1
On April 16, the "Create 2024 Baidu AI Developer Conference" was held, and Robin Li brought three major AI development tools, namely AgentBuilder, AppBuilder, and ModelBuilder.
Robin Li has drawn a grand future for Baidu to the outside world, based on the Wenxin Yiyan model, natural language will become the new general-purpose programming language, as long as you can speak, you can become a developer. His exact words are "everyone is a developer".
AgentBuilder is an agent development tool, AppBuilder is an AI native application development tool, and ModelBuilder is a model customization tool of various sizes.
On April 15, Robin Li and Jiyue CEO Xia Yiping jointly broadcast live and showed off autonomous driving. He said that automotive robots are the future, and the basis of automotive robots is also the multimodal model of Wenxin Yiyan. He believes that the biggest application scenario of the vision model is autonomous driving.
The future drawn by Robin Li can be summarized by "three horizontal and one vertical", which refers to the three development platforms of AgentBuilder, AppBuilder, and ModelBuilder, which are aimed at toB developers, autonomous driving, and toC consumers.
Refinement granularity, different from the PC era and the Internet era, the boundary between toB and toC in the AI era is disappearing, based on Wenxin Yiyan, Baidu AgentBuilder and AppBuilder as a platform can be directly connected to C with app and agent as a springboard. Only ModelBuilder is pure toB.
Autonomous driving is toC from the product dimension, directly for car drivers and passengers, but commercial implementation is toB, cooperating with vehicle manufacturers such as Jiyue.
Different from traditional toB software products such as database, office collaboration, and CRM, the self-driving system is a platform, and the platform is an ecology, including software ecology and hardware ecology, similar to the OS in the PC and mobile Internet era, which is much larger than the imagination space of software products.
Regardless of "three horizontal" or "one vertical", the base is a large model of Wenxin Yiyan, which is the infrastructure of Baidu World, similar to an operating system. The "three horizontal and one vertical" based on the Wenxin Yiyan model is Robin Li's "Longzhong pair" for the AI era.
Baidu was founded in 2000 and has gone through 24 spring and autumn seasons, C-end information services, B-end marketing services, Baidu has never broken through such boundaries. In the face of the new era, Baidu has greatly broken through its own boundaries by coupling toB and toC, from tools to applications, from platforms to ecosystems.
From the perspective of social empowerment, Robin Li's "three horizontal and one vertical" liberates people, everyone is a developer, and also liberates enterprises, paraphrasing Alibaba's "there is no difficult business in the world", that is, "there is no difficult company in the world, and there is no difficult business".
2
In 1995, Jack Ma, who was unwilling to be a teacher, decided to start his third business, his first venture was to engage in translation, and foreign languages were his specialty. The translator didn't make any money, so he began to sell small gifts, focusing on flowers, and it was the second time he started a business, which was also not popular.
Jack Ma's third venture is called China Yellow Pages. The business of China Yellow Pages is okay, but there is a conflict with the major shareholder Hangzhou Telecom, Ma Yun resigned from the company he founded, and Ma Yun, who doesn't know what to do, began to drift north. After three years of drifting north to nothing, he returned to Hangzhou in 1999 to found Alibaba.
Robin Li is four years younger than Jack Ma, but Robin Li's starting point for entrepreneurship is much higher and smoother than Jack Ma.
After graduating from Peking University in 1991, Robin Li went to the United States to study, and in 1994, he joined Wall Street Jones and successively served as a senior consultant for Dow Jones' subsidiary and a designer of the real-time financial information system for the online version of the Wall Street Journal. In 1997, Robin Li left Wall Street and went to Infoseek, a well-known search engine company in Silicon Valley. In between work, he also wrote a book "Silicon Valley Business War", which is concise and a good reading book on Silicon Valley history.
In 2000, he founded Baidu with a $1.2 million raised from the "hyperchain analysis" patent. Robin Li and Jack Ma have become competitors on the same track: in essence, Alibaba (corporate BBS), China Yellow Pages, and search engines all connect supply and demand digitally.
They didn't know each other at the time, and they didn't expect to become commercial competitors. The climax of the story is in August 2005, when Alibaba acquired Yahoo China, Jack Ma and Robin Li wrestled their wrists face-to-face, and Baidu and Alibaba saw bayonets in the search market.
2007 is a key node in the history of China's Internet, at this time the development of the Internet has gone through ten years, the past ten years is the decade of the Internet boundary gradually blurred, the decade after 2007 will be the Internet giant boundaries gradually clear ten years, information, e-commerce, games, entertainment, entrepreneurial heroes are good at winning, each leading the way.
Robin Li is indeed the most tech-savvy boss, Baidu dominates the search industry, Google loses, 360 and Sogou barely survive; Ma Yun is indeed not good at computers, but it is not English, but business, and e-commerce has become Ma Yun's base camp.
Robin Li insisted on the direction of "connecting supply and demand in a digital way", and Jack Ma gave up this direction, one to the left and the other to the right, but this did not prevent the success of the two men. Alibaba transfused Taobao, and Ma Yun "can't find an opponent with a lantern" in the field of e-commerce. They all bring their talents to their respective areas of expertise.
In the real society, the social subject is divided into two categories, individuals and institutions, China Yellow Pages, Alibaba (BBS), enterprise station, and search keywords all solve a problem: how does the institutional subject of the enterprise exist in the Internet world, that is, to present, show itself, and collaborate and communicate with the outside world.
China Yellow Pages, Alibaba (BBS), enterprise websites, and search keywords are the forms of existence of enterprises and organizations in the PC era, and in the era of mobile Internet, they have become new forms of existence such as official accounts, mini programs, and live broadcast rooms.
What about the era of AI models? Certainly not in the form described above. Throughout 2023, there will be a lot of debate about large models, and Agent has gradually become the consensus of the industry about the existence of enterprises and organizations in the AI world. Unlike all previous forms, agents can interact with consumers and customers as if they were real people.
At the beginning of this year, some people predicted that Agent would become the hottest direction for large models this year. Robin Li's opinion at the "Create2024 Baidu AI Developer Conference" confirms this prediction.
AgentBuilder helps enterprises run into the era of large models, and AppBuilder lowers the threshold for enterprises to enter the Internet world. The relationship between the App and the Agent is not a flat relationship, but a coexistence relationship, just like the parallel coexistence of the App, the enterprise website, and the search keyword.
3
The boundaries between giants change from chaos to clarity, and then from clarity to chaos, repeatedly, which is the "cyclical law" of the Internet industry.
From 2007 to 2017, the boundaries of the Internet were clear and chaotic again ten years later, and the emergence of new forces such as Douyin, Pinduoduo, Meituan, Xiaomi, and Didi expanded the boundaries of the entire Internet industry.
In the field of cloud computing and AI large models, Robin Li once again bumped head-on with Jack Ma. In the field of autonomous driving, Baidu is actively deploying, and Ali has tried several times, and under the logic of returning to the core business, Alibaba is unlikely to invest too much in the field of autonomous driving.
This time it is no longer the competition of small start-up companies, in addition to Alibaba, Baidu, Byte, Tencent, Douyin, Huawei and other giants are gathering momentum, there is also an influx of entrepreneurs like Wang Xiaochuan. Everyone is aware that the technology industry has entered a new cycle of innovation. If you miss this time, you may have to wait another ten years.
The moment when the temptation is the most is also the most dangerous, one wrong step, one wrong step, and finally you may step on the empty era and lose ten or even twenty years. In the history of the development of the Internet, there are countless such stories. This is especially true for large companies, where small companies have a small ship and it is difficult to turn around, and large companies have a big difficulty turning around.
At such a tempting and dangerous moment, systematic strategic thinking is particularly crucial, and Robin Li's "Longzhong pair" is the result of systematic strategic thinking about the future.
A key dimension of this kind of systematic strategic thinking is "the distribution of the right to speak in all aspects of the industry", from the PC Internet, mobile Internet, to the current AI model, the industry pattern is changing, and the distribution of the right to speak in all aspects of the industry is also changing.
In the PC Internet era, software and hardware are separated, Microsoft's Windows and Intel's chips are platforms, and there are two ecosystems on the platform, one is the application software ecology, the other is the hardware ecology, and the most powerful one is the platform, that is, the "Wintel architecture".
In the era of mobile Internet, software + hardware + Internet, Apple created the "triathlon" model, the right to speak began to shift to the terminal, and the pattern of platform monopoly power was broken. Even in the Android ecosystem, terminal manufacturers have a much stronger voice than in the PC era.
In the architecture of "three horizontals and one vertical", the bottom layer is the Wenxin Yiyan model, the middle layer is the secondary model developed by enterprises based on ModelBuilder, and the top layer is APP and Agent to support enterprise business innovation and management innovation.
Autonomous driving as a special zone, based on the Wenxin Yiyan model, the car will become another intelligent networking terminal after PC, TV, mobile phone has become an industry consensus, the large model is accelerating the arrival of autonomous driving. The bigger imagination is that autonomous driving can be an ecosystem of its own.
4
In 208 AD, Liu Bei looked at the thatched house, Zhuge Liang suggested that "first take Jingzhou as his home, and then take Xichuan to build a foundation, in order to become a full momentum", this is not the end of Longzhong's strategy, the end game is the next step, is "then the Central Plains can be mapped".
"When there is a change in the world, then a general will be ordered to send the troops of Jingzhou to Xiang Wan and Luo, and the general will lead the people of Yizhou to go out of Qinchuan, and the people will not eat pot pulp to welcome the general?" Zhuge Liang's answer was this.
"Three points have become a dream, and future generations complain out of nothing", and the reason for the failure of the "Longzhong pair" strategy is not a strategic failure, but a problem in implementation. After "Guan Yu carelessly lost Jingzhou", the Northern Expedition could only cross the Qinling Mountains from Hanzhong to attack from a long distance, and the journey of thousands of miles, people were sleepy and horses were scarce, and it was more difficult to transport grain and grass.
The endgame of "three horizontal and one vertical" is not a three-point world, but a plan to unify the world, which includes a platform (Wenxin Yiyan model) + middle layer (ModelBuilder) + application ecology (AppBuilder, AgentBuilder), similar to Apple's "software + hardware + mobile Internet" layout in the mobile Internet era.
Whether it is the PC Internet era or the mobile Internet era, Chinese enterprises are only innovating at the application level, and there is no room and opportunity for innovation in the field of platforms and underlying technologies. If the "three horizontal and one vertical" is successful, Chinese companies will prove themselves in the field of platforms and underlying technologies.
"Three horizontal and one vertical" is a grand strategic layout, and the biggest challenge for Baidu is also in organization and execution. Baidu is known as the "wolf factory", learning from Huawei, Huawei advocates military culture, from the perspective of efficiency, in Chinese enterprises, belongs to the ceiling of combat power. It makes sense for Baidu to choose Huawei, and in order to achieve such a grand cause as "three horizontal and one vertical", it is indeed necessary for Huawei to execute.
Another risk of Baidu's "three horizontal and one vertical" strategy is that the trend of technology development is uncertain. There are currently two modes of development of large models, one is the open source model, and the other is the closed-source development, the model of OpenAI and Baidu is closed-source, and the model of Google abroad and Alibaba in China is open source.
Which path will win in the future is a big gamble for the participating manufacturers.