Large model and generative artificial intelligence have been ＂out of the circle＂ from the technical field! AI’s ＂golden decade＂ is just the beginning.
Large-scale model and generative artificial intelligence (AIGC) have gone out of the technical field and become one of the hottest topics today. On the one hand, the key reason lies in the great opportunities it may bring to global development, and on the other hand, its powerful and disturbing ability. At the World Artificial Intelligence Conference that opened yesterday, global industry leaders focused on big models, talked about innovation and trends, and also talked about how to step on the brakes.
AI should not only "read thousands of books", but also "walk Wan Li Road".
Many heavyweight guests unanimously believe that starting from the breakthrough of the big model, we will usher in a "golden decade" led by AI (artificial intelligence). Now, it’s just the beginning of this great era. In order to successfully cut into this cadenza movement, AI developers also need to carefully polish the large model so that it can be accurately adapted to thousands of industries.
Hu Houkun, the rotating chairman of Huawei, borrowed the requirements of traditional culture for personal cultivation —— reading thousands of books and walking Wan Li Road, and described the development direction of large model technology in his eyes. The so-called "reading thousands of books" is to let AI learn massive data and complete "basic knowledge accumulation", but this is not enough to fully empower and change traditional industries. Therefore, the big model should "follow the Wan Li road", that is, deeply integrate AI with the real economy and forge a subdivided industry model. For example, Huawei has used its own software and hardware platform to connect large models with hundreds of industries. Among them, the AI model trained by using meteorological big data in the past decades can give a forecast of the global weather in the coming week within a few seconds.
In fact, regarding the development of big models, there are two schools of views in the industry. One is the "universal" school, which means that the AIGC fever triggered by ChatGPT is likely to make people find a breakthrough towards universal artificial intelligence; The other school believes that the vitality of AI lies in landing, and it is too early to talk about general AI.
Most of the guests who appeared on the stage yesterday held the latter view. The most determined one is Yang Likun, winner of Turing Award and one of the "Three Masters of Deep Learning". He simply said that at present, there is no universal AI in the world, "every intelligence is dedicated."
When it comes to how to connect AI with the industry, Tencent and Baidu can be said that great minds think alike. Tang Daosheng, senior executive vice president of Tencent Group, believes that many AI models seem to be omnipotent, but in the face of various industries, they can only solve 80% of the problems, and the remaining 20% are easy to make mistakes, which leads to confusion. To this end, AI companies need to create exclusive "small models" in addition to the general big model according to the business characteristics of customers. According to this model, Tencent planned the so-called MasS ("model as a service") business, which is to help customers quickly complete the customized development of "small model".
Baidu’s big model business is also similar in design. Wang Haifeng, chief technology officer, revealed that enterprises need industry models, but they can’t solve the problems of cost and difficulty. To this end, Baidu hopes to learn from the chip foundry model, and provide automated and standardized AI model customization services for various industries through its own super computing power, super-large data and hundreds of billions of models.
Machine assistants, digital avatars, and new technological breakthroughs are concentrated.
Too much talk about business models will inevitably be a bit boring. In contrast, hard-core and cutting-edge technical topics may attract more attention. Yesterday, a number of industry leaders and cutting-edge players took turns to stage, showing the most proud and optimistic AI technology.
Yao Qizhi, winner of Turing Prize and president of Shanghai Zhizhi Research Institute, introduced the latest achievements of one of his young teachers. It is the latest reinforcement learning algorithm, and compared with the traditional technology, the learning speed of this algorithm has been improved by hundreds of times. In the past, it usually took several months to train an intelligent robot, but it can be taught in a few hours through new algorithms.
As an important investor and partner of OpenAI (the developer of ChatGPT), Microsoft is working hard to make the big model more deeply bound with its Windows Windows system, Office suite and cloud computing services. Hou Yang, senior vice president of the world and chairman and CEO of Greater China, revealed that at present, Microsoft has launched 50 AIGC pilot services, the most important of which is CoPilot intelligent assistant, which can quickly generate PPT speeches instead of users or write software at 150% speed. In addition, another revolutionary technology released by Microsoft, the ——AI Plug-IN system, will also become an interface between the big model and various specific application scenarios. It is expected that the total number of plug-ins in ChatGPT will exceed 1,000 when it is officially released soon.
He Xiaodong, vice president of JD.COM Group and president of JD.COM Exploration Research Institute, first played a video of him introducing the progress of AI research and development on JD.COM Live Channel. Later, he "revealed" the truth. In fact, he and the host in this film are not real people, but digital people generated by AI. At present, to generate such a "digital avatar" that can be confused with the real thing, the threshold is very low: just take a picture of the party for 5 minutes, AI can model in real time and complete the perfect reproduction from appearance and expression to accent and tone.
In addition to introducing the products and technologies that have come out, the future development direction and breakthrough of the big model was also a hot topic yesterday.
Yang Zhilin, an assistant professor at the Institute of Interdisciplinary Information of Tsinghua University, put forward a very interesting direction: Can AI have "collective memory" like people? He said that although the big model can remember the context to a certain extent when talking with people, once the chat is over, the context will no longer exist. Can AI quickly enter the role every time through new technologies? He thinks this application is very promising.
Pan Xingang, an assistant professor at the School of Computer Science and Engineering of Nanyang Technological University, is optimistic about the field of video generation. He said that the current big model is quite powerful when creating pictures; In the future, we should guide it to further upgrade and provide more help in the generation of dynamic pictures including three-dimensional content.
In addition, whether enough and good data can be used to "feed" AI is directly related to the ability and performance of the big model. However, at present, "data shortcomings" are still widespread. Speaking of the way to crack, Sun Yuanhao, founder and CEO of Xinghuan Technology, said that what AI lacks most at present is actually knowledge data, that is, how to interpret a news event and how to solve a mathematical problem. He believes that the technical community needs to develop corresponding tools and databases at present, and extract human methodology with natural language. Only in this way can we really make good use of large models.
AI needs responsible governance, but it is overly worried or unnecessary.
Since the AIGC and the big model show the powerful intelligence beyond expectations, the outside world is worried that AI will defeat the human brain and challenge mankind. Yesterday, the voice that the reporter heard in the venue about how to supervise AI seemed to be much more frequent than before. Of course, the big coffee makers in the industry are not as anxious about the current AI as ordinary people. They are more based on rationality, focusing on the long-term, and giving suggestions on the governance of AI.
Yang Likun believes that it is necessary to set goals for AI, define nature and build safety barriers. For example, AI can be required to be honest, obey and not hurt people; Once it is violated, AI should get emotional feedback similar to fear or pain. Because the realization of technology is very challenging, Yang Likun suggested starting with low-level AI, such as putting AI equivalent to mouse intelligence into a "technical sandbox" first, studying how to train it into a good mouse, and then gradually improving AI ability until it is close to human intelligence.
Hou Yang said that in Microsoft’s view, to build a responsible AI, we need to follow six principles, namely fairness, reliability, security, transparency, privacy and tolerance. "We advocate global technology companies to form a consensus through cooperation to ensure that AI technology benefits all mankind."
Of course, there are also some AI industry leaders who have reservations about the concerns of the public and peers. The hottest drawing AI in the world
David Holtz, the founder of MidJourney, said that the technological progress of AI so far is to enhance the capabilities of people themselves, and there is no strong artificial intelligence era coming in the near future. Holtz said that AI is not an individual, it is just a kind of "entanglement" or "flow" of wisdom. For example, painting AI is by no means an artist who works side by side with you, but an extension of your thoughts. The entrepreneurial star stressed: "If you think too much, you may miss many opportunities."
If you think Holtz’s expression is too mysterious, then the way Tang Xiaoou, a professor at the Chinese University of Hong Kong, used to comfort the "AI phobia" is more grounded. The expert who has led the team to achieve a lot of world-class achievements in the AI field likes listening to a cross talk before going to bed. Every time at this moment, he would listen and think, "How can a machine surpass such an interesting soul? I don’t believe it. "
Source: Zhang Yi
Editor: Fan Jing
* The exclusive manuscript of Wenhui, please indicate the source.
Reporting/feedback