On May 15th, at the Renaissance Hotel of Dongsheng Science and Technology Park, Zhongguancun, Beijing, Zhang Yijia had a peak conversation with Zhu Xiaohu, Fu Sheng, Li Zhifei, Wang Yu and Zhang Jianzhong.
Today, the speed of AI development is suffocating: Sora, which made people think that the world's model was born, is no longer the top of the industry's discussion after only one month; Robin Li said that the open source model will fall behind more and more, and three days later, Meta released the "strongest open source model" Llama 3.
I can't help but think: in the current era of AI, does the truth exist? How long is the shelf life of the truth?
There are many more specific questions:
- Has the AI technology paradigm converged, or is it still bifurcating? For example, Scaling Law and Transformer, does the Iron Throne exist?
- How to balance the "moon exploration" of AI technology and the "money scene" of business?
- Across cycles, from the Internet to AI 1.0 to AI 2.0, what has happened to the business model?
- In the current AI ecosystem, what are the different positions of "car makers" and "road builders"?
- Is the AI infrastructure of China and the United States on the same path?
- Who is stronger, open source or closed source?
- Is going public a starting point, a turning point, or simply replacing it with dividends?
……
One question is unanswered, and another one comes one after another. Moving forward with questions, every AI practitioner has their own pride and rationality in their hearts.
Identity determines position, faith determines action, and the AI exploration journey is a blind man touching an elephant, and "Jiazi Lightyear" wants to make this elephant puzzle more complete.
To this end, Zhang Yijia, founder and CEO of Jiazi Lightyear, invited five heavyweight guests in the AI industry, Zhu Xiaohu, Fu Sheng, Wang Yu, Li Zhifei, and Zhang Jianzhong, to discuss these important but perhaps unanswered questions at Dongsheng Science and Technology Park in Zhongguancun, Beijing, on May 15.
The reason why these five guests are invited as the protagonists of the summit forum is because they represent five distinct and extreme roles leading to the era of China's AI creation, and the permutations and combinations between them are reflecting the most concerned topics in the AI industry today. Like what:
- Li Zhifei vs Fu Sheng - a tortuous transformation and many final listings, a good old but new start; One is a native entrepreneur of the AI cycle, and the other is a serial entrepreneur who has crossed the cycle from the Internet to AI;
- Zhu Xiaohu vs Fu Sheng - Differences and reconciliation are the epitome of rapid iteration in the AI era;
- Wang Yu vs Zhang Jianzhong - both are infrastructure builders, one is a typical Tsinghua school, and the other has a background as a global executive at NVIDIA, and they are very suitable to answer the question: will China and the United States go the same way in terms of AI infrastructure;
- Of course, there are other permutations and combinations: the two-way rush between academics and industry; car building (application layer) and road building (infra layer); Those who invest money, burn money, and make money......
Investors VS Entrepreneurs: Investing money, burning money, making money
As the only investor among the guests, Zhu Xiaohu, managing partner of GSR Venture Capital, has always taken a clear stand.
Zhu Xiaohu is regarded as a typical "market believer". It's not that he doesn't believe in AGI, but he believes in AGI that can be commercialized. In his view, because of the existence of open source large models, the technical barriers of AI are not as high as imagined, and many projects are easy to fall into homogeneous competition, so it is more necessary to pay attention to the implementation of projects in a specific industry and specific scenarios.
"It doesn't make much sense for China to make hundreds of large models, but there is a lot of innovation at the application level. China far surpasses the United States in terms of data and application scenarios. ”
I don't know what sharp points the fast-talking Zhu Xiaohu will throw out for us on the day of the conference?
Entrepreneur VS Entrepreneur: Restarting and Continuous Transformation
Since joining 360 in 2005 to create 360 Security Guard, a popular product in the PC era, Fu Sheng has been known as the first generation of product managers in China, and has witnessed China's transformation from PC Internet to mobile Internet all the way. With the rise of AI, in 2016, Fu Sheng restarted to create Orion Starry Sky and focused on the development of artificial intelligence.
In Fu Sheng's view, the big model is not just a special session for large companies. Large model technology has given rise to many new entrepreneurial opportunities, especially large models for different applications and verticals built on top of large models.
Many of Fu Sheng's views don't seem so mainstream in the spotlight of the AI industry. In a conversation with Miss A a while ago, he expressed that when everyone thinks that large models are a big deal, it is "a red flag" for entrepreneurs; He believes that bipedal humanoid robots "don't want to land in five years"; He also believes that "engaging in AI can no longer burn money like this".
As an entrepreneur spanning the Internet, mobile Internet and AI eras, Fu Sheng has experienced many highlight moments and a period of serious self-confusion. In this wave of vigorous large models, he gave up "heroism" and chose a seemingly conservative path, not to roll up the 100 billion large model, but to release the Orion Star 10 billion model in January this year, providing customers with the choice of privatization large models.
Compared with Fu Sheng, Li Zhifei's questioning process is also tortuous. In 2012, Li Zhifei left Google to found Mobvoi. After the voice interaction software period in the mobile Internet era, the intelligent hardware period in the AIoT era, and in 2020, he opened a new era of large-scale model commercialization with the industry's first AIGC product "Moyin Workshop", Li Zhifei has embarked on a continuous transformation path for Chinese entrepreneurs, and finally completed the "small goal" of listing in the near future.
Nowadays, Mobvoi has become a company that "combines production and model", that is, it makes both products and models. In past conversations, Li Zhifei pointed out that "the difference between OpenAI is that after GPT-3, it has become very product-driven", while "Google, as a public company, has an unbridgeable gap from algorithm prototype to product launch".
Li Zhifei judged that 2024 will be the year of the explosion of AI application products. In the era of large models, the organizational structure of enterprises and the production relations of society will inevitably develop in the direction of the highest overall efficiency.
Fu Sheng, Chairman and CEO of Cheetah Mobile and Chairman of Orion Star, and Li Zhifei, Founder and CEO of Mobvoi, what kind of deep understanding will these two "entrepreneurial old boys" contribute to us?
China VS United States: Will the construction of AIInfra between China and the United States "step into the same river"?
Compared with the United States, China's AI ecosystem has a gap in terms of capital, computing power, and technology. To promote the closed-loop combination of production, education and research, Wang Yu practiced it.
Wang Yu is a professor and head of the Department of Electronic Engineering at Tsinghua University, a distinguished young person from the National Natural Science Foundation of China, and an IEEE Fellow. In 2016, Wang Yu participated in the establishment of Shenjian Technology, an AI chip company, and quickly became one of the leading AI chip start-up companies in China.
Today, Wuwen Xinqiong, established by Wang Yu as the initiator, is solving the problem of efficient and unified deployment of large model algorithms to large computing power chips, building an algorithm-hardware joint optimization platform, and improving the deployment efficiency of large models to chip platforms.
Zhang Jianzhong, the founder of Moore Threads, also plays an important role in promoting the development of AI infrastructure. In October 2020, Zhang Jianzhong established Moore Threads to provide customers with core infrastructure and one-stop solutions to accelerate computing power.
Zhang Jianzhong's resume also has the name of Nvidia. In 2006, Zhang Jianzhong joined NVIDIA as global vice president and general manager of China. NVIDIA is an important foundation for today's AI infrastructure, and can even influence the development trend of today's AI large models.
On the one hand, the academic entrepreneurship of Chinese universities, and on the other hand, the entrepreneurship of executives of American technology giants, the dialogue between Wang Yu and Zhang Jianzhong will definitely bring us a fierce collision.
Of course, there are countless similarities and differences in these 5 guests that can be arranged and combined. It is necessary to sell it here.
On the morning of May 15th, the Renaissance Beijing Zhongguancun Dongsheng Science and Technology Park Hotel, you must not miss this wonderful summit dialogue.