wu yongming of alibaba: the threshold for advanced ai models will reach tens of billions of dollars, and the robotics industry will usher in huge changes
2024-09-19
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
alibaba's wu yongming believes that the investment threshold for advanced ai models may reach tens of billions of dollars.
on september 19, wu yongming, ceo of alibaba group and chairman and ceo of alibaba cloud intelligence group, delivered a keynote speech at the 2024 yunqi conference.
on september 19, the 2024 yunqi conference opened in yunqi town, hangzhou. wu yongming, ceo of alibaba group and chairman and ceo of alibaba cloud intelligence group, said in his keynote speech: "in the past 22 months, ai has developed faster than any other period in history, but we are still in the early stages of the agi revolution. the biggest imagination of generative ai is by no means to create one or two new super apps on mobile phone screens, but to take over the digital world and change the physical world."
this is also the first time that wu yongming has given a speech at the yunqi conference since he took office as ceo of alibaba group in september 2023. wu yongming believes that the investment threshold for the next stage of advanced ai models will reach billions or even tens of billions of dollars: "to achieve true agi, the next generation of models will need to have a larger, more generalized knowledge system, and will also have more complex and multi-level logical reasoning capabilities. the investment threshold for the global advanced model competition will reach billions or even tens of billions of dollars."
regarding the view that the industry is waiting for the emergence of ai super applications, wu yongming pointed out that the imagination of ai will not be limited to this. he took the internet wave in the past 30 years as an example: "the essence of the internet wave is connection. the internet connects people, information, business and factories. through connection, it improves the world's collaborative efficiency, creates huge value, and changes people's lifestyles. but generative ai creates new value through the supply of productivity, thereby creating greater intrinsic value for the world, that is, it generally improves the productivity level of the entire world. this kind of value creation may be ten or dozens of times the value of mobile internet connections."
at this year's yunqi conference, big models, computing power, autonomous driving and robots became the four key words. wu yongming pointed out that after the automotive industry, robots will be the next industry to undergo a huge change: "in the future, all movable objects will become intelligent robots. they can be robot arms in factories, cranes on construction sites, porters in warehouses, firefighters at fire scenes, including pet dogs, nannies, assistants in the family... now every urban family has one or two cars, and in the future every family may have two or three robots to help people improve the efficiency of their lives."
in addition, alibaba cloud has observed that in the new computing power market, more than 50% of the new demand is driven by ai, and ai computing power demand has become the mainstream. wu yongming said that generative ai will bring about fundamental changes in computing architecture: "in the future, almost all software and hardware will have reasoning capabilities, and their computing cores will become a computing model with gpu ai computing power as the main and cpu traditional computing as the auxiliary."
at this year's yunqi conference, alibaba cloud also launched a series of updates in terms of large models, including upgrading the tongyi basic model, releasing the open source model qwen2.5 series, and launching more than 100 full-modal models such as language, audio, and vision. it is reported that the cumulative downloads of tongyi open source models have exceeded 40 million, and the total number of tongyi native models and derivative models has exceeded 50,000, making it a world-class model group second only to the american llama model series. at the same time, alibaba cloud announced that the three main models of tongyi qianwen have been significantly reduced in price again, with the highest reduction of 85%, and the price of one million tokens is as low as 0.3 yuan.
alibaba cloud chief technology officer zhou jingren said that alibaba cloud is setting a new standard for ai infrastructure and investing in upgrading ai infrastructure to turn data centers into supercomputers. in addition, zhou jingren specifically mentioned that in the past two years, the size of the model has increased thousands of times, but the computing cost of the model is continuing to decline, and the cost of using the model for enterprises is also getting lower and lower: "this is the technological dividend brought about by the comprehensive innovation of ai infrastructure. we will continue to invest in the construction of advanced ai infrastructure to accelerate the entry of large models into thousands of industries."
in an interview with the media, including a reporter from the paper, after the meeting, zhou jingren added that alibaba cloud's price cuts were mainly achieved through technology optimization: "today, not only are the models themselves iterating rapidly, but the model's reasoning architecture and system optimization are also constantly improving. through the efforts of the past few months, we have been able to further improve the efficiency of all aspects of model reasoning... we recognize that today we need to effectively promote the development of the entire ai industry. only in this way can we more effectively deploy various ai applications."
zhou jingren also said that he does not think this is a price war for large models, just as people "could not imagine 20 years ago that 200 yuan could be used for dozens of gb of traffic today." the current price of large models is not low enough, "and it is still too expensive for future large-scale applications."
as for alibaba cloud's layout of both open source and closed source models, zhou jingren said that on the one hand, when doing open source, we hope that the ecosystem can develop, and on the other hand, when doing closed source, we hope to serve enterprises more effectively: "since last year, alibaba cloud has been unwavering in doing open source. this is very closely related to our original intention of doing cloud computing. we hope that more people can use the capabilities of open source."
hu hanyan, reporter of the paper
(this article is from the paper. for more original information, please download the "the paper" app)
