news

currently, the "2024 yabuli entrepreneurs 20th summer annual meeting" is being held in guangzhou. as the hottest technology track in the past two years, big models are naturally the focus of discussion among entrepreneurs at this summer annual meeting.

at present, big models are being applied in many industries such as finance, medical care, and education, becoming a key force in promoting industrial upgrading. however, on the other hand, the implementation of big models still faces multiple challenges such as "difficulty in importing enterprise data, lack of security, and low accuracy". on the evening of august 30, zhang wenzhong, founder of wumart group, shi qinghua, vice president of baidu, fu sheng, chairman and ceo of cheetah mobile, and other industry leaders attended the annual meeting and delivered speeches, answering the core questions of the big model industry such as "whether to be anxious" and "how to make money".

image source: provided by the organizer

“opportunities definitely outweigh anxiety”

since the emergence of big models, the "ai anxiety" in all walks of life seems to be getting worse. many companies and individuals are anxious about the application of big models; big model manufacturers themselves are even more "anxious" because the commercialization problem remains unsolved.

a reporter from the "daily economic news" learned that at the opening forum of the "2024 yabuli entrepreneurs' 20th summer annual meeting", representatives of major model manufacturers and large model application companies all stated that in fact they were not as anxious as the outside world said.

liu cong, vice president of iflytek and director of the research institute, said: "maybe people are not as anxious as you think. what we see is that opportunities are definitely greater than anxiety. open ai lost $5 billion last year, but it is still doing well and doesn't seem to be particularly anxious."

"when will people become anxious? anxiety only makes sense when you feel completely unsure about where the future lies and what to do next." zhang wen zhong said that it is a consensus that artificial intelligence is the future of mankind. what we need to do now is to fully embrace digitalization and intelligence. once we know where the future lies, we must do it unswervingly.

shi qinghua also emphasized: "to be honest, baidu is determined to invest in artificial intelligence, big models and their development prospects."

recently, former google ceo schmidt mentioned at an exchange at stanford university that china lags behind the united states by ten years in the field of ai.

in this regard, fu sheng believes that there is a gap between the top level of big models between china and the united states, but the gap is definitely not as big as it has been in ten years. qi ruifeng, co-founder of zero one everything, believes that in the era of big models, the united states was ahead of china in to c applications, but to date, china is ahead of the united states in to b applications.

how to make money?

as a user of big models, 58.com ceo yao jinbo said that the more big model manufacturers there are, the better. "it would be best if there were 50 of them, and the lower the price of the products they can provide, the better, and the higher the product performance, the better." he believes that there are still obstacles to applying big models on a larger scale today. the first is that they are not smart enough, and the second is that big models are still expensive. "it would be best if you all provide us with services for free or at a loss, which is what we would like to see most," said yao jinbo.

but yao jinbo is also willing to pay. "i especially hope that someone will come to me with a model (project) and i can immediately calculate the accounts. i will definitely be willing to pay. i will not pay according to the cost of the big model company, but according to the value created by the big model for the user, or the cost saved by the user. maybe that day will be the day when our big model manufacturers can really run." yao jinbo said.

shi qinghua responded to yao jinbo on the spot. he said that he was very happy to hear customers say that big models are indeed useful at the conference. this is what big model companies want to hear most, because customers are willing to pay for value, which is what big model manufacturers ultimately pursue. he also mentioned that training big models is inseparable from big data, big computing power and big models, which is not only time-consuming but also very costly. "for example, the training process of big models requires computing power. not to mention the construction of an intelligent computing center, the investment in building a wanka cluster alone is in the tens of billions."

liu cong said that for large model manufacturers, it is necessary to explore scenario needs and then cooperate deeply with customers, and to be industry-leading based on the development of core technologies. if these can be achieved, future profits will be natural. he emphasized that the current application of large models in the g-end and b-end represented by central state-owned enterprises is expanding from point to surface and is ready to go.

zhang wenzhong said that artificial intelligence can definitely create value, and consumers will pay for it. the combination of software and hardware is also a way for artificial intelligence technology to "make money." "those of you who are engaged in large models, go forward boldly, there will be people who will pay for it," zhang wenzhong said.

daily economic news