(Original title: Kai-fu Lee, Chairman of Innovation Works: Every existing APP may be reconstructed by AI2.0, and the technology wave of related large models will change everyone’s life)
From May 25th to 30th, the 2023 Zhongguancun Forum will be held in Beijing.
On May 28, at the parallel forum of “Development of Artificial Intelligence Large Models” of this forum, Kai-fu Lee, Chairman and CEO of Sinovation Ventures, gave a keynote speech on “New Opportunities from AI1.0 to AI2.0”.
He said that the AI 2.0 era will subvert many industries, such as search engines, e-commerce, finance, education, film and television entertainment, medical care, games, etc. In the future, AI will enter the entity, and it will reconstruct human life and work. “AI large model is a historical opportunity that China cannot miss.”
During this parallel forum, Kai-fu Lee accepted media interviews including reporters from Daily Economic News.
The fastest realization in the era of AI2.0 is productive application
When asked which fields will be impacted first in the AI2.0 era, Kai-Fu Lee believes that the fastest implementation should be productive applications, such as Microsoft Office for Word, Excel, PowerPoint (PPT) or Photoshop applications. Their users will be replaced.
For example, he said, for example, the picture drawn by a graphic designer now may be completed very well by using AI to draw with text in the future. In the future, a poster may be ready only through verbal communication. In the past, it may have been done in a few days or even a week, but now it may be done in a few minutes or even a few seconds.
“So this will bring huge productivity improvements and cost reductions. It will also bring about great changes in work distribution and so on. Maybe graphic designers will gradually find that there are fewer and fewer job opportunities, but those who are very creative If a person can use his language to become a poster, then he can create greater value.”
Li Kaifu said that in the future, similar changes may occur in various fields such as media, medical care, economy, and entertainment. Therefore, similar productivity tools in these fields are fields that allow white-collar workers to generate content, and may be the first to be replaced. .
The second thing that will be replaced is certain tools or scenes of entertainment. For example, it is already possible to use language to form games, that is to say, what kind of game the user wants to play, he can not only help you make out the characters, props, scenarios, scripts, scenes, even the codes and dubbing in it.
“Of course, the games we make today are still very simple, but in the future, I think a group of players may discuss together, write a long manuscript and input it, and the game will be made, so the entire game production process will also change. ” Li Kaifu said that in terms of video production, for example, today’s Douyin uploads the UGC shot by many individuals, and at the same time distributes the content that users are most interested in, but in the future, AI can completely target users’ interests. Produce short videos .
“So every app that exists today may be refactored by AI2.0.” Li Kaifu said that some applications that cannot be done because the technology is not good enough, such as intelligent assistants and certain applications in the Metaverse, will be in the future because of Relevant technologies are becoming more and more mature, and products that cannot be made today can be made.
Therefore, Kai-fu Lee believes that the entire AI2.0 model will bring a new wave of technology, which should be many times larger than Windows and Android in the past. It can rewrite all imaginable applications, and then change the life of each of us. Life.
There are three different ways to invest in AI2.0
At present, AI technologies such as AIGC (generative AI) and large-scale models are relatively hot. Should investment pursue these hot spots? Will there be bubbles in related fields?
In this regard, Kai-Fu Lee said that investing in platform-based large-scale companies should be a huge opportunity, and their ceilings are very high, such as Microsoft and Google. Their growth opportunities are particularly large, and of course they are also risky. Today there are so many giants to participate in the big model, but each one faces the challenge of not necessarily being the first.
“So, every time a huge wave of technology comes, if you can catch the best company, it may bring huge returns, but if you make a wrong investment, you may lose everything. This is not too big because it is a big model or AI. Big difference.” He said that investing in platform companies is betting on the value of the platform, that is, high risk and high return.
In Lee Kai-fu’s view, there may be more opportunities for the application of large-scale investment models, and the risks are not so great. Because AI2.0 will bring subversive changes in almost every production field, which will bring opportunities.
However, he also mentioned that there are several key points to grasp when investing in AI2.0 applications. This type of enterprise must have a suitable compound cross-type founding team, not only must understand the technology but also understand the industry, the execution must be fast and strong, and it will continue to iterate according to changes in technology and the market. “Of course, from the perspective of application, if the platform is too powerful, it will be squeezed. Therefore, for pure application developers, they must be able to build their own moat early enough.”
Kai-fu Lee emphasized that don’t ignore investment in infrastructure, especially computing power. He mentioned that the computing power required by a search is about 1/100 of the computing power required by a GPT. If we think that every application will be refactored in the future, the usage of GPT must be many times that of search in the future. Assuming it is 10 times the search, the computing power it needs is a 1000 times increase. “The 1,000-fold increase in computing power cannot be supported by today’s Internet, servers, and clouds.”
In this context, various new algorithms, new chips, new network protocols, and new distributed system development may be produced, all of which are of great value. “If each of these technologies can reduce the cost of computing power by half, it will be a potential start-up company.” He said, so there are actually three different investment methods for investing in AI2.0 – one is to invest in the platform The second is to invest in application-oriented companies, and the third is to invest in companies in the infrastructure field, such as companies that can improve computing power or bring new algorithms and new chips. He emphasized that don’t think that you must invest in a platform-based model company.
In the current upsurge of large-scale models, which category can go to the end? In this regard, Kai-fu Lee said that there should be several large models in the end. The text mockup is not necessarily used in the same way as the multimodal mockup, so it may come from two companies, or it may come from one company.
In his opinion, AI2.0 is here to “turn the table”. It will take these existing strongest companies in the past, no matter in which field such as entertainment, games, e-commerce, short video, search, medical education, etc. , must make subversive things, so that the existing companies feel “I don’t dare to do it”.
Just as it was difficult for Kodak to accept the fact that digital cameras were going to replace traditional cameras at that time, it can also be seen that many powerful traditional retailers have never played e-commerce, traditional companies that are strong in long-term video have never played short-video, traditional payment Strong companies have never made electronic payments. “The advent of every technology is an opportunity to ‘turn the table’.”
Faced with such an opportunity, the advantage of a big factory is that it can quickly integrate large models, “Just like Microsoft embedded GPT into Office software, but if you really need to subvert Office in the future, will it be done by Microsoft or someone else? , I will bet that someone else did it.” Li Kaifu said that the large-scale model of a big factory can quickly generate value and generate applications, which will help financial reports, but to achieve subversion, it is likely to be a new developer. Do.
Disclaimer: The Securities Times strives for truthful and accurate information, and the content mentioned in the article is for reference only and does not constitute substantive investment advice, so operate at your own risk
Download the “Securities Times” official APP, or follow the official WeChat public account, you can keep abreast of stock market trends, gain insight into policy information, and seize wealth opportunities.