The hot AI "breakout" is not a battle alone

Original source: Harvard Business Review

Author: Zhu Dong

Image source: Generated by Unbounded AI‌

From 2023 to today, if you can’t find a common topic with others and don’t want to be embarrassed, then it’s definitely right to talk about artificial intelligence.

The development of AI in 2023 is obvious to all. The topic of concern among enterprises is inseparable from the “big model”, and different fields have begun to think and try the “surprises” that AI can bring to the industry.

Recently, the 2023 World Artificial Intelligence Conference was held in Shanghai. At the conference, we conducted dialogues with experts, entrepreneurs, and managers in the field of artificial intelligence, and discussed the opportunities and challenges that AI brings to enterprises and industries in the digital business era.

The next surprise may be years away

Artificial intelligence has been around for more than 60 years.

If we briefly summarize the leapfrog development of artificial intelligence in the past five years, Zhang Yaqin, academician of the Chinese Academy of Engineering and dean of the Intelligent Industry Research Institute (AIR) of Tsinghua University, believes that the real big breakthrough is actually deep learning. Especially in the past three years, massive data, large models, and algorithm computing power have all made breakthroughs from quantitative changes to qualitative changes.

**The first is the leap from perception to cognition. **In the past, AI was more about speech, image, character recognition, etc. in terms of human perception. In the past two years, it was more about cognition, reasoning and thinking about our understanding of language and the semantic understanding of videos.

**The second is from artificial intelligence with dedicated algorithms to artificial general intelligence (AGI). **In the past, it included voice, image, automatic driving, protein analysis, etc., and more relied on specific algorithms, specific models or specific data sets, but now GPT4 at least provides us with a channel from specific to general artificial intelligence.

**The third is from discriminative or analytical AI to generative AI. ** For the first time, human beings can create and generate new things, such as text, language models, generating pictures, generating protein structures, etc.

However, in Zhang Yaqin’s eyes, no matter how good generative AI is, it is still a tool. True creativity, creativity, imagination, and inspiration still depend on scientists, experts, and people themselves, which cannot be replaced by AI. It’s just that many of these technical problems can be accelerated by using AI.

Regarding the progress of artificial intelligence, Professor Deng Zhongliang, academician of the International Eurasian Academy of Sciences and professor of Beijing University of Posts and Telecommunications, also expressed emotion, ** “30 years ago, my teacher said in an artificial intelligence class, ‘Robots are good at everything, but they can’t talk about love’. Now, robots can even be human emotional consultants.”**

Talking about the progress of domestic AI technology and the differences between European and American countries, Deng Zhongliang said that from the perspective of the country’s strategic layout, from the scientific research of AI and the early research and exploration of key problems, the development of AI in China has gone very fast. And the supporting capacity of basic conditions, including the manufacturing capacity of basic components such as chips, is also accelerating. The follow-up is the continuous research and gradual improvement of the big data model.

Deng Zhongliang believes that in the field of artificial intelligence, there will be a big leap in China, and whether it will reach the world-leading level in some aspects requires joint efforts. At present, in terms of algorithm computing power, our level is similar to that of European and American countries, but there is still a lot of room for development in terms of the supporting capacity of some infrastructure.

Moreover, there is not much risk for technological progress to promote industrial development. The potential risks come from the application level and corresponding laws and regulations. For example, in the case of AI automatic driving, once an accident occurs, how should the rights and responsibilities be divided? When a robot has a “temper”, who is responsible for an accident? How to prevent and control? There will be more and more similar things in the future.

In addition, Deng Zhongliang bluntly said that he is not worried about the employment pressure that AI will bring to certain industries that many people worry about. Humans are the producers and creators of AI, and artificial intelligence serves human society, not replaces human beings. **Currently, ChatGPT, GPT4 or other large AI models have set off a new climax of artificial intelligence, and this climax should continue to develop for a period of time. AI’s next “surprise” may be years away. **

“AI may also form an independent industry in the future, but at present it is more of an integration with other industries.” Deng Zhongliang said.

AI research and development and investment, rolling alone is not as good as “rolling” together

Indeed, AI is gradually being applied to enterprises and industries in different fields.

Chris Young, executive vice president of business development at Microsoft, once used the “Cambrian Explosion” to describe the growth rate of companies engaged in AI services, as well as the changes in the scale of companies that are trying or heavily investing in AI research and applications. After all, AI has not only become a “hacker-level” means to improve productivity and operational capabilities, but also opens up new practical paths and creates new business opportunities.

Guo Fan, director of the “Wandering Earth” series of films and vice chairman of the Beijing Film Association, said when talking about AI at the Artificial Intelligence Conference that when the second part of “The Wandering Earth” was filmed, the number of on-site personnel had reached more than 2,000, and the overall team size was close to 30,000, involving a huge process management system. “Perhaps, in the future, more AI technologies will be used to assist tens of thousands or even millions of people to collaborate at the same time. I believe that soon, the intervention of artificial intelligence will lead the entire film industrialization to 3.0.” How to create by then? How to shoot? Will there be any new changes in film post-production or even viewing mode? are full of many unknowns.

Gu Liang, Chief Technology Officer of Jingtai Technology, called the integration of biomedicine and AI as “the right time for birth”. If it is too early, it will be too late.

Gu Liang explained the reasons for this judgment to us: First, ChatGPT has proved the feasibility of large AI models, proving that with enough data samples and sufficient computing power, an intelligent model can be obtained to assist enterprises and industries in reasoning, demonstration, verification, and exploration.

Second, in the field of life sciences, the application of AI technology caters to the needs of drug development and innovation as well as the needs of human health. With the goal of extending the length and breadth of human life, it has become feasible for AI to empower drug research and development to reduce costs and increase efficiency. The technology platform of AI+ robots has been verified in the field of life sciences, and it is also showing its strength in new materials, chemical and other industries, and constantly expanding new scenarios and new fields.

At present, platform-based technology companies, including Jingtai Technology, have already had successful cases. When Pfizer developed the oral solid new crown drug PAXLOVID, the Pfizer team cooperated with Jingtai Technology, using AI prediction algorithm combined with experimental verification, which greatly shortened the research and development time. It took only six weeks to confirm the advantageous crystal form of the candidate drug, which was used for subsequent development and production, and accelerated drug listing.

However, Gu Liang also added that the large-scale application of AI to enterprises and businesses requires the cooperation of the entire industry chain, collaborative breakthroughs, rapid trial and error, and constant search for more effective methods. The ultimate goal is to lower the cost of research and development and benefit more people. In this process, the pioneers and demonstrators in the industry, and the “players” who are willing to provide platform tool empowerment, cooperate and explore with more companies. This model may be more sustainable.

At the same time, the innovation and transformation of the industry in the future also requires new technologies to enter the “Internet” era from the “Pay Yellow Pages” era. **That is to say, the integration of automation and intelligence requires sufficient samples and data based on innovative research and development, so that the AI model can iteratively operate and become more and more intelligent after being continuously verified. Therefore, the independent research and development of AI by state-owned enterprises and investment will inevitably cause significant waste. If the industrial chain can be connected in series, and various supply chain links can cooperate with each other to achieve a win-win or even win-win situation, it will be very viable. **

He Tao, the global vice president of Nvidia, also has a deep understanding of this point. He mentioned that from the United States to China, many companies and institutions have invested a lot of money and energy in building basic large-scale models and cloud services. The improvement of AI computing power is the direction of the entire industry, but the construction of computing power is not a one-day effort. It is wrong to blindly expand computing power. The biggest first-mover advantage is to drive the development of the entire ecosystem. Wu Yunsheng, vice president of Tencent Cloud, head of Tencent Cloud Intelligence, and head of Youtu Lab, also believes that the development of AI large-scale model technology and industrial exploration are inseparable from industrial chain collaboration and ecological co-construction.

And the synergy of ecology, jointly promote the innovation and implementation of AI in the industrial field, the end result is the progress and development of human beings.