Sunday, April 28, 2024
HomeBusinessKai-Fu Lee Faces Criticism in the AI Arena: A Tale of Ambitious...

Kai-Fu Lee Faces Criticism in the AI Arena: A Tale of Ambitious Ventures and Controversial Claims

Amidst the turbulent waves of the domestic and international AI scene, touted as the “first windfall” of the AI era, a “shell game” controversy is unfolding.

Kai-Fu Lee, known as the “Father of Chinese AI,” finds himself deeply embroiled in a major setback in his forte, the field of AI.

I

In recent days, the global tech landscape has been exceptionally lively, with the central figure being Sam Altman, the “Father of ChatGPT.” His sudden removal from the board of OpenAI has triggered a “coup drama” regarding his future destination, stirring the nerves of the tech world. Upon learning the news, Tesla founder Elon Musk commented, “The public should know why the board made such a resolute decision. If it’s related to AI safety, it will affect the entire planet.”

OpenAI, the global leader in the tech arena, has unleashed a frenzy with its creation, ChatGPT. During this time, ChatGPT, large models, AIGC, and AI chips have dominated the global capital market. With a suction effect, the heat of AI has reached unprecedented levels, leading tech giants such as Huawei, Tencent, Alibaba, ByteDance, JD.com, Baidu, Xiaomi, and Meituan to secure tickets to the forefront.

However, the recent surge in domestic AI is not orchestrated by the tech giants but led by Kai-Fu Lee.

As one of the most prestigious figures in the Chinese venture capital scene, Kai-Fu Lee’s life story is legendary, marking him as an early explorer of “institutionalized” investment in China. In 2009, recognizing the growth opportunity in Chinese internet traffic, Kai-Fu Lee decided to bid farewell to Google and established the well-known early investment institution, Innovation Works, in the Haidian district of Beijing.

Benefiting from Kai-Fu Lee’s renown, Innovation Works invested in a batch of early-star projects, earning a seat on various major investment lists globally. Investments in Zhihu and Meitu brought substantial financial returns to Innovation Works. However, in 2013, Kai-Fu Lee faced a turning point in his career when he was diagnosed with lymphoma. He returned to his hometown in Taiwan for recuperation, temporarily leaving a void in the venture capital scene.

Upon returning to the venture capital scene, Kai-Fu Lee’s mindset underwent significant changes. He became less concerned about short-lived trends, redirecting his focus toward cutting-edge technologies with a “long-termism” perspective. In 2016, while the world was immersed in the O2O battle of the “Story King,” Kai-Fu Lee mentioned AI, stating, “In the next ten years, AI will spawn the most unicorn companies.” Consequently, Kai-Fu Lee became the biggest advocate for AI.

Innovation Works shifted its investment focus from the internet to AI, and Kai-Fu Lee was labeled the “Father of Chinese AI.”

In the following years, Kai-Fu Lee wielded significant influence in the AI field. Major domestic companies almost had to invite the “Father of Chinese AI” to speak at their events. Kai-Fu Lee lived up to expectations, not only excelling in AI investment but also contributing to the success of star enterprises like Megvii Technology, Fourth Paradigm, and Innovation Qichuang. The latter two are currently valued at HKD 24.4 billion and HKD 5.6 billion, respectively.

Undoubtedly, Kai-Fu Lee’s transition to AI investment once again yielded tremendous success.

However, his sharpness also exposed Kai-Fu Lee to an unprecedented AI public opinion storm.

II

“The Father of Chinese AI stumbled in the field he excels in.” This is a rather ironic joke.

This year, OpenAI stirred up the world’s largest tech storm, with tech giants competing in large models, AIGC, AI chips, and more. Naturally, Kai-Fu Lee, known as the “Father of Chinese AI,” was not willing to be left behind. He quickly incubated a large model project, “01Wanwu.”

Unlike before, Kai-Fu Lee actively participated in the entrepreneurship of 01Wanwu as a founder. This special involvement led the company to secure investment from Alibaba Cloud, a subsidiary of Alibaba Group, within a few months of its establishment, with a valuation of $1 billion.

01Wanwu entered the unicorn ranks shortly after its establishment, a sight unseen in the venture capital scene for many years.

This indirectly reflects that “currently, AI is undoubtedly the most sought-after dish in the eyes of capital.” So, why did 01Wanwu achieve a peak valuation of $1 billion so quickly? Leaving aside Kai-Fu Lee’s endorsement, it must have shown outstanding and unconventional performance.

Facing a new entrepreneurial journey, Kai-Fu Lee elevated 01Wanwu to an unprecedented height. In public, he stated, “From the first day of establishing 01Wanwu, my goal has been to create a world-class company that can enter the top echelon globally.”

Subsequently, to showcase the strength of being a “world-class company,” 01Wanwu released a series of open-source large models, Yi-34B and Yi-6B. Among them, Yi-34B is a bilingual basic model trained with 340 billion parameters. Kai-Fu Lee confidently claimed, “Yi-34B’s data collection, algorithm research, and team composition are all at the forefront globally, comparable to OpenAI and Google.”

“Yi-34B is the world’s most powerful open-source model, ranking first globally in various indicators such as general ability, knowledge reasoning, and reading comprehension.” Regardless of the actual strength of 01Wanwu, Kai-Fu Lee’s words elevated the description to the extreme.

Perhaps even Sam Altman would not dare to be so ambitious. It is unclear whether 01Wanwu can compete with OpenAI and Google, but Kai-Fu Lee’s grandiose claims and expectations for 01Wanwu as a “world-class company” have stimulated domestic AI peers.

The first to express dissatisfaction was Jia Yangqing, former vice president of Alibaba, known internally as an “amazing” AI scientist. During his doctoral studies, he founded and open-sourced the deep learning framework Caffe, adopted by Microsoft, Yahoo, and NVIDIA.

Jia Yangqing has always been a “pragmatist” who doesn’t like to tell stories. Rumors had it that he left Alibaba this year to venture into large model entrepreneurship. Logically, Jia Yangqing’s old workplace is Alibaba, and 01Wanwu is backed by Alibaba. Considering that Jia Yangqing’s technology has been used by Microsoft, and Kai-Fu Lee is Jia Yangqing’s senior, he shouldn’t have “looked away.”

But Jia Yangqing couldn’t hold back.

III

“For the first time, the ‘Father of Chinese AI’ faces questioning in his area of expertise.”

According to Jia Yangqing’s revelations, “while helping overseas clients adapt to a new model in China, a friend informed me that the model was actually based on the LLaMA architecture, with only a few variable names changed in the code.” Although Jia Yangqing, considering the seniority of his predecessor, did not specify the name of the company that developed the new model, all signs pointed to Kai-Fu Lee’s “world-class company,” 01Wanwu.

This incident has had a significant impact on the venture capital scene, bringing up the long-discussed topic of “shell games.”

There was a time when tech entrepreneurs were quite sensitive to taking money from early investment institutions. A founder, feeling frustrated, once expressed on social media, “I discussed financing with a well-known investor. While the investor smiled triumphantly, he kept suppressing the valuation of the project, saying, ‘If you think the money is too little, I won’t invest. I’ll quickly incubate a project myself, do it on my own, and by the way, squeeze you out.'”

Some venture capitalists in the industry are not shy about this topic, seeing it as “an efficient path to maximizing investment.”

Is Kai-Fu Lee engaged in a “shell game”? Entrepreneurs find it “hard to believe.” After all, Kai-Fu Lee is a leading figure in AI, and if he were to engage in deceptive practices, the so-called “world-class company” would become a laughingstock, and Kai-Fu Lee’s public image would collapse.

Given Kai-Fu Lee’s reputation in the industry, he is definitely not a “big shot” with a “mouth full of nonsense.”

Realizing the seriousness of the public opinion bombardment, 01Wanwu responded to the “shell game” allegations. They stated, “The structural design of its large model is based on the mature GPT architecture, drawing inspiration from top-notch public achievements in the industry. Due to the early stage of large model technology development, maintaining a consistent structure with the industry mainstream is more conducive to overall adaptation and future iterations. At the same time, based on 01Wanwu’s team’s understanding of models and training, they have done a lot of work and continue to explore breakthroughs at the model structure level.”

01Wanwu’s response cast a shadow over Kai-Fu Lee, who was already in the spotlight of public scrutiny. In other words, 01Wanwu’s Yi-34B still drew inspiration from external achievements, striving on the shoulders of giants, not as miraculously described by Kai-Fu Lee.

In fact, even before Jia Yangqing’s revelation, foreign experts pointed out that the Yi-34B model by 01Wanwu basically adopted the LLaMA architecture. However, 01Wanwu responded to the questioning after the cat was out of the bag, expressing apologies for the oversight in renaming. Inadvertently, it dealt a blow to the “Father of Chinese AI.” This is the first time Kai-Fu Lee has taken a big tumble in the field he excels in, but he handled it with grace.

Kai-Fu Lee’s grace lies in 01Wanwu’s willingness to confront questioning and not lead the discussion into a “roshomon.”

IV

Establishing 01Wanwu and releasing Yi-34B, Kai-Fu Lee seems a bit impatient.

According to a seasoned media person who has long been following the AI market, Kai-Fu Lee’s current round of entrepreneurship appears hasty.

“You can see how 01Wanwu operates, as if rushing against time.” The media person attributed “Kai-Fu Lee’s haste” to the fact that he has always lacked a world-class masterpiece to prove his capabilities in the field of AI. Although Megvii Technology and Fourth Paradigm are noteworthy domestically, they lack hardcore competitiveness compared to foreign giants. Kai-Fu Lee, preaching AI to gain global recognition, urgently needs a bold move.

In September of this year, the authoritative American magazine “Time” visited 25 “AI leaders” worldwide, including Sam Altman, Elon Musk, Huang Renxun, and Kai-Fu Lee. This indicates that Kai-Fu Lee’s contributions to AI are no less than those of the aforementioned big shots. At least in the eyes of “Time” magazine, Kai-Fu Lee is on the same level as Sam Altman, Elon Musk, and Huang Renxun.

However, from a business achievement perspective, OpenAI and NVIDIA are solidly “far ahead.” Kai-Fu Lee is eager to create a company that can compete with the aforementioned tech giants or, from his perspective, fill the gap of such companies domestically.

So, the outside world saw Kai-Fu Lee, who rarely boasts, starting to boast. By rapidly raising the valuation of 01Wanwu to shorten the distance from a “world-class company.” If no one comes forward to question, tech unicorns often continue to initiate the next round of financing after launching a new product, doubling their valuation. When the plan is disrupted, it’s not necessarily a bad thing for a peer to come forward and pour cold water.

“The ‘Father of Chinese AI’ is impetuous, how can other companies develop calmly?”

An angel investor expressed on social media, “Calmness doesn’t necessarily mean being right; the speed of AI iteration is unimaginable.” The investor said, “Before the National Day holiday, domestic projects regarded GPT-4 as a benchmark, and after the holiday, foreign countries produced GPT-4V.”

Calmness doesn’t necessarily mean being right, but being too impetuous is certainly not the solution. Blindly chasing trends and riding on popularity can also bring chaos to the industry. Hopefully, this big stumble will help Kai-Fu Lee regain clarity and use technology to build a “world-class company in Chinese AI.”

Most Popular

Recent Comments