Solve the problem of AI.
文 | 华商韬略 耿康祁
In 2023, an AI storm will come, which will immediately set off a wave of large-scale model involution of many top technology companies around the world. For a time, the enthusiasm and investment of countless investors, science and technology enterprises and entrepreneurs were all attracted by the word "big model".
However, on December 16, at the Geek Park Innovation Conference 2024 held in Beijing, Robin Li, the founder, chairman and CEO of Baidu, clearly proposed that the real value of the era of large models lies in AI native applications.
[Baidu, one step ahead]
"There must be thousands, if not millions, of AI-native applications on top of the basic model for the value of the large model to be reflected. ”
At the Geek Park Innovation Conference 2024, Robin Li, who came here for the sixth time, threw out a sharp insight into the general trend of AI as soon as he took the stage.
In a simple sentence, it contains at least three meanings: the real dividends of the era of large models have not been fully explored, and the opportunities are great; AI native applications are the greater future; Baidu, which has seen this trend, will inevitably take the lead in "doing things".
2023 is undoubtedly a turbulent year for the AI industry, and the competition around large models has triggered many industry crazes.
Whether it is entering artificial intelligence many years ago, or Baidu fired the first shot of the release of domestic large models with Wenxin Yiyan at the beginning of the year, and then continued to iterate and build end-to-end optimization solutions with large models in the middle of the year, leading the technological transformation of AI large models, and deeply integrating into industrial development, Baidu has always been one step ahead of others, taking the lead in layout, and then defining and leading the track.
Behind this, it is Robin Li who is lighting the fire for Baidu and illuminating the future of the AI industry.
In 2020, Robin Li wrote on social media: "I actually turned out the notes from the artificial intelligence class at Peking University 30 years ago. ”
As a "Mr. AI" who has a deep relationship with the AI field, Robin Li has always been an optimist who understands and believes in AI technically and commercially, and has made Baidu the first "crab eater" in China to deploy AI in depth more than ten years ago.
On March 16 this year, at the "Wenxin Yiyan" press conference, Robin Li revealed that the first version of the Wenxin model in 2019 has been released, and there have been new iterations every year since then.
The long-term deep cultivation with great foresight has become the biggest hole card for Baidu Wenxin Yiyan to quickly lead the outbreak of large models in the domestic market.
When people's evaluation of the application of AI large models mostly stays in the impression of "an easy-to-use intelligent software", Robin Li took the lead in proposing in China that this AI wave is comparable to the "fourth scientific and technological revolution".
On June 26, at the Nishan Dialogue on Digital Civilization at the Internet Conference, he once again emphasized that "the big model is a major technological change that affects human development, an engine that drives global economic growth, and a major strategic opportunity that must not be missed." ”
Another business tycoon who agrees is Bill Gates from across the ocean, "Whether you like it or not, this will lead to a new industrial revolution that goes beyond the invention of the Internet and the iPhone." ”
As a result, in the face of the pessimism that the first-mover advantage of foreign large models is too strong, Robin Li decisively chose "All in AI", and also chose the path of going against the current for Baidu - "One day we will catch up".
This prediction soon took the first step: Baidu's Wenxin Yiyan became the first generative AI product released by a major global manufacturer.
The iterative optimization of Baidu Wenxin Yiyan has also been accelerated. Version 3.5 of the Wenxin large model released in June has greatly improved the training speed and inference speed. Just four months later, it was replaced by the fully upgraded Wenxin Model 4.0, and Robin Li bluntly said that its comprehensive capabilities are "not inferior to GPT-4".
This year, Baidu "shined" on the AI stage, and Robin Li was the "conductor" in front of the stage who dominated everything.
With his forward-looking insights, he not only allowed Baidu to maintain its leading position in domestic AI and wrestle with the world's top AI giants, but also one of the leaders who drove the continuous acceleration of change in the development cycle of the domestic AI industry.
On September 7 this year, Time magazine announced the first list of the world's top 100 AI figures, and Robin Li became the highest-ranked Chinese representative in the "AI Leader" quadrant, which shows his influence.
At Baidu World 2023 in October, Baidu made another big move to trigger industry focus: releasing more than ten AI native applications such as new search and new maps. Facing the camera, Robin Li boldly predicted: "We are about to enter an era of AI native. ”
At the end of the year, Robin Li once again issued a call for "must roll up AI native applications", and its key significance is self-evident.
[Beware of "dead ends"]
If 2023 is called the year of "AI big models", it will not be an exaggeration at all.
According to public information, as of October, 238 large models were released in China alone, while in June the number was 79, a threefold increase in less than half a year.
The AI community seems to generally regard large models as the key to unlocking the battlefield of the future.
The reality is that both ChatGPT and Baidu Wenxin Yiyan will immediately go to the next battlefield of the "application scenario" after the large model is successfully polished. They know that technology is only valuable if it is truly implemented in specific scenarios and needs.
AI entrepreneurs and investors who are obsessed with building the next "big model" can easily break into a "dead end". OpenAI founder Sam Altman even warned, "I'm skeptical of all startups that try to train their own big models." ”
This point, Robin Li's sharing from the beginning of the year to the end of the year, is very clear and even urgent.
He has pointed out more than once that under the catalysis of "AI for all", many industries and enterprises want to train their own dedicated large models from scratch, which may be meaningless.
This is because only when the scale of model parameters is large enough, the amount of training data is large enough, and continuous investment and long-term iteration can intelligence emerge and the large model can have the ability to bypass classes.
This means that the time, money, and risk costs of building a large model from 0 to 1 are huge, and the profitability may be far away.
For example, Baidu made the Wenxin model, which took more than ten years and invested nearly 150 billion. And the mature large model is constantly evolving, and players who enter too late are likely to burn out their money but can't catch up with others, and they don't even have the opportunity to drink soup.
At the Shenzhen 2023 Xili Lake Forum on November 15, Robin Li poured cold water on the "vigorously produce miracles" thinking of the 100-model war: "What we need is 1 million AI native applications, not 100 so-called large models." ”
At the Geek Park Innovation Conference 2024, Robin Li mercilessly pointed out that the 100-model war is a great waste of corporate resources and social resources, and more resources should be placed on the exploration of AI super applications.
This heralds a huge shift in the market focus of the AI field.
In September of this year, Sequoia USA released an article proposing that "generative AI is moving from Act 1 to Act 2, and the hype and rapid display of large models will be replaced by real value landing and complete product experience." ”
It reveals that the sheer scale of unlocking potential killer applications and end-user needs is the real AI dividend.
In this regard, Robin Li gave a more accurate analogy: in a more long-term way, the large model will eventually become a basic base similar to the operating system, and entrepreneurs will rely on a few large models to develop a variety of native applications.
At the conference on December 16, Robin Li, who often presents himself as the "helmsman" of technical decision-making, was rarely a little emotional: "Only the native application of volume AI is valuable, and the progress of large models is not an opportunity for the vast majority of people." I see that the main excitement of the media, society, and the public is still on the basic model, and it has not moved to AI native applications, so I am more or less anxious. ”
Robin Li is really worried that a large number of AI entrepreneurs will step into the "trap" of large models, and they will miss the real AI explosion dividend.
[AI Entrepreneurship "Incubator"]
At the press conference of "Wenxin Yiyan" at the beginning of the year, Robin Li pointed out that in the era of large models, the real biggest opportunity is "precisely in application".
At the conference on December 16, Robin Li gave AI entrepreneurs three "reassurances": a good environment, a large market, and a high ceiling.
First, there are relevant industrial policies to encourage the development of AI native applications based on large models; second, AI native applications on top of large models can affect the key indicators of most enterprise businesses, bringing revenue and profits; and it is very likely that startups will make three or five super applications, and it is more likely that they will make hundreds or thousands of very valuable vertical applications.
Microsoft's Office 365 Copilot, priced at $30 a month, may earn $5 billion a year, which is many times larger than OpenAI's annual revenue. Robin Li believes that the innovation of AI products and the great value of the world's largest technology companies are inspiring enough.
Baidu itself is also a beneficiary of running into the AI-native era. At the beginning of November, after the launch of version 4.0 of the Wenxin model, which charges a subscription fee, the number of paid users has maintained a high growth rate every day.
The business model with a broad business scope and easy to run through, as well as the advantages of not burning too much money and its own monetization ability, make AI-native application startups more friendly and more attractive. But starting a business is never an easy road.
This year, Robin Li has no shortage of experience and methodological thinking on how to implement and win AI entrepreneurship, which can be called a hand-in-hand teaching.
At the Wenxin Cup Entrepreneurship Competition in September, he proposed three conditions for entrepreneurs to build AI-native applications: first, they must be able to interact with natural language, which is the most fundamental change, second, they can make full use of the ability to understand, generate, logical, and memorize that they did not have in the past, and third, each application should interact with no more than two menus.
At the Xili Lake Forum, Robin Li emphasized the necessity of implementation, proposing that "embracing the native era of AI needs to be driven by CEOs and top leaders, because only top leaders pay attention to the role of new technologies in promoting key performance indicators".
At the conference on December 16, Robin Li further answered key questions such as the basic principles, development paradigms, entrepreneur characteristics, and product managers of AI native applications.
He mentioned that the truly successful product manager of AI native application is likely not to be a certain type of person, but a comprehensive talent with strong Xi learning ability, product and market sense, and at the same time not afraid of technology.
Under the leadership of Robin Li, Baidu has also used Baidu PaddlePaddle Deep Learning Xi Platform, Baidu Intelligent Cloud, Galaxy Large Model Community and other tool platforms to allow more AI native application entrepreneurs, developers and enterprises to have an "incubation base camp" and no worries.
Throughout the year, Robin Li has played multiple roles in the AI wave: he is an AI business leader, a technological changemaker and leader in the AI industry, and a mentor and evangelist for AI entrepreneurship.
The upcoming 2024 is destined to be swept into a new storm of "AI native applications".
For the industrial question of "where and how AI is going", Robin Li and Baidu, who have submitted high-scoring answer sheets, have also prepared textbook-level answers for what to do and how to do AI entrepreneurship in 2024.
——END——
Welcome to pay attention to [China Business Strategy], know the people, and read the legend of Taostrategy.
All rights reserved, unauthorized reproduction is prohibited
Part of the picture comes from the Internet
If there is any infringement, please contact us to delete