Zero One Wanwu, a large model company under Kai-Fu Lee, is online, with dozens of core members in place

The house comes from Aofei Temple
Qubit | Official account QbitAI

Who is China’s most competitive large-scale entrepreneurial dream team?

Zero and ten thousand things .

Yes, at least starting from today, this name will start a new stage of being noticed, discussed, and tracked.

There is no other reason - this is the AI ​​2.0 company incubated by Dr. Kai-Fu Lee who "rolled up his sleeves" and personally led the team. In late March this year, Sinovation Ventures chairman and CEO Kai-fu Lee officially announced his entry into the large-scale model under the name of Project AI2.0... And Zero One Thousand Things is the official company name of this startup. The official website 01.AI is also today Online together.

Qubit learned that after nearly three months of preparation, Zero One Thing has also made a series of progress, including:

Company: The headquarters is registered in Beijing, but it recruits top talents from all over the world.

Route: Self-developed large model. A self-built platform builds thousands of GPU clusters for training, tuning and inference, increasing the amount of effective data parameters and the density of high-quality data used to build continuous, controllable and scientific training capabilities.

Progress: Within three months, the internal test of the model with tens of billions of parameters has been completed, and the model is being expanded to a scale of 30 to 70 billion parameters.

It is reported that in just 3 months, Zero One Thousand Things has gathered a team of dozens of core members , focusing on large model technology, artificial intelligence algorithms, natural language processing, system architecture, computing power architecture, data security, product R&D and other fields. Among them, members of the joint creation team who have joined include the former Alibaba Vice President, the former Baidu Vice President, the former Didi/Baidu chief algorithm leader, the former Google China executive, the former Microsoft/SAP/Cisco/Vice President, algorithm and product team The backgrounds are from major domestic and foreign manufacturers.

Since returning to China to establish Microsoft Research Asia 25 years ago, Mr. Kaifu has become the most well-known mentor and talent in the Chinese computer industry, objectively boosting the development of China's Internet and AI.

Now, standing at the juncture of large-scale models and the wave of AI 2.0, former students, former colleagues and admirers are following and gathering again.

01Wanwu is the seventh company incubated by the Sinovation Ventures AI Engineering Academy founded by Kai-fu Lee . This special approach will combine fund investment and spire incubation to accelerate the creation of a new entrepreneurial ecosystem for AI 2.0.

Zero One Thing, self-developed universal large model

Public industrial and commercial information shows that Beijing Lingyiwu Technology Co., Ltd. (hereinafter referred to as Lingyiwu) was established on May 16. At that time, it was exactly 2 months after Kaifu Li’s circle of friends officially announced that he would personally join the large model business.

What is the explanation for "zero one thing"?

It is not difficult to understand 01. The underlying rule binary of the computer, 01, is the native language of engineers. But Qubit further heard that in Kaifu's view, zero one - 01 also represents the entire digital world, from zero to one, and even all things in the universe. The so-called Tao gives rise to one... gives birth to all things, which means "01 intelligence, all things are endowed with wisdom" "can" ambition.

Industrial and commercial information also shows that Lingyiwan’s business scope covers artificial intelligence mass entrepreneurship and innovation service platform, artificial intelligence basic resources and technology platform, engineering and technology research and experimental development, data processing services, etc.

This scope is also consistent with Kaifu’s thinking and judgment about AI2.0. After ChatGPT led to the craze of large models, Kaifu Lee publicly stated that the AI ​​2.0 era, with the base large model as a breakthrough, will set off multiple technologies, platforms and applications. level revolution. Just as Windows has driven the popularity of PCs and Android has given birth to the mobile Internet ecosystem, AI2.0 will create platform opportunities that are ten times larger than the mobile Internet. It will rewrite existing software, user interfaces and applications, and will also give birth to a new era. batch of AI-first applications and the birth of AI-led business models.

More business details of Zero One Thousand Things have also been disclosed:

The concept is AI-first , the driving force is technical vision , backed by the excellent Chinese engineering heritage , and the breakthrough point is the large base model , covering multiple levels from technology, platform to application .

More specifically, the large model entrepreneurship route chosen by Zero One Wanwan is self-developed large models .

And it has already begun - this company has begun to build an experimental platform and built a GPU cluster with thousands of cards for training, tuning and inference. In terms of data, the focus is on high-quality data density that increases the amount of effective parameters and usage.

From the announcement to the appearance, in the past three months, Lingyiwuwu has implemented an internal test of the model with tens of billions of parameters, and is expanding to a scale of 30-70 billion parameters. In the future, it plans to launch an open source model version.

It is reported that after the universal large model, the next step for Zero One Wagon is to launch a multi-modal model that combines pictures, videos, and 3D .

The entry-level Super App is also part of the plan, and platform middleware and developer tools will also be gradually released to the outside world.

At the same time, it was also exposed that Zero One Wish’s “seven module technology routes” in the AI ​​2.0 era .

In short, Lingyiwu believes that in order to build a large model platform, it needs to recruit top scientists and engineers in seven major directions, namely:

• Pre-training framework (Pre-train) technology that defines the ceiling of model capabilities

• Post-train technology to improve model implementation capabilities

• AI Infra technology that supports efficient model training

• More diversified multi-media multi-modal model technology

• Practical Training Scaling Technology

• Platform middleware and tool chain technology

• Inference and service deployment technologies

Based on these 7 technical modules, Zero One Wanwan has recruited a technical team in a targeted manner.

Qubit learned that the team currently has dozens of core members in place. The algorithm and product team backgrounds are all from major domestic and foreign companies , focusing on large model technology, artificial intelligence algorithms, natural language processing, system architecture, computing power architecture, Data security, product research and development and other fields.

There are elite soldiers and strong generals. Judging from the lineup of the joint creation team that has been confirmed to join, it looks like an "AI 2.0 dream team":

Former Alibaba Vice President, former Baidu Vice President, former Didi/Baidu Chief Algorithm Head, former Google China executive, former Microsoft/SAP/Cisco/Vice President...

It can be said that the core team has been deeply involved in the construction of some of the world's largest software platforms.

The evolution of China's large models is divided into three stages. Only the view from the top of the tower can be fully utilized.

Since starting the incubation of the spire of Zero One Thousand Things, Teacher Kaifu has also started an intensive preparation mode, which is very busy and very low-key.

But the qubit also learned its latest thinking and judgment about the evolution of large models. In Kaifu Lee's view, the evolution of China's large language model will not be achieved in one step, but will go through three stages from L1 to L3.

At this stage, most companies in China are in the L1 stage. The next step is to cross the "emergence" threshold and become a mainstream large language model at the L2 level. Finally, they will catch up with L3 global first-tier companies and become a powerful AI 2.0 "technical infrastructure" to achieve high-quality, High performance goals.

The core reason why Lingyiwu chose the road of self-developed large models is that the technical implementation of open source models and small models is absolutely unable to achieve the performance of self-developed models by overseas manufacturers .

Judging from actual experience at this stage, many open source models that claim to be similar to ChatGPT are limited to simple conversations. They are limited by the structure and size of the model itself in tasks such as long text generation and reasoning, and cannot adapt to slightly generalized or complex tasks. .

In addition, within Zero One Wish, it is also determined that from the perspective of medium and long-term development, over-reliance on models means that we can only know the results but not the process, "knowing what is happening but not knowing why." Your own ability ceiling and boundaries are defined by others, and you will soon hit the ceiling of your ability to accomplish a variety of tasks. Situations, cultures, user habits, laws and regulations at home and abroad are also different. Overseas open source technology may be shut down, switched to high charges or restricted regional access at any time, and other risks.

Only by fully integrating Chinese language data can the independent and controllable development of AI 2.0 be realized.

However, Lingyiwu also revealed that it plans to launch an open source model version in the future to attract outstanding talents to jointly improve the model and promote the global AI 2.0 developer ecosystem. Therefore, large-model startups should embrace open source, but not be overly dependent on the open source model. Only by establishing our own technological advantages as soon as possible and building a strong ecological moat can we truly seize the historical opportunities of AI 2.0.

Kaifu Lee believes that the diversification of large model companies will give rise to a wave of innovation ecology from AI 2.0 models, AI 2.0 infrastructure, to new AI 2.0 applications. “We predict that once China has a truly original, high-quality large model launched, High-quality and creative applications will flourish - just like the era of mobile Internet and AI 1.0."

There will be a few large model companies in the global market with abundant resources to support the demand for computing power, driving the prosperity and development of various downstream applications, gradually improving the new ecology of China's AI 2.0, exploring new business models, and thereby creating huge economic value and social value. “If the market is flat or there are only a few players, then we really need to worry about being left behind.”

Kai-fu Lee also believes that the AI ​​2.0 market is huge and can accommodate three levels of corporate competition: giants, small and medium-sized enterprises, and startups.

However, "the technical threshold for large models is very high, and it is very necessary to integrate an efficient team with both scientific research capabilities and development capabilities to formulate and implement the company's technology and product strategies and execute them consistently from top to bottom."

Objectively speaking, China may be a few years behind in R&D breakthroughs, but it is good at technology iteration and has extremely excellent engineering capabilities and execution speed. The future is promising. Compared with large companies, startups with top-notch technology and the ability to focus on mastering large-scale model play have the advantages of concentrated resources, flexible strategies, the ability to fight tough battles, and strong implementation capabilities. On the contrary, they have a better chance of leading the industry, just like the OpenAI company behind ChatGPT. Same.

And this may be the core reason why Kai-fu Lee chose to enter the game personally and incubate Zero One Thousand Things at Spire - starting a business with a self-developed large model base requires not only technology and talent, but also the integration of Spire resources and the cognitive vision of Spire.

也只有李开复本人躬身入局，零一万物才显得如此不同。在众多的创新梯队中，零一万物具有显著优势和竞争特色。李开复拥有逾40年国际AI经验，亲自牵头筹组，全球范围的顶级AI专家陆续加入，目前组成的核心团队深度参与过世界最大几家软件平台的构建，深度理解平台成功的方法论。

"AI is the explanation of the human learning process, the quantification of the human thinking process, the clarification of human behavior, and the understanding of human intelligence... AI is the last mile for humans to know and understand themselves. I hope to join this new bloom. , a promising future scientific field."

This is Kai-fu Lee’s application letter for studying for a PhD at Carnegie Mellon University in 1983. As one of the pioneers in the field of artificial intelligence, he has been active in artificial intelligence research and industry for 40 years.

In 1998, Kaifu Lee founded Microsoft Research China (the predecessor of MSRA) , which has become the “AI institution” that cultivates the most artificial intelligence talents and outputs the highest quality talents in China. The young talents of that time are still shining brightly in the artificial intelligence trend today.

It is this time-tested achievement that has allowed Dr. Kaifu Li to gradually become a word-of-mouth AI pioneer . It is also this time-tested achievement that has led more young engineers who are full of faith in AI 2.0 and AGI to respond to the call. , seize opportunities in the wave of the technological era and change the world.

According to the latest news, 01Wang’s recruitment of global talents continues. The resume channel on the official website 01.AI shows:

http://01.ai/ (To jump, please click at the end of the article to read the original text)

Lingyiwu welcomes outstanding talents with AI 2.0 technical strength and AGI beliefs to join us to build a new AI 2.0 platform and accelerate the arrival of AGI.

—Contact the author— ​

Click here ???? Follow me and remember to star~