20+ big names debate China's AIGC industry! The scene was crowded, and more than 2 million online netizens said: 'It is necessary to watch it again in full'

What kind of new world will ChatGPT bring? Industry leaders have a lot to say:

Qubit AIGC Summit: Baidu, Microsoft and other 20 big names debated! It was fully booked offline and over 2 million people watched it online

This is an opportunity for all mankind!

The global "AI re-engineering" has begun, and companies that cannot keep up will completely lose their future competitiveness. Model as a service will create a trillion-level market.

Rather than being the “iPhone moment,” this wave is more like the early days of the PC Internet.

Has AI really created a wave of unemployment? I don't think so.

The site of the first China AIGC Industry Summit turned out to be a stage for heated debates among big names and experts. Just looking at this scene of "people following the crowd" is enough to see how exciting and exciting it is.

##The online live broadcast was also very lively, and everyone followed the guests closely. Share to start a discussion.

## There are even netizens leading formations in the group: Very It is necessary to watch it again in its entirety.

#It has been recognized that we are at a critical juncture . The new era of AIGC led by ChatGPT is redefining our production and life.

How should everyone grasp this global storm? What new opportunities and challenges do Chinese enterprises have?

Looking at the speeches and roundtable sharing throughout the conference, nearly 20 industry players and venture capital representatives have comprehensively answered the above two aspects. issues and look forward to a new future.

Including executives from well-known industry players such as Microsoft, Baidu, iFlytek, SenseTime, Xiaoice, Kunlun Wanwei, etc.; also Representatives from innovative companies include Zhou Ming, founder of Lanzhou Technology, Lin Yonghua, chief engineer of Zhiyuan Research Institute, Yuan Quan, founder of Qiyuan World, and Zhang Peng, CEO of Zhipu Huazhang.

The speech content is mainly organized into three dimensions for discussion, including

AIGC new world, the path and thinking of ChatGPT commercialization, and New value of AI industry

.

In addition, this summit also has two roundtable forums that integrate industry and venture capital experts to directly address the most concerned issues at present. Two major topics: AIGC commercialization and new opportunities for AIGC in China.

Without further ado, I will take you through this article.

##AIGC New World

Lanzhou Technology Zhou Ming: Entering AI 2.0 does not mean that we must abandon 1.0 technology

The first thing to start is today's One of the most accomplished Chinese in the field of NLP,
Zhou Ming

, was the former chairman of ACL. In June 2021, Zhou Ming established

Lanzhou Technology, a cognitive intelligence company based on NLP technology. After open-sourcing the lightweight "Mencius Model" in 2021, he launched Mencius in March this year MCaht controllable large model.

What is the difference between ChatGPT and the previous large model? Zhou Ming explained his understanding.

First of all, ChatGPT understands the user's
intention

very clearly and more accurately; secondly,

multiple rounds of dialogue The modeling is very good, and there are some analytical abilities. However, from an industrial perspective, many tasks that require parsing capabilities can be better achieved with a separate engine than ChatGPT. Therefore, it is very important to find the application scenarios of ChatGPT. Zhou Ming proposed two considerations:

> When To C, the more AGI capabilities the better , users complete all calls in a simple interface;

> In To B environment, the ability to understand problems and multiple rounds of dialogue is connected to the company's own core engine and combined well.

This time, Zhou Ming mainly stood from the perspective of To B and shared his views on the digital transformation of enterprises in the AI ​​2.0 era. .

After the emergence of ChatGPT, capabilities emerged that could never be achieved with a single model or single task, thus entering the AI ​​2.0 era. Its characteristics It is a model with N tasks. In the future, even one model can solve infinite tasks.

# However, when used internally by an enterprise, you will find that many functions of AGI are not used, and some capabilities may cause trouble.

#With this as the background, when enterprises carry out digital transformation, should not pursue AI 2.0 but must abandon AI 1.0. Enterprises should try to seamlessly connect the AI ​​1.0 functional engine and some AI 2.0 capabilities (such as problem understanding and multi-round dialogue capabilities) to form a vertical and horizontal synergy. The team that practices the model and the team that uses the model continue to work together. Generate interactions and jointly improve efficiency.

When Lanzhou made a large model, he especially considered the issue of business implementation. You don’t have to do ChatGPT. In the field of AIGC, focus on keyword generation, custom templates, text and graphics work, etc., make the model light and good, and combined with user needs, enterprise data can be used wisely stand up.

##Xiaobing Xu Yuanchun: The birth of any technology generates industrial value

#The value of AI content generation lies not only in reducing costs and increasing efficiency, but more importantly, in solving problems, improving efficiency, and creating scenarios.

The above three points are Xiaobing COO Xu Yuanchun’s summary of AIGC’s creative capabilities based on past experience.

#Among them,

improving efficiency is The most obvious and easiest to understand point. For example, in the field of video, during the 2022 World Cup in Qatar, more than half of the programs of commentator Liu Jianhong were based on his own recordings and generated by AI replicating images and synthesizing sounds.

The key to improving efficiency lies in high concurrency and stability. The reason for emphasizing stability is that many technologies have achieved algorithmic value in the laboratory, but in the end The one-kilometer landing is very difficult, so the industrial application standards of the technology must meet stability requirements.

In terms of

solving problems, all AI companies or technology companies should realize that the problems that need to be solved should be The real pain points that industries, especially traditional industries, urgently need to pay attention to and change.

For example, using AIGC capabilities to create and generate virtual digital employees means that without changing or diluting the existing IP of the enterprise, Being able to create more content in multiple time and space and on multiple occasions - content is the most important asset and core productivity of an IP.

Also

Expand new scenes. In addition to using digital twin technology, virtual humans are used to produce and disseminate content to obtain a wider range of traffic, decoupling the operation team from IP, decoupling content from image, and meeting a large number of content output needs.

#For example, digital employees can live broadcast more sessions. It is mentioned here that live broadcasts are no longer limited to video recordings, but also real-time live broadcasts, including using local languages ​​and local texts to respond to questions in the live broadcast room overseas, generating real-time interaction and actual communication with users, and ultimately triggering purchasing behavior.

In this scene, the integration of virtual people and AIGC reaches a deal and promotes sales. From this perspective, AIGC not only produces content, but also Real sales revenue was generated. Based on this, how to make the model more suitable for recommendation is a more traditional application in large models.

Kunlun Wanwei Fanghan: The greatest social significance of AIGC is to provide low-cost ultimate solutions

As a senior Internet practitioner,

Kunlun Wanwei CEO Fang Han talked about some of his recent thoughts.

First of all, in terms of the implementation of the AIGC business model, he observed two laws:

First, the threshold for user creation is lowered, and the number of content creations increases tenfold; second, the business logic of C-side tools is community, while the essence of B-side is the completeness of functions.

The second is the impact of AIGC on producers.

Regarding the understanding and expression of existing knowledge, AIGC will achieve a level of 80 points at a low cost. If we humans can only achieve 80 points, then the job will definitely be replaced; only people with 80 points or above can manage "AIGC operators", complete the work that once required waist workers, and greatly increase production. , and waist workers will most likely be eliminated.

In addition, Fang Han also mentioned several observations:

First, Open source large models are a powerful supplement and replacement for commercial closed source large models.

Just like ChatGPT can be imagined as early Windows. The existence of Windows is a cornerstone of the entire computer business software field and is also the production tool for most people. Linux has turned itself into a Windows replacement through 30 years of hard work. Therefore, large open source models like "Linux" will definitely appear.

Second, Only the open source model can meet the long-tail needs of users. For example, Stable Diffusion, its quality and performance are lower than DALL·E2 and Midjourney, but its users are growing rapidly, and it has also spawned many secondary development tools to solve the rigidity problem. As the last high wall for AI creation, controllable generation is very likely to have further breakthroughs in the foreseeable time.

Third, The capabilities of various AIGC tools are still limited by pre-trained large models, which appeared in GPT-4 It became more prominent after that.

The greatest significance of AIGC to the entire society in the future is to provide the ultimate low-cost solution; secondly, it will completely change the way of artistic creation and production, and finally, it will be concrete When it comes to content generation, the emergence of a large amount of creative content makes VR and the metaverse more feasible.

##ChatGPT commercialization

Microsoft Guan Weiya: Every application can be driven by AI

Currently the world’s most watched The number one player in AIGC is undoubtedly Microsoft, which has successively released two disruptive applications in search engines and productivity tools.

Microsoft Greater China Chief Strategy Officer Weiya Guan came to the scene to talk about "New Opportunities in the Industry, AIGC Opens a New Generation of Productivity Revolution" .

##More and more organizations and institutions are turning to the cloud Transformation of native and AI technologies. You can imagine that in the era of artificial intelligence, every application can be driven by AI to achieve higher efficiency and better user experience.

She gave an example from the cooperation with OpenAI. OpenAI's success can be attributed to four aspects: a strong talent echelon; advanced technology base; long-term research and investment; and a culture of open cooperation.

In addition, Microsoft has also made positive contributions to the development of OpenAI: a strategic investment of US$1 billion and the creation of the world's first Five supercomputers also provide diversified and commercial paths.

#The two companies combined to not only train multiple large-scale AI models that shocked the world, but also deeply integrated AI technology with Microsoft's full line of products. , allowing enterprises and individuals to enjoy the potential and possibilities brought by AIGC.

AIGC’s entrepreneurial path can be divided into three functional dimensions:

The first category is to help enterprises reduce operating costs, such as the recently launched Microsoft 365 Copilot;

The second category is to improve customer satisfaction degree and product experience, for example, through voice interaction technology, it helps different types of companies provide consumers with personalized products. This is particularly evident in digital human technology.

The third category, as a revolutionary technology, AGI also bears very important social responsibilities. For example, provide more convenient life services for people with disabilities.

Similarly, AIGC’s business vision in different industries is also real and specific, such as manufacturing, smart cars, finance, medical, Games, entertainment, education and other fields.

For enterprises, how to upgrade AI strategy to the CEO level? Which scenarios should be the first choice for pilot implementation? How to position the core capabilities of enterprises and talent quality models in the AI ​​era? How to reshape the company's processes, organization, and culture to better meet the new generation of productivity revolution? These are questions that all entrepreneurs need to think deeply about.

##Baidu Yuan Foyu: Wen Xin’s words will change the rules of cloud computing game

Domestic applications represented by Wen Xinyiyan have shown the potential for productization and commercialization.

Baidu Group Vice President Yuan Foyu said that Wen Xin’s words will fundamentally change the rules of the game in the cloud computing market. Due to the two major breakthroughs of large models and generative AI, the global "AI re-engineering" has begun, and companies that cannot keep up will completely lose their future competitiveness.

## will be explained specifically from three aspects.

First of all, this is Baidu’s years of technology accumulation and keen insight. On the one hand, without a 13-year investment of up to 100 billion yuan in technology research and development, there would be no big model like Wen Xinyiyan. On the other hand, last year we accurately predicted the directional changes in technology, namely generative AI. This year Wenxinyiyan will integrate with Baidu search, Xiaodu, Apollo autonomous driving and other businesses, but the bigger story lies in cloud computing. Baidu Intelligent Cloud is confident to become a leader.

Secondly, the rise of generative AI will bring "double-effect" improvements to production and operation efficiency and user experience, and will completely change everything industry, accelerating the realization of the "intelligent transition" of society. This can be seen from the capabilities released by Wen Xinyiyan, which can help thousands of industries achieve AI re-engineering.

Furthermore, today’s IT technology stack has adapted to the four-layer architecture of AI technology development

(chip, framework, model, application) . The mainstream business model of cloud computing will change from IaaS (Infrastructure as a Service) to MaaS (Model as a Service).

China has the most advanced and complete industrial chain in the world, which can generate a large amount of real industry needs and user feedback, which will greatly Promote the rapid development of generative AI, and as generative AI is implemented in the digital and real economies, Model as a Service (MaaS) will give birth to a trillion-level new market.

## iFLYTEK Gao Jianqing: Cognitive intelligence large model will bring major industry disruption and opportunities

Gao Jianqing, Executive Vice President of iFlytek AI Research Institute
, shared some progress of iFlytek in AIGC technology exploration and application innovation.

## Overall, iFlytek ranks among the top three AIGC There is accumulation and progress in modules, namely audio creation, visual creation and text creation.

Taking the key technology of audio creation as an example, Gao Jianqing believes that there will be two key trends in the future:

•

The interactive field achieves better personification, emotion and even colloquial language.

• In the field of content production, better control over semantic information such as timbre, rhythm, and accent.

Currently, speech synthesis has been used in more than ten fields such as news broadcasting, documentary narration, game commentary, and live advertising.

Under the wave of AIGC, there are also some key development trends in speech creation: large language models, such as the Audio LLM solution, can be used to control key factors in speech synthesis; AIGC can also be better utilized to pursue Sound art improves the speech synthesis experience by inserting sound effects and environmental sounds.

#Visually, iFlytek It has accumulated complete video creation capabilities with virtual humans as the core, including virtual human generation and reproduction. Using AIGC, the development trend of virtual humans is to form full-stack visual generation capabilities from text to sound to images centered on virtual humans.

##In terms of text generation, starting in 2023, iFlytek will also begin to integrate text generation capabilities into industry applications, especially in Medical and educational fields.

Gao Jianqing said that the cognitive intelligence large model will bring major industry disruption and opportunities: it will change the model of information distribution and acquisition, content production modes and modes of human-computer interaction. iFlytek will combine its understanding of scenarios and data to make layouts in education, medical care, human-computer interaction, office and other directions.

It is expected that in May
of this year, iFlytek will release the progress of large models in multiple industries.

​

Zhipu Huazhang Zhang Peng: Pre-trained large models have become the foundation of the AIGC era

Also as a representative player of AIGC, Zhang Peng
, founder and CEO of Zhipu Huazhang, showed us a series of preliminary studies on productization directly at the scene. , let netizens say: full of useful information.

#For example, code generation plug-in

CodeGeeX

, behind it are 13 billion parameters, more than 20 programming language code generation pre-trained models, with code generation, translation, annotation and other functions, helping programmers write 4 million lines of code every day.

##There are also those who started the first phase of internal testing some time ago ChatGLM, a ChatGPT-like product, has been tested internally by nearly 2,000 people and has generated more than 7.3 million tokens. It can cosplay and chat and solve physics/mathematics problems.

In addition, Zhipu has also open sourced one of the large models, ChatGLM-6B, with a scale of 6.2 billion, which supports running on a single GPU. , in third-party evaluation, equivalent to 70% ChatGPT capability level.

Based on this model, Zhipu has explored the business model of MaaS (Model as service)

, which can be The end-to-end service integrated from training to final application development can also provide API calls like OpenAI, and can also be directly used in a commercial manner for everyone to use.

The specific practical operation is the big model application group BigModel.ai, including solutions, products, demos, etc., such as programming assistance tools CodeGeeX, Big Model-assisted writing applications include Writing Frog, personalized robot Derpy, etc.

Zhang Peng believes that the AIGC industry base is a pre-trained large model for two reasons:

First, universal generalization ability, which is a very critical feature for reducing costs and increasing efficiency of AI research and development; second, incorporating more knowledge to enable better simulation of large models Human intelligence.

But at the same time, it also faces some challenges, such as high cost, huge training data, and long cycle.

Finally, Zhang Peng also shared a piece of work that recently excites him: Google uses the PaLM model plus the ViT visual model to make it work in the real world. Robots can complete complex tasks autonomously.

New opportunities in the era

Zhiyuan Lin Yonghua: Great The model has risen from a language model to a cognitive model

In this new wave of AI, for ChatGPT and various types of ChatGPT In terms of application, the most important thing is their underlying base: the large model.

Beijing Zhiyuan Research Institute once released the 1.75 trillion parameter large model Enlightenment 2.0. This time, Zhiyuan Research Institute Vice President and Chief Engineer Lin Yonghua brought the thinking that big What can models bring to the industry?

## In the past few years, big models have appeared frequently. Two trends can be summarized. One is that the model size is getting larger and larger, starting with 100 million parameters, and now it has exceeded one trillion parameters; the other is that single language mode has become multi-modal. Therefore, when paying attention to the new trend of large models, we must not only prove its innovation in the research field, but also face the new challenges brought by industrial implementation.

Starting from the generalization capabilities exploded from ChatGPT and GPT-4, Lin Yonghua believes that large models have risen from language models to cognitive models. Model.

She personally summarized the most important challenges for large models in the next 10 years:

•

Extremely large number of parameters. How many parameters can support application requirements? Should we blindly pursue a 100 billion model, or is a 10 billion model enough?

•

Training data. How much training data can feed a 10 billion or 100 billion model? How much information and data can be passed to downstream tasks?

•

Evaluation task. From single modality to multimodality, how to evaluate models that have moved toward cognition?

•

Continuous learning and fixed-point error correction. How to absorb knowledge and information at low cost? How to perform fixed-point correction when an error is discovered?

•

Efficiency issues.

It can be seen that the iceberg above the water surface reveals AIGC applications such as Vincent’s conversational task generation model, but there are still thick layers below the water surface. A thick iceberg technology stack is used to support the tip of the iceberg above the water. Therefore,

Large model evaluation method

has been placed in a more important position.

Currently, Zhiyuan Research Institute is uniting multiple research teams, institutions, and manufacturers to

build an AI basic large model evaluation system

. In addition, Zhiyuan is also building its own AI large model intelligent computing platform to optimize multi-tasking for large model training, and is also researching the next generation AI compiler technology across chip architectures.

##In the past 10 years, the rapid development of AI is inseparable from
Open source and openness

promotion. Lin Yonghua said that Zhiyuan has now built a technology open source system such as

FlagOpen around the bottom-up technology system of large models, open sourced algorithms, data, models, evaluation systems, etc., so that more developers can Enterprises will join in to jointly build full-stack technology for large models. SenseTime Yang Fan: The window period between technological breakthroughs and the formation of industrialized applications has been greatly shortened

ChatGPT has opened up new outlets for many downstream applications, but "how to implement it with high efficiency, low cost, and scale" has become an urgent problem to be solved. Yang Fan, co-founder of SenseTime and president of the large device business group, brought his thoughts.

The first opinion expressed by Yang Fan is: When the whole society is paying attention to ChatGPT and large language models, we cannot ignore the overall development of AI in recent years. The core breakthrough is actually all-round.

#Secondly, the technical progress of ChatGPT reveals a characteristic, That is, the window period between technological breakthroughs and the formation of industrial applications has been greatly shortened, replaced by a stronger acceleration effect in this link. It is not difficult to see that in the current market, there are many specific applications that are being explored and matured in various countries around the world.

Third, behind all this, we can not only see the aesthetics of violent data, such as large computing power, the aggregation of ultra-large-scale computing resources, the design of ultra-large-scale model structures, and the aggregation of massive data, but also all links Strong experience accumulation. In other words, current large-scale model research is inseparable from a large number of software engineering system issues, and core power consumption requires continuous precipitation and accumulation.

What changes will such technology bring to the AI ​​industry?

#Firstly, there are major changes in the production paradigm, and secondly, because the core capabilities have a large number of C-side factors, they can naturally create a data closed loop that is more in line with AI technology iterations. , forming a strong flywheel effect. This means that the product has long-term sustainability and business barriers are easier to establish.

## Yang Fan shared his views to emphasize that in response to such changes AI 2.0 and AI 1.0 companies will not be completely disrupted. SenseTime has been making large pre-training models since 2018, which supports the sustainable development of SenseTime's overall business.

The "one platform" in the "one platform and four pillars" of the SenseTime platform content refers to the "big device" of artificial intelligence infrastructure. Years of continuous investment stem from SenseTime’s insistence that AI infrastructure will support a more prosperous industrial ecosystem in the future. Speaking from the infrastructure level, a large amount of model design and system tuning experience can support large model services and help more people iterate their own models.

Overall, the core value of this matter is that the entire AI infrastructure will be flexible and scalable. By providing all-round capabilities, SenseTime hopes to provide more basic support capabilities for the entire artificial intelligence infrastructure, market and industry.

Qiyuan World Yuan Quan: AGI’s Chinese-style Exploration

##ChatGPT After breaking the circle, people also have more understanding and understanding of general artificial intelligence. As a pioneer in domestic AGI exploration, Yuan Quan

, founder and CEO of Qiyuan World, shared their exploration process in the past few years.

##From the perspective of Qiyuan, AGI starts from the game and moves towards The Internet and the metaverse are broken. The first is exploring this AGI in the game world, including playing StarCraft from scratch. The second stage is to break out of the Internet world. Especially after the release of Transformer, the commercial value of AGI on the Internet is 100 times greater than that of games. The third piece is the metaverse, which is one of the most important things in the next 5-10 years.

The essence of the metaverse is the combination of rich corpus and rich interactions on the Internet, plus 3D open world games.
How AIGC gives birth to the Metaverse and how to train AGI

in the Metaverse are very valuable propositions in the future.

The last step may be
AGI virtual and real migration

, service or application in the real world, this is Qiyuan’s response to the trend judge.

Qiyuan’s exploration in this process is mainly divided into two stages.
The first stage

is mainly based on various The game mainly trains large decision-making models, with parameter levels within tens of millions or 100 million, allowing AI to use small samples or even start from scratch to explore the boundaries of this type of task.

The second phase

is a language model with 1 billion to 10 billion parameters that we will start from 2021. Based on the large-scale decision-making model technology platform, Qiyuan's current goal is to develop AGI that can inspire and accompany people. Last year, it began to focus on chatting with NPCs. This is not only the field of games, but also for the wider virtual world, such as some historical figures and two-dimensional characters, and users also want to interact with them in depth.

For example, we made a companion NPC based on Liu Bei. You ask why Liu Bei likes to cry? What do you think of Adou? He can give interactions and dialogues based on personality and facts.

#Qiyuan World believes that AGI can serve any scenario. Inseparable from Understanding. Qiyuan's original intention and vision is to inspire and accompany more people.

#Roundtable Forum

Commercialization process after the explosion

"Hello New World" Summit Settings A roundtable forum was held with the theme: Commercialization of AIGC.

From last year’s explosion to this year’s hotness, AIGC’s technology iteration and commercialization are progressing almost in parallel. The four representative guests invited this time are:

• Yunbo Technology founder and CEO CEO, Mei Song, the pioneer of AIGC virtual content platform；

##• Partner of Source Code Capital, Huang Yungang continues to pay attention to technology-driven innovation;

##• 特zanCo-founder, 王喆## who focuses on the field of content technology #；

#•

TiamatFounder and CEO, domestic leading AI image generation technology service practitioner青gan.

#Head of Qubit Think Tank
Liu Meng Under the auspices of Yuan

, it mainly focused on

3 topics: the commercialization reasons behind the explosion of AIGC, entrepreneurship at the model layer and application layer, and new scenarios and new expectations.

The commercial reasons behind the explosion of AIGC

In Qinggan’s eyes, excellent generation effects are the driving force for AIGC’s commercialization. Images are naturally communicable. From relatively fragmented effects to the ability to achieve high-quality and rapid generation, it has attracted a lot of attention to promote the development of technology and brought judgment on the commercial value of image generation.

#Wang Zhe concluded that the first reason for the popularity is that the effect is good and the obvious progress in effect has brought image generation to the public; the second reason is that Open sources such as Stable Diffusion provide the possibility for large-scale reproduction. In addition, due to the high level of attention among young consumer groups, brands are also trying to use advanced technology as a bridge to communicate with consumers.

Huang Yungang believes that this is related to the fact that ChatGPT can be experienced by a large number of C-side users. ChatGPT essentially sets a direction for the technical roadmap, and its huge improvements in capabilities have also been transformed into productivity, allowing all walks of life to use this technology to do it all over again. Technological changes will gradually affect the industry, form products, and form businesses.

Mei Song’s view is that on the one hand, the evolution of AI technology has allowed its effects to reach a usable threshold, which can be converted into direct productivity, thus Reflect commercial value; on the other hand, the commercialization path of the first popular image and text generation is relatively short.

There is a certain gap between the commercialization of AIGC at home and abroad in the current market. Everyone generally believes that it depends on the business environment, market size, and user usage habits. Closely related to willingness to pay. Moreover, AIGC has actually greatly improved work efficiency in China, but because companies use it "moisturizing things silently", it is not like AIGC start-up companies that can receive timely attention from the media and the public.

Model layer VS application layer

AIGC's commercialization path is mostly divided into two ideas, one is to train its own large model, and the other is to innovate the application layer based on the existing large model. Which side is more worthy of being favored?

As an AI painting service provider, Qinggan believes that for a long time to come, innovation at the model layer and application layer will go hand in hand and be inseparable. . Service providers need to collect enough real data at the application layer and further optimize based on user feedback at the model layer.

Mei Song believes that the number of large models needed in the future will not be too many, so model layer entrepreneurship is more suitable for large companies, and more entrepreneurs should find new opportunities in the application layer. He also pointed out that one thing that needs to be considered is that application layer entrepreneurship will be restricted by the development of large-model technology, so it is necessary to try to grasp core customers and expand the industrial chain as much as possible.

Huang Yungang elaborated on his views on the widely discussed issue in the industry: “In the AIGC era, should all applications really be redone?” He said that investors and start-up companies take this opportunity to reshape the industry and subvert large companies; giants hope to use the AIGC era to widen the distance from latecomers.

In Wang Zhe’s view, in the face of technological innovation and technological iteration, all industries cannot be called “do it over again.” Innovation is inevitable. Some innovations in knowledge are small and structured, while some innovations bring disruptive changes. Only by defining structural changes can we redistribute value and have the opportunity to become a major manufacturer.

New scenes and new expectations

AIGC brings new opportunities. In terms of its commercialization, what scenarios are most worth looking forward to?

When Qinggan answered this question, the answer he gave was "How to attract consumers when content production becomes faster and better?" . When technology iterates and content becomes rampant, how to grab users’ attention will make the commercialization of the industry more interesting.

Huang Yungang's answer is "It's hard to answer" because he is very optimistic about many scenes and it is difficult to choose one with the most potential. . Moreover, as new product forms continue to emerge, new business forms beyond imagination may emerge, and heroes cannot be completely judged based on the current existing scenes.

Live streaming and games are the pan-entertainment scenarios that Mei Song is most optimistic about. Speaking of the reasons behind it, in addition to Mei Song’s own enthusiasm for these two fields In addition to being particularly liked and paid attention to, there is also a considerable market size - for entrepreneurship, it is very important to find a big market that you like, and AIGC is "definitely a big benefit" to the global game industry, and it is likely to explode into extremely valuable new products. Business type.

Wang Zhe is very concerned about the changes AIGC has brought to the marketing track. AIGC lowers the threshold of creation and improves work efficiency, which is very beneficial to consumer entertainment; at the same time, brands and consumers will move from one-way communication to frequent two-way communication in the AIGC era, during which there will be great business prospects.

​New Opportunities for China AIGC

##The topic that has attracted much attention is New Opportunities for China AIGC. Regarding this topic, representatives of enterprises and VCs who are at the forefront of the industry also launched a heated debate on the spot.

Zhu Lei, co-founder and COO of Yuanyu Intelligence. Representatives of the first batch of Chinese version of ChatGPT companies, their product ChatYuan attracted the attention of many people in the industry as soon as it was launched.

#Co-founder of Unbounded AI

马千里. As a pioneer in generative painting in China, Wujie AI currently covers more than 1 million users across the entire network.

VP of commercialization of Huayuan Shusapien

Lin Lenny. Huayuan Computing has completed the transition from computational intelligence to perceptual intelligence and then to cognitive intelligence.

#There are also those who have always been paying attention to the fields of deep technology and cutting-edge technology Fengrui Capital Investment Partner

Chen Shi

.

#Editor-in-Chief at Qubits
金磊

Under the auspices of , discussions were held around topics such as intuitive feelings, model homogeneity and miracles, and the Chinese version of ChatGPT.

Intuitive feeling

First of all, when talking about their feelings under the ChatGPT storm, several representatives mentioned positive feedback such as shock, impact, and excitement.

But at the same time, as an entrepreneur in related fields, Ma Qianli, co-founder of Unbounded AI, feels anxious because the speed of technological development is really It is too fast, but it still feels like progress; partners in the entire industry are also anxious, and they are worried that they will be eliminated if they fail to participate in this wave.

Vigorously miracle

As an inevitable trend in the current large model development ecosystem, Unbounded AI co-founder Ma Qianli and Fengrui Capital investment partner Chen Shi both believe that this is a proven breakthrough in the field of unsupervised learning and does not require humans to be so powerful. intervention.

In this regard, Ma Qianli also added: Whether it is "vigorous" or other methods, as long as a "miracle" can occur, it is a good thing.

#From a business perspective, Chen Shi, investment partner of Fengrui Capital, also said that most entrepreneurial opportunities may still be at the non-model level.

##中国版ChatGPT

HuayuanshusapienVice President of Commercialization林laneni holds a relatively positive attitude on this issue and believes that China will also have its own OpenAI company in the future.

Yuanyu Intelligence co-founder and COO Zhu Lei suggested that everyone pay more attention to some non-commercial projects. Especially the progress of domestic open source data sets and open source models, which may be the cornerstone of the future. Just like OpenAI was built for non-profit purposes from the beginning.

##New Opportunities for China AIGC

## Regarding the new opportunities of China AIGC, several big names also talked about their opinions.

Zhu Lei described this wave as being like the early days of the

PC Internet; Lin Laini focused on the issue of talent employment , believing that it will not create a wave of unemployment.

Both Ma Qianli and Chen Shi believe that this is an opportunity for all mankind. Even though there is still a certain gap compared with foreign countries, there is still a certain gap. Eventually we were able to catch up. Chen Shi also compared this wave to the mobile Internet back then. Everything was just a matter of time.

The above is the detailed content of 20+ big names debate China's AIGC industry! The scene was crowded, and more than 2 million online netizens said: 'It is necessary to watch it again in full'. For more information, please follow other related articles on the PHP Chinese website!