutions.
Zhang Jianzhong emphasized the importance of industry applications in technology development, and believed that technological innovation needs to be closely integrated with the industry to achieve real commercialization, especially in the field of medical and health care, and the application of AI has broad prospects. Zhang Jianzhong also mentioned the importance of computing power, emphasizing the mission of Moore Threads in providing cost-effective computing power to promote the popularization of AI technology. At the same time, he also pointed out the changes in the current AI R&D model and the key role of GPUs in AI development.
2. Seven speeches to understand AI trends
Fang Han, chairman and CEO of Kunlun Wanwei, shared the theme of "Tiangong Model Drives New Changes in Application in the AI Era".
Fang Han elaborated on the company's latest progress and future vision in the field of AI. Since 2020, Kunlun Wanwei has devoted itself to the research and development of Chinese pre-trained large models, and in 2022, it open-sourced China's first 1.3 billion parameter Chinese model. The launch of Tiangong 1.0 in 2023 and its subsequent versions, especially the 400 billion parameter Tiangong 3.0 released in 2024, marks the company's significant progress in the field of AI technology.
Fang Han, chairman and CEO of Kunlun Wanwei
Fang Han emphasized that technological dividends and product innovation are the key to attracting and stabilizing users, and Kunlun Wanwei is committed to improving user work efficiency and promoting personalized expression through AI search and AI-UGC platform. In addition, he mentioned the achievements of the music grand model, which is able to compose music based on random content and has a global leadership in dialect processing and natural vocal generation.
Fang Han believes that despite the challenges of insufficient data volume in vertical fields such as music and 3D models, Kunlun is confident that it can sustain the development of the technology and promote the commercialization of AI technology through innovative methods such as synthetic data. He finally expressed Kunlun's mission to achieve artificial general intelligence, so that everyone can better shape and express themselves.
In the afternoon session of the conference, Gong Ruinan, Chief Operating Officer of Zhongguancun Dongsheng Science and Technology Park, delivered a keynote speech on "Science and Technology for Good, People-oriented", in which he began by quoting the poem "New Jiazi with the Fortune of the Calendar, Goodbye to the Old Renchen in the Days", paying tribute to the organizer of the event "Jiazi Light Year", and introducing the Compo cycle theory, pointing out that we are in a period of great technological change, and artificial intelligence is expected to play a key role in the new economic cycle.
Gong Ruinan, Chief Operating Officer of Zhongguancun Dongsheng Science and Technology Park
From a historical perspective, Gong Ruinan emphasized the development process from language to text to information theory, and believed that the current AI model drive represents a significant reduction in the cost of knowledge acquisition and use, and a new cognitive era has arrived, behind which are huge business opportunities.
As a result, he led to the second line of the poem, "This situation can be recalled, but it was already lost at that time", in order to describe the mental journey of the entrepreneur. Gong Ruinan discussed the different stages of business growth, including early start-up, market validation, scale to capitalization, and emphasized the experience and advantages of Dongsheng Science Park in serving enterprises, reducing the cost of trial and error, and improving the success rate.
The third line of the poem, "Go to the poor place of water, sit and watch the clouds rise", was used by Gong Ruinan to express his insistence on facing the uncertainty in the field of AI, emphasizing that technology should serve human happiness, rather than leading to human alienation, and proposed that technology companies should take science and technology for good and people-oriented as the core values.
Gong Ruinan revealed in his speech that Dongsheng Science and Technology Park is building a next-generation service form, aiming to build an intelligent ecology and promote value integration and exchange, so as to improve the probability of enterprise success and invite AI companies to create such a value community.
Lu Zhang, a well-known investor in Silicon Valley, founding partner of Fusion Fund, and visiting lecturer at Stanford University, delivered a keynote speech "Cost efficiency of AI, the future of infrastructure" in Silicon Valley, which deeply discussed the key role of AI infrastructure and its importance in promoting technological innovation and industrial integration.
Lu Zhang, Founding Partner of Fusion Fund and Visiting Lecturer at Stanford University
Zhang Lu mentioned that we are at an exciting time, especially in Silicon Valley, where the development of the AI ecosystem is booming. In this process, innovation in AI infrastructure has become particularly critical, including hardware chips, cloud computing, data processing, cybersecurity, and edge computing.
The commercialization and industrial application of AI technology need to be achieved by reducing costs, especially in terms of computing power and energy consumption. Zhang Lu listed the four major challenges facing the current AI infrastructure, and proposed a series of technological innovation directions, such as GPU optimization, new computing architectures, energy consumption reduction technologies, and data privacy and security solutions.
Zhang Lu also mentioned the value of open source ecology in AI development, as well as the advantages of edge computing in reducing energy consumption and improving security. She encouraged entrepreneurs and investors to have a clear understanding when choosing partners and determining competitive strategies, and emphasized the importance of high-quality data in AI applications. Zhang Lu firmly believes that through continuous technological innovation and international cooperation, AI will bring more surprises and value to mankind.
Li Zhifei, founder and CEO of Mobvoi, delivered a keynote speech "How can AIGC products that combine production and model play with?" 》。
Mobvoi founder and CEO Li Zhifei
In his speech, Li Zhifei deeply discussed the importance of the combination of products and models in the field of artificial intelligence, emphasizing that in the context of the current rapid development of technology, close collaboration between products and models is crucial. He believes that although large model technology brings great potential, there is also a disconnect between the product and the model, which requires companies to clearly define the product goals and the technical specifications of the model.
Li Zhifei proposed that having an independent model can provide strong iterative ability and market competitiveness for products, and at the same time, high-quality data based on products is the key to building lasting competitive barriers.
As the helmsman of the newly listed "AIGC No. 1 Stock", Li Zhifei pointed out that the company must have a clear product goal to guide the direction of R&D, rather than conducting aimless R&D in the field of AI. In addition, he shared his exploration of business models, especially new markets for professional consumers, and shared several of the company's products, including the AI dubbing platform Moyin Workshop, the three-second clone vocal DupDub, and more.
Finally, he emphasized the importance of finding and sticking to the company's core values in the rapidly changing AI industry, that is, truly finding the "soul of the company", and encouraged community communication and feedback to promote innovation and progress.
Next, Yang Fan, co-founder of SenseTime, delivered a keynote speech on "Construction and Application of Artificial Intelligence Infrastructure under the Wave of Large Models".
Yang Fan, co-founder of SenseTime
First of all, Yang Fan summarized two key technical characteristics of the large model era: the first is AGI, "if we treat both humans and AI as information systems, then today's new generation of AI information systems can provide feedback more actively"; The second is that the versatility of AI has been significantly improved, "the cost has decreased and the cost performance has improved, which will support the sustainable and healthy long-term development of the AI industry and improve the cost performance problem."
Secondly, Yang Fan raised the importance of AI infrastructure. He believes that the three elements of AI: computing power, algorithms, and data infrastructure, are the inevitable trend of industrial development. In terms of application scenarios, Yang Fan mentioned that the medical field is a hot spot for AI applications, and programming ability is the core competitiveness of AI, which is not only about rational thinking, but also the key to promoting the closed-loop action.
In the application landing session, Chris Pereira, founder and CEO of Meixun, delivered a speech on the theme of "Empowering Brand Globalization: The AI Wave of Chinese Enterprises' Overseas Business Expansion".
美讯创始人兼CEO Chris Pereira
Chris Pereira shared: "I personally don't like the word 'go to sea', the moment you buy your ticket, you're already out on the sea. From China to overseas, localization is the most important process, and the process of localization is to build trust and make friends with local people. ”
In his view, content localization, team localization and third-party recognition are the three major levers for Chinese enterprises to localize overseas.
Speaking about the importance of building local trust and relationships, Chris Pereira mentioned that important connections and trust can be built through social activities such as coffee with locals. He quoted Huawei's Ren Zhengfei quote that "a cup of coffee absorbs the energy of the universe" to emphasize that even informal social events can bring great business value and avoid detours, thereby building trust and relationships in overseas markets.
Zhang Audi, deputy general manager of eclickworld technology center, delivered a keynote speech entitled "Demand production capacity from AI, make decisions from data, and build a new ecology of overseas marketing".
Zhang Audi, deputy general manager of Eclickworld Technology Center
Zhang Audi said that localization, data growth and IT infrastructure are the main problems that domestic marketing may face when going overseas.
On the whole, "asking for production capacity from AI and making decisions from BI" is the core strategy of eclickworld to use AI and data analysis (BI) to improve the effect of overseas marketing. He explained that by combining AI models and industry experience, an intelligent decision-making chain can be built to improve the accuracy and efficiency of marketing campaigns. At the same time, BI is used to make more accurate business decisions to achieve a more effective overseas marketing strategy.
3. Four roundtables to dismantle the AI ecosystem
The roundtable forum of this conference discussed the development trend of AI from different layers of AI ecosystem.
In the roundtable session of "AI Technology Paradigm Change: What is the End of Scaling Law", Hou Zhenyu, Vice President of Baidu Group, Lu Zhiwu, Professor of Hillhouse School of Artificial Intelligence of Chinese People's University, Zhang Yu, Founding Partner of Qingzhi Capital, Zhang Junlin, Head of New Technology R&D of Sina Weibo, Luo Xuan, COO of RWKV, and Wang Yi, Senior Analyst of Jiazi Lightyear, discussed the development trend of Scaling Law.
Hou Zhenyu is a staunch supporter of Scaling Law, believing that Scaling Law is still valid and has great potential, at least for the time being. He also mentioned Moore's Law and Andy Beer's Law, arguing that Scaling Law has a similar effect in the field of AI, where the growth of computing power is consumed by the demand for algorithms and data.
Lu Zhiwu expressed conditional support for Scaling Law, believing that it may not be sufficient in achieving AGI. He suggested that a better model architecture might be needed, noting that the Transformer model might not be the final solution. Lu also highlighted that there are more factors to consider for commercialization success, not just the size of the model.
Zhang Yu believes that Scaling Law is effective in the short term, but it may reach its limits in the long run. He illustrates the limitations of Scaling Law through metaphors, such as the inability to climb trees to reach the moon, suggesting that new methods or technological breakthroughs may be needed.
Zhang Junlin sees Scaling Law as an empirical formula, which is based on a large number of experiments and data. He feels that at this stage, Scaling Law is valid, but may see its effects slow down over time.
Luo Xuan believes in Scaling Law very much, but points out that the time and space complexity of the Transformer architecture leads to low computing power and data utilization. He believes that there will be new algorithms to improve the efficiency of data and computing power in the future, and mentioned that the architecture of RWKV may represent this direction of development.
At the roundtable forum of "AI Creates Everything, Who Will Feed AI", Yang Zhen, Head of Strategy and Marketing of Nortel Digital Intelligence, Shen Ou, Vice President of Qingyun Technology, Liang Shuang, Vice President of Beijing Luchen Technology Co., Ltd., Yu Yi, COO of Shansi Enlightenment, and Zhang Yingjie, Managing Director of Dachen Caizhi, had a wonderful discussion.
Zhang Yingjie first mentioned that AI computing infrastructure is facing the challenge of adapting to rapid technological change. Yang Zhen emphasized the difference between AI infrastructure and the IT era, and proposed a redefinition of hardware, software, and field levels. He introduced the construction of the Beijing Digital Economy Computing Center, emphasizing its role as a public computing power provider, as well as the comprehensive layout of the computing power layer, scheduling layer and industry empowerment layer, and the importance of providing a framework and tool chain.
Shen Ou shared Qingyun's experience in the field of cloud computing, emphasizing the reliability and performance requirements of the computing platform. He discussed the importance of domestic computing power adaptation and the ecological construction of the platform, and emphasized the challenges of operating the computing power platform, including cost control and the provision of technical services.
Liang Shuang mentioned the diversification and flexibility of hardware requirements for large models, as well as the limitations of GPUs in large model training. She introduced Luchen Technology's Colossal-AI system, emphasizing its advantages in distributed training. Yu Yi pointed out the importance of forward-looking deployment, customized configuration, and paying attention to advanced technology trends, as well as forward-looking planning for the construction of the computer room.
Next, in "The "Money Scene" of AI: How to Form a Business Closed Loop? 》In the roundtable forum, many people, Vice President of Xiaomi Mobile Phone Department, Vice Chairman of Xiaomi Technical Committee, General Manager of Xiaomi Robot Division, Xiong Youjun, General Manager of Beijing Humanoid Robot Innovation Center, Duan Xiaolei, General Manager of Global Startup Ecosystem Development of Huawei Cloud, Liu Dexin, President of aiXcoder, and Liu Maoliang, General Manager of Xinzhi Juan, had a wonderful discussion. Zhang La, co-founder of CSSTV and senior engineer of the Institute of Automation, Chinese Academy of Sciences, served as the moderator of the roundtable.
Many emphasized the importance of survival, pointing out that companies need to have a clear business positioning and focus on sustaining costs, as the development of the AI industry is a long-term process. Duan Xiaolei also expressed a similar opinion: "Survive, after all, we are a company. Only by saying that we can survive can we bring greater influence to the world when we are stronger in the future. ”
In the long run, how can companies survive and live better and better? Liu Dexin, president of aiXcoder, believes that even if there is no technical problem, we should pay more attention to the actual needs of customers and polish the product to better meet these needs.
"The industry has pain points, the industry has demand, AI is growing, and AI is iterating. Now the next step is how to integrate AI with the entire industry. Liu Maoliang, general manager of Xinzhi Juan, said.
Regarding the future of AI, Xiong Youjun, general manager of Beijing Humanoid Robot Innovation Center, believes that embodied intelligence is the future of AI, and humanoid robots, as representatives of embodied intelligence, will be the key to future development.
In the roundtable forum of "Towards the Era of AI Creation: Two Sides of the Coin", Wang Tianmiao, Honorary Director of Beihang Robotics Research Institute and President of Zhongguancun Zhiyou Research Institute, Liu Daoquan, Founder and CEO of Al wisemodel and Deputy Secretary-General of the AI Big Data Special Committee of Tsinghua Alumni Association, Liu Changwei, Chairman of Yixin Technology, Wu Bingjian, Partner of Heart Capital, Li Jiaying, Co-founder and Product Leader of RPGGO, and Li Genji, CEO of Whale Sea Shellfish and head blogger of artificial intelligence, had a "debate on the future" 。
When talking about "whether embodied intelligence is the ultimate form of AI", all four guests gave a "no" answer.
Wang Tianmiao believes that there are many ultimate forms of AI, including the spiritual level, the interactive level and the physical operation level. In Liu Daoquan's view, embodied intelligence is a form, but the ultimate form of AI may be in the cloud, and the edge may only have simple perception or processing capabilities.
Liu Changwei's point of view is that the logical generation of embodied intelligence requires the summary of the laws of large models, and there may be better models in the future. Wu Bingjian said that the ultimate form of AI may be composed of multiple models, and embodied intelligence is just one of them. From the perspective of needs, Li Jiaying believes that the form of embodied intelligence depends on the way to meet human needs, and does not necessarily have a specific physical form.
Lee Gengi was the only guest who voted "yes". In his view, "embodied intelligence is an important direction for the development of AI, and it may be one of the ultimate forms." He emphasized the need to pay attention to the complexity of the brain structure during human evolution, and the impact that this structure may have on the future development of AI.
In this roundtable forum, guests from different fields shared their professional insights on the development of AI, providing a comprehensive and diverse perspective on the intersection of AI and society, law, ethics and philosophy.
For example, in the discussion of "Will AI synthetic data be a lifesaver for model training", Wang Tianmiao believes that AI synthetic data is very effective in training underlying intelligent behaviors, especially in the field of embodied intelligence, and the operation ability of robots can be improved through data generation training. Liu Daoquan, on the other hand, holds the opposite view. His AI-generated data is essentially the same as human-generated data, both are the source of the data.
The roundtable forum of this conference is not only an elaboration of technology and products, but also a discussion of practical development issues, and an outlook on the broad prospects of AI. Just like Leonardo da Vinci's endless curiosity, in this fast-growing field of AI, maintaining an inquisitive heart and asking insightful questions may be the golden key to humanity's future.
Finally, "Jiazi Lightyear" also released the [Star 100] 2024 Innovative Enterprise List at the conference. Through the release of this list, "Jiazi Lightyear" hopes to recognize the outstanding contributions made by relevant enterprises in technology research and development, business models, application scenario innovation and ecological construction, and inject more vitality and impetus into the future of China's AI industry.
In the long river of technological development, mankind is standing at a special historical intersection, and "Jiazi Lightyear" looks forward to exploring the infinite possibilities of the era of AI creation with everyone. As Zhang Yijia said in the interpretation of the report: "In the era of AI creation, we pay attention to how the technological paradigm converges, how the technological leap forward reshapes thousands of industries, and pays more attention to the far-reaching impact of every decision on the future social and economic structure." ”