Shulou(Shulou.com)12/24 Report--

This year, ChatGPT was born, giving a shot in the arm to the commercialization of AI. When AI as an advanced productive force has gradually penetrated into thousands of industries, it may become the next "hydropower and coal" in the future. Focusing on the moment, we can no longer ignore the capacity dividend that AI is about to release in thousands of forms of consumption.

Under the double-wheel drive of "culture + science and technology", the new era of national tide is coming. Recently, Liangzhu 2023 Science and Technology National Tide Industry Conference was held. Under the guidance of the China Association for the Promotion of Brand Construction and hosted by the people's Government of Yuhang District of Hangzhou, the conference invited 100 guests, such as experts and scholars, entrepreneurs, investors and industrial service providers, to create a national tide of science and technology. Huang Weipeng, co-founder of Tuoyuan Wisdom and general manager of Yuanfen platform, was invited to attend the meeting and revolved around "AI enables large consumption of the whole industry chain, how to reduce the cost and increase efficiency?" The theme has a round-table conversation with a number of artificial intelligence companies to share the source of digital split. Yuanfen is exploring the experience and challenges of AI and consumer applications. The round-table forum is hosted by Shen Yuebo, vice president of Yinxinggu Capital Investment.

Liangzhu Science and Technology National Tide Conference "consumption + AI" Forum

Huang Weipeng, General Manager of Yuan Fen platform

The following is a transcript of the speech, hoping to bring you more inspiration and thinking.

Shen Yuebo: the large model is far away from large-scale commercial use, what other hurdles do we need to break through?

Huang Weipeng: because Tuoyuan Wisdom is committed to building an enterprise-level vertical scene AGI application platform, we have done a lot of work at the large model level, working with Huawei's Noah's Ark Lab to build a Wukong model since 2017. Recently, we have helped some enterprises to build a large vertical scene model. Based on these practices, I think there are still four stuck points before large-scale commercial use.

First, cost. The cost of the large model at the training and reasoning level is very high. Take ChatGPT as an example, the cost of one training is at the level of tens of millions of dollars, and each update will emphasize reducing the cost of token.

Second, controllability. In terms of controllability and accuracy of content generation, hallucinations are tolerable in pan-entertainment consumption scenarios, but if there is no controllability, it is difficult to use them on a large scale in serious scenes.

Third, product form. The large model is still in a very primary state, the scene is still very single, limited by the product form, the value generated is relatively limited, can only be used as an assistant in production or consumption, and requires manpower for secondary processing.

Fourth, security. How to prevent the data of large models from being contaminated and stolen, and how to regulate fraudulent content, is still a great challenge.

Shen Yuebo: AI technology has entered a new intelligent stage, and the technology iteration is very unstable. How can new technology enterprises survive the painful period of scientific and technological development?

Huang Weipeng: I have been engaged in related work in the field of science and technology for nearly ten years since I started my business in the university. There are three main points about how to get through the painful period of scientific and technological change.

First, when new technologies become products, early products are often not mature enough, resulting in unclear application scenarios. In this regard, we can interview with approachable enterprises, communicate with customers who go deep into the front line, excavate scenes, design plans, verify the feasibility of products, and find out whether customers are really willing to use money to vote.

Second, because the product is not mature and the scene is not so clear, it is difficult to find suitable customers, that is, angel customers. In this regard, it is necessary to tap the customers who are most suitable for their own enterprises, circle them into the enterprise system through a certain mechanism, let customers become the promoters of early technology and products, and build a reliable and loyal customer pool as soon as possible.

Third, because there are no continuous customers, there is no reliable and sustainable income during the labor pains. It's important to be alive, to be a lean team during labor pains, not to expand too much, to avoid breaking the capital chain before achieving sustainable revenue, and to look extensively at what's going on in the market.

Shen Yuebo: what is the space and trend of AI technology to empower the consumer industry chain?

Huang Weipeng: the customer demand in the consumer field reveals the trend of future development.

In the creative design, after the large model has a certain degree of randomness, it can provide inspiration, and can also be applied to the simulation and prediction of new material combination production.

In the production process, the biggest trend is embodied intelligence, that is, robots. Robots can replace a large number of people working in factories, and can increase efficiency to a state of "terror".

In the consumer end experience, you can try it on and try it out through applications such as AR.

In the marketing promotion link, the generation of marketing content, pictures and videos of goods, and the application of digital people in live broadcast and short video transmission is also an obvious trend. There is also a bigger direction is to truly implement personalized recommendations. The emergence of large models breaks the limitations of content production. For example, when placing advertisements before, you need to prepare different advertising content for different segments of the population. If you use AIGC, you can push the tags and data of the crowd to the large model, which completes the production, push and delivery of the content, which will bring about a truly personalized era of thousands of people and thousands of faces.

In the process of service and after-sales, agents' understanding of emotion will be enhanced based on the progress at the bottom of the big model. When digital humans and robots have the ability to understand, identify, interpret emotions and resonate, they can truly achieve humanized service.

Huang Weipeng, General Manager of Yuan Fen platform (left)

Shen Yuebo, Vice President of Yinxinggu Capital Investment (right)

Shen Yuebo: 2023 is an extraordinary year. A new wave has arrived. AI represents new technology, consumption represents everything we come into contact with every day, and AI technology can do the whole chain of consumer industry chain again.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.