Discussing the Industrial Value of AI Large Models: Surprises and Long-Term Considerations | WISE2024 King of Business

The environment is constantly changing, and the times are always evolving. The "Business Kings" follow the trend of the times, insist on creating, and seek new driving forces. Based on the current large-scale transformation of the Chinese economy, the WISE2024 Business Kings Conference aims to discover the truly resilient "Business Kings" and explore the "right things" in the Chinese business wave.

From November 28 to 29, the two-day 36Kr WISE2024 Business Kings Conference was grandly held in Beijing. As an all-star event in the Chinese business field, the WISE Conference has now reached its twelfth session, witnessing the resilience and potential of Chinese business in an ever-changing era.

2024 is a year that is somewhat ambiguous and has more changes than stability. Compared to the past decade, people's pace is slowing down, and development is more rational. 2024 is also a year of seeking new economic momentum, and the new industrial changes have put forward higher requirements for the adaptability of each subject. This year's WISE Conference takes Hard But Right Thing as the theme. In 2024, what is the right thing has become a topic we want to discuss more.

Robots, artificial intelligence, and data technology are helping to achieve a leap in efficiency and creativity through empowerment and optimization: Technological empowerment has promoted a comprehensive upgrade from smart communities to business operations; Intelligent products have created more convenient and personalized services for society; At the same time, AI has become an accelerator of innovation, stimulating new boundaries of human creativity.

In the UP Roundtable session on November 28, five guests from Elephant Robotics, Qujing Technology, Shui On Xintiandi, WENGE.AI, and NetEase Yunshang discussed the industrial value of AI large models from the perspective of the AI industry.

From smart communities to the remodeling of business models, technology is not only getting closer to people's daily lives, creating a connection of warmth and value, but also revealing the stories behind the information and promoting the continuous evolution of society. In the future, the core of technology lies in empowering and amplifying human creativity, rather than replacing it.

Source: 36Kr

The following is the verbatim record of the conversation, edited and organized by 36Kr:

The rise of domestic chips is not only a powerful proof of our country's scientific and technological self-reliance, but also provides a solid hardware support for many cutting-edge technologies.

When it comes to cutting-edge technologies, the large model technology that has attracted much attention in recent years has to be mentioned. In recent years, artificial intelligence technology represented by large models is becoming the core driving force leading a new generation of industrial changes. Combining with the practical application scenarios in production and life and empowering and improving efficiency for thousands of industries is the inevitable development direction of artificial intelligence large models.

In this industrial change led by AI large models, what exciting surprises have we gained? And how should we plan ahead and properly deal with the challenges and risks that may be encountered? Next, we will enter the "UP Bureau" session, handing the stage to Ms. Deng Yongyi, a senior author of 36Kr. At the same time, please welcome Mr. Chen Haotian, CMO of Elephant Robotics, Mr. Ai Zhiyuan, CEO of Qujing Technology, Ms. Wei Tiantian, Assistant General Manager of Community Innovation of Shui On Xintiandi, Mr. Wen Hao, Vice President of WENGE.AI, and Mr. Zhou Dan, Deputy General Manager of NetEase Yunshang to bring us the theme sharing. Welcome!

Deng Yongyi: Welcome everyone to the WISE Conference. Thank you very much. I'm Deng Yongyi from 36Kr. The WISE Conference is a rather special day. In two days, it will be the second anniversary of OpenAI's release. I believe that in these two years, the whole world has felt the value of AI. For people in the industry, the sense of urgency is even stronger. I often hear interviewees saying: When I wake up, the world has changed again, and I don't know how to do my job.

So I'm very glad to invite five guests to share with us the feelings and observations of AI at the implementation level.

First of all, I still want to ask the five guests to introduce themselves and talk about the amazing moments they have seen in the past two years where AI has changed life or work.

CMO Chen Haotian of Elephant Robotics

Chen Haotian: Hello everyone! I'm the CMO of Elephant Robotics. Elephant Robotics has three product lines from industrial, commercial to consumer: assisting robotic arms, simulation pets, and humanoid robot products. In the past, we have had many connections with artificial intelligence. Let me share two points:

First, less than two months after OpenAI launched ChatGPT, Microsoft officially released a case. What does this case do? It uses ChatGPT to automatically write code and automatically train the model to make the robotic arm do the sorting mode.

Second, the concept of spatial intelligence, what does it do? It is equivalent to enabling the robot to perceive, understand, and even make inferences, and finally be able to independently complete the movement in the physical world and complete the corresponding tasks. This is closely related to the field where our current robots are. This year in China, the team at Tongji University also replicated such a case, and based on our robotic arm products, a DEMO was released at this year's Shanghai Artificial Intelligence Conference. These cases can be shared on Bilibili and the company's official WeChat account.

Illustration: CEO Ai Zhiyuan of Qujing Technology

Ai Zhiyuan: I'm from Qujing Technology. We are an AI manufacturer of large model inference acceleration from Tsinghua University.

Maybe everyone is not particularly familiar with this field. Now there are underlying chips and many large models, but how to make the large models run faster on the chips? This requires work in inference optimization, and that's what our company does. For example, due to the separated inference architecture design of Mooncake, Kimi can increase their traffic acceptance by more than 75%.

Our goal is to lower the entry threshold for the implementation of billion and trillion-scale large models, so that they can be applied in enterprises and units in thousands of industries.

From my perspective, the current large models have not yet reached the level of completely replacing our work, but it can assist or handle our daily relatively repetitive and cumbersome work.

But what surprises me is that many programmers in our company find it difficult to spend money. They would rather do it themselves than buy it. But I found that the R & D colleagues in the company basically have two screens, with GPT on the left and the development environment on the right. They also recharge the account for OpenAI GPT every month. The change in consumption habits proves that GPT is really useful and helpful to them.

The emergence of large models makes people feel that they need such tools to assist and speed up in the daily work process, rather than completely replacing them. This has a relatively large impact on me.

Illustration: Wei Tiantian, Assistant General Manager of Community Innovation of Shui On Xintiandi

Wei Tiantian: Our company is called Shui On Xintiandi, a Hong Kong company. The projects it develops include a series such as Shanghai Xintiandi. My department does something special, which is to do community business innovation and venture capital. Therefore, we will pay attention to the underlying structural changes in the future city, and focus on making the city more sustainable, intelligent, and people-oriented in the future. We carry out business content incubation and early-stage technology project investment.

I think an interesting point is that our understanding of AI is often from the application end or the consumer demand end. In the past two years, the pet economy has been very popular. Everyone also knows that many changes have taken place offline. I saw a news that a team in Silicon Valley made a video that shocked many people. A host held his own GPT, said a human language, and translated it into a dog's language. The dog understood and brought the remote control back.

Such a scene excites us because from our perspective, we saw the progress of technology for the first time - it stands at a more subdivided human demand perspective, not only a functional response, but also an emotional response. This is also the point where we pay attention to technology and find it very interesting. We look at new opportunities at the intersection of humanity, care, and technology.

Vice President Wen Hao of WENGE.AI

Wen Hao: Hello everyone, those who can persist until now are true fans.

I'm Wen Hao from WENGE.AI. The establishment of WENGE.AI originated from the story of a teacher leading students to start a business. This teacher is a member of the 1985 class of the Juvenile Class of the University of Science and Technology of China. He went to the United States to study artificial intelligence in 1995, studying under the famous artificial intelligence professor Katia. She is one of the proposers and promoters of multi-agent. In 1995, he published relevant academic papers on reinforcement learning, Bayesian learning, and multi-agent, with thousands of academic citations, which have played an important role in promoting research in related fields. After returning to China in 2007, he began to lead many students, including the chairman and CEO of WENGE.AI, to undertake multiple national-level big data projects and continuously provide services to multiple government departments.

Here, I would like to share a case - the Intelligent Media Project of Xinhua News Agency. Xinhua News Agency has reporters in more than 100 countries around the world, providing them with a media editing, publishing and distribution platform. This case is a small epitome of the team's industrialization process. In 2017, coinciding with the country's encouragement of innovation and entrepreneurship, the team established WENGE.AI based on the accumulation of the original industry-university-research-application. Since its establishment, we have been continuously making progress and innovation in the fields of big data, small models, artificial intelligence large models, language large models, and video multi-modal large models.

When it comes to the practical application of artificial intelligence, we cannot fail to mention traditional Chinese medicine. As a treasure of Chinese culture, traditional Chinese medicine has helped us overcome plagues many times in history, showing unique therapeutic value. Last year, we jointly developed the "Dayi Jinkui" large model with the China Academy of Chinese Medical Sciences, successfully combining the traditional "black box" diagnosis and treatment method with artificial intelligence, creating a new application mode. It is worth mentioning that "Dayi Jinkui" obtained a high score in the Traditional Chinese Medicine Practitioner Qualification Test.

Based on the "Dayi Jinkui" large model, we have launched the same series of TCM diagnosis and treatment APP. A user gave feedback after using this application. He had multiple rounds of inquiries due to fever symptoms. Through the first diagnosis, it was determined that he had a cold due to wind and cold, rather than wind and heat. He also received corresponding dietary suggestions and OTC drugs. The user's symptoms were relieved after taking the medicine, and he highly praised our service.

Artificial intelligence is constantly bringing us surprises. In the future, we are full of expectations for the development of related technologies. Thank you!

Deputy General Manager Zhou Dan of NetEase Yunshang

Zhou Dan: Hello everyone! I'm Zhou Dan from NetEase Yunshang. Our NetEase Yunshang is a SaaS operator company for enterprises. We released our own intelligent customer service product in 2016, which more or less has been used in everyone's life and work.

When we established the project within NetEase Group in 2016, we took intelligence and AI as our differentiating points. Perhaps it is precisely because of this that we have smoothly developed to the present in several rounds of customer service product iterations.

In the third quarter of 2023, we have integrated the large model into our products, including the launch of the first vertical large model in the customer service field - the Shanghe Large Model in September. Everyone knows that NetEase Yunshang is a company that is relatively good at creating products. In the process, we still think it is relatively stable and steady.

There are two points that impressed me. The first is that when our employees, product managers, and R & D are developing large model products, they will spontaneously buy some AI tools. We found that it can indeed improve the personal efficiency of employees. The second point is that when AI really starts to operate in the organization, we will find a phenomenon that the knowledge in the employees' minds can be quickly transformed into the common assets of the organization. The talents or precipitations of excellent employees are more likely to be absorbed by other employees and become assets in the organization. In fact, my boss, our CEO, also said last year that the organization should self-iterate and self-update like AI. In fact, after our organization made AI products at that time, it also self-updated like AI. This point is still quite impressive to me.

How Does AI Move from Online to the Real World?

Deng Yongyi: Thank you for the sharing of several guests. From ordinary individuals to the entire organization, everyone can feel the value of AI. Today, in fact, the fields of the five guests are relatively diverse. We can see that, for example, like Mr. Chen from the hardware field, Mr. Ai, Mr. Wen, and Mr. Zhou are exploring the value of AI at the model layer or application layer, while Ms. Wei is in the offline business space.

The second question I would like to ask the guests to discuss is how AI has demonstrated its value in their respective fields in the past two years. This can be combined with their own businesses to talk about. In the past two years, for example, what AI explorations have we made and what is the actual value that AI has generated?

Zhou Dan: AI has been implemented relatively early. When the product was implemented, we already took AI as the core skill point. At that time, the AI technology was mainly NLP search-related technology, providing capabilities such as classification and summarization, but this ability was still relatively primary, so we could only do some relatively simple.