The post-90s Ali model boss left his job to start a business

The ever-changing world of AI has also added some new possibilities.

The latest news, Zhou Chang (nickname: Zhong Huang), the technical director of Ali Tongyi Qianwen large model, will resign from Alibaba Cloud.

Zhou Changyuan is one of the core technical backbones of the algorithm team of Tongyi Laboratory, and this is a normal resignation. The R&D and open source work of the Tongyi large model is still in progress, and the current person in charge of the Tongyi Lab is Zhou Jingren, CTO of Alibaba Cloud. Zhou Chang is the person in charge of the large model of "Tongyi Qianwen".

According to his resume, Zhou Chang graduated from Fudan University with a bachelor's degree in computer science and technology in 2012 and a doctorate degree in computer software and theory from Peking University in 2017. He successively worked as an algorithm expert in the Damo Academy, and then as the head of the large model of Tongyi Qianwen. He specializes in deep learning, graph mining, and distributed computing, and has published more than 30 articles at top international conferences in the fields of machine learning, data mining, and databases. It is worth mentioning that according to multiple media reports, Zhou Chang has decided to start a business and point to AI. However, the exact direction is not yet known. Interestingly, there is currently news from many headhunters on the Internet that Zhou Chang's entrepreneurship has been favored by many investors. This also includes Yuanjing Capital, a subsidiary of Alibaba's Wu Ma and Wu Yongming.

There is no specific information about Zhou Chang at present, but according to his undergraduate graduation in 2012, Zhou Chang is probably a post-90s generation.

This post-90s generation, which is closely connected with the glorious history of the "Tongyi Large Model Family", ushered in a turning moment. Or will transform from Alibaba Group's AI technology gem to a member of the post-90s AI entrepreneurial army. It will be interesting to see how the story will develop next.

The mainstay of Ali Tongyi's large model

Now looking at Zhou Chang's LinkedIn account, I find that it still shows that he is a senior algorithm expert in Alibaba.

According to the information, Zhou Chang joined Alibaba in July 2017 and served as a senior algorithm expert in the Intelligent Computing Laboratory of the Damo Academy. During his time at joining Alibaba, Zhou Chang's academic research has yielded fruitful results, with his papers being included in major summits and Google Scholar being cited 8,000+.

At the same time, Zhou Chang played an important role in a number of projects during his time at the Damo Academy, such as the development of the ultra-large-scale commodity chart signature algorithm APP, the user representation framework ATRank, and the vector recall algorithm CLRec series based on self-supervised contrastive learning.

After joining Alibaba, Zhou Chang also witnessed the development of Alibaba's large model and was an important promoter in its development process.

From 2020 to 2021, he led the team to design and implement the ultra-large-scale multimodal pre-training model M6, making breakthroughs in the number of parameters and low-carbon training mode. According to Alibaba Cloud at the time, the M6 model has a parameter scale of up to 100 billion, making it the largest model in the history of multimodal pre-training. Under the push of Alibaba Cloud on the release of the M6 model, Zhou Chang once left a message: "Welcome all talents to join our team to continue this work!" "

To some extent, this M6 model is the predecessor of the Ali Tongyi model that the public is familiar with today.

By this year, the domestic AI industry has passed the embryonic stage, and the basic large model pattern has been formed. Startups such as Zhipu AI, Minimax, Dark Side of the Moon, and Baichuan Intelligence have emerged. Tencent, Baidu, ByteDance and other large manufacturers have also entered the game, of course, Alibaba is naturally indispensable.

The Ali Tongyi model, with Zhou Chang M6 as its predecessor, will be launched as early as April 2023, and users are invited to test the experience. As the "Ali version of GPT", Tongyi Qianwen has functions such as multi-round dialogue, copywriting, logical reasoning, multi-modal understanding, and multi-language support.

Four days after the launch of "Tongyi Qianwen", Daniel Zhang announced that all Alibaba products will be connected to Tongyi Qianwen in the future. In addition, Alibaba Cloud also said that it will open the capabilities of Tongyi Qianwen to the public to help enterprises build their own large models.

By the end of 2023, "Qwen-72B" has achieved remarkable results in the latest open-source large model ranking released by ugging Face, ranking first among all pre-trained models with a comprehensive score of 73.6. The performance of Tongyi Qianwen surpassed Llama 2-70B and many commercial closed-source models, and became the strongest open-source model in the industry.

It is reported that this list is recognized as the most authoritative list in the field of large models, bringing together hundreds of open source large models from all over the world, and conducting comprehensive evaluation through six evaluation dimensions, including reading comprehension, logical reasoning, mathematical calculation, factual question and answer, etc. It is in this high-profile ranking that Alibaba highlights the group's technological strength in the field of artificial intelligence, and also shows the leading position of Chinese technology companies in the research and development of large models.

On May 9, 2024, the first anniversary of the release of the Tongyi large model, Alibaba Cloud officially released the Tongyi Qianwen 2.5, and the model performance has fully caught up with GPT-4 Turbo and become the strongest Chinese large model on the surface. At the same time, Tongyi Qianwen's 110 billion parameter open source model has achieved the best results in multiple benchmarks, surpassing Llama-3-70B and becoming the most powerful model in the global open source field.

In this era of large models rising together and involuting each other, Alibaba has handed over an excellent answer sheet by relying on Tongyi Qianwen. Behind this answer sheet, there is no doubt that Zhou Chang is an important witness and witness.

In 2022, the Chinese Institute of Electronics announced the list of "2021 Science and Technology Awards of the Chinese Institute of Electronics", among which the "ultra-large-scale high-performance graph neural network computing platform and its application" completed by Zhejiang University, Damo Academy, and Alibaba Cloud won the first prize of scientific and technological progress. Interestingly, Yang Hongxia has already left her job in 2022 and started a business, and now Zhou Chang will also leave her job to start a business.

High-yielding bulls go to the sea to start AI entrepreneurship

A similar scene is continuing to be played out in the domestic AI community. The launch of ChatGPT in 2022 has made AIGC popular, attracting countless outstanding entrepreneurs to target this track for entrepreneurship. Many backbone players like Zhou Chang have been exposed to solo entrepreneurship, and the entrepreneurial boom in the AI industry is in full swing.

Previously, a financial reporter from Rongzhong interviewed the AI marketing company Ultron Lightyear. Up to now, the company has completed three rounds of financing, and the investment institutions include Wanwu Capital, Zhen Fund, Cathay Capital, Meitu, etc.

According to the news revealed by the VP of the Marketing Department in the interview, in 2023, the contract amount signed by Altron Lightyear customers will increase by more than 300%, exceeding 100 million yuan. At the same time, Altron Lightyear has successfully served more than 100 first-tier brand groups and e-commerce platforms such as Alibaba, PepsiCo, L'Oreal, Unilever and Johnson, and many customers are still in the stage of cooperation or contract renewal, and continue to expand the depth and scale of cooperation. For an AI startup, such a performance is not easy.

It is worth mentioning that the two founders of Ultron Lightyear, Xu Zhe and Yang Hai, were previously staff members of Alibaba. The two met in Alibaba, and left Alibaba one after another to go their separate ways, breaking out of the world in different fields. For a part-time worker, the two of them were already standing at the pinnacle of their respective careers at that time.

After that, the two investigated the No. 1 position of 100 companies, explored the pain points of the industry, and finally decided to devote themselves to AI entrepreneurship and established the current Altron Lightyear. Xu Zhe and Yang Hai have also become the perfect transition from executives of large factories to AI entrepreneurs.

And there are many AI entrepreneurs of the same type like Zhou Chang, Xu Zhe, and Yang Hai.

Not long ago, on July 16, Li Ping, vice president of Baidu Research Institute and director of Baidu Cognitive Computing Laboratory, resigned to found Ai Infra company vecml and served as CEO. According to Baidu Encyclopedia, Li Ping graduated from Stanford University in 2007 with a Ph.D. in statistics, a master's degree in computer science and a master's degree in electrical engineering. After graduating with a Ph.D., Li Ping taught at Cornell University and Rutgers University, and became a tenured professor in the Department of Computer Science and Statistics in 2013. At the end of 2017, Li Ping joined Baidu Meiyan as a Distinguished Scientist (T10) and became the head of the Big Data Laboratory of Baidu Research Institute. In 2018, Li Ping was promoted to T11, and the Big Data Lab was renamed the Cognitive Computing Lab.

According to media reports, during his time at Baidu, Li Ping and his team published a number of international top conference papers in the fields of deep learning algorithms, knowledge graph and information extraction, graph-based approximate neighbor retrieval and maximum inner product retrieval, tree model, hash algorithm and approximate neighbor retrieval (ANN) algorithm, basic theory and method of machine learning, recommendation and LTR technology, natural language, big data, and data mining. Baidu Maps and other teams have carried out in-depth cooperation to apply the research results to various products, produce remarkable results, and can develop new research directions and topics from practical problems. The company he founded, vecml, is committed to building a lower-cost, more secure, and more accurate AI infrastructure for the next generation of large model applications.

Also not long ago, World Labs, founded by Li Feifei, known as the "godmother of AI", also had new news, which has exceeded a valuation of $1 billion and has become a new unicorn in less than 4 months; and Qi Yuan, the former vice president of Ant Group, whose entrepreneurial direction was also revealed, and he founded a large model company Infinite Light Years in parallel with production, education and research, and was exposed to being a "credible large model"; Li Mu, who resigned as Amazon's chief scientist, created Boson.ai and has now launched the first large-scale model results.

Such similar news is not only happening in China, like foreign technology giants such as Google, Microsoft, and Open AI, which have internal engineers going to the sea to start AI startups. For example, Kapasi, the head of the AI department of Open AI and Tesla, recently announced his entrepreneurship and founded a new AI-native school, Eureka Labs.

In today's fast-paced era of technology, artificial intelligence (AI) has become a key force driving innovation in all walks of life. As AI technology matures, the demand for AI professionals surges, providing a platform for executives to showcase their talents and realize their personal value. In large factories, they may have reached the peak of working as workers, but entrepreneurship can provide them with a bigger stage.

Compared with direct entrepreneurs who have no experience, the experience of being an executive in a large factory has allowed them to accumulate a certain amount of experience in both technology and business operations, and they also have a certain understanding of the "way", which also makes them more comfortable on the road of entrepreneurship.

With the addition of these bigwigs, the ever-changing AI world has also added some new possibilities.

A new era of AI industry is coming

The AI boom is unprecedented. At the recent China Internet Conference, a group of dazzling data was announced.

At present, there are 421 national-level intelligent manufacturing demonstration factories, more than 10,000 provincial-level digital workshops and smart factories, and more than 4,500 artificial intelligence enterprises in mainland China...... According to data released by the China Academy of Information and Communications Technology, in the field of artificial intelligence, the application process of the mainland's artificial intelligence industry will continue to advance in 2023, and the scale of the core industry will reach 578.4 billion yuan. In terms of industry investment and financing, 815 investment and financing events have occurred in China's AI industry in 2023, with a total financing of 263.1 billion yuan, an increase of 51% over last year.

Artificial intelligence, which continues to explode, has also triggered the next step of thinking among industry experts. Over the past few decades, AI technology has moved from esoteric theories in the lab to widespread real-world applications. This transformation is not only a technological advancement, but also a complete revolution in business models, ways of working, and even lifestyles. With the maturity of machine learning, deep learning and other technologies, it has witnessed how artificial intelligence has found its value in many industries such as healthcare, finance, education, and manufacturing.

Nowadays, artificial intelligence is making great strides towards "application", which is about to give birth to a new round of AI industry transformation.

Previously, Luo Wenyong, executive dean of Moore Threads Industry College and general manager of the investment and financing department, said at the Rongzhong 2024 New Quality Productivity Innovation Enterprise Summit: "AI has now become the largest application scenario, just like the mobile Internet 20 years ago, which can empower thousands of industries. ”

For example, it is a well-known fact that the real economy is the foundation of the country, and the manufacturing industry is the key to the real economy. Previously, the state has issued policies to emphasize improving the development quality of the manufacturing industry, driving development with innovation, accelerating the empowerment of digital technologies such as artificial intelligence, comprehensively promoting industrial upgrading, giving birth to new models and new functions, and realizing the rapid growth of the manufacturing industry. Six departments, including the Ministry of Science and Technology, issued the "Guiding Opinions on Accelerating Scenario Innovation and Promoting High-quality Economic Development with High-level Application of Artificial Intelligence" to encourage in-depth exploration of artificial intelligence technology application scenarios in key industries such as manufacturing, and promote the high-end and efficient development of the intelligent economy; On the other hand, the "14th Five-Year Plan" for the development of intelligent manufacturing and other policies emphasize that it is necessary to strengthen the research of key core technologies, research and development of artificial intelligence, 5G, big data, edge computing and other applicable technologies in the industrial field, and enhance the new momentum of integrated development.

SenseTime, one of the leading AI companies in China, has also provided many internal examples before. For example, in response to the problem that "it is difficult for the elderly to use financial apps", SenseTime and Bank of Shanghai have jointly developed a financial model, and the back-end financial model has learned 4,000 large model products to answer various complex questions raised by the elderly. The front-end interaction has made a 4K ultra-realistic digital human, just like the offline lobby manager and financial consultant, giving the elderly very good professional services and chat dialogues, which has been well received by the market.

It's an indisputable truth that technology is the source of innovation in society. Mao Jiming, partner and vice president of engineering of Jijia Technology, also made it clear: "As a technology, artificial intelligence must complete the implementation of application scenarios and provide support for industrial upgrading in all walks of life in order to play its due value." At the same time, in the process of combining AI technology with scenarios, domain data and experience will also be sunk, feeding back the AI technology base and enabling it to evolve rapidly. Technology and scenario application, the two will promote each other, produce a flywheel effect, which is also one of the important means for enterprises to have sustainable competitiveness. ”

The development of AI has entered a new era, and at the same time, the elites such as Zhou Chang have collectively poured into AI entrepreneurship, what new stories will they tell under the new development logic?