Written by | Tian Xiaomeng
Edit | Lee Shin Ma
题图 | IC Photo
From the second half of 2022 to the beginning of 2023, ChatGPT attracted a lot of attention, and its emergence made the public quickly accept and start using generative AI, and artificial intelligence also set off a new round of trends.
OpenAI reports that ChatGPT attracted 100 million users within 60 days of its release, compared to TikTok that took nine months to reach that milestone. Dall-E 2 is visited by about 1.5 million users per day. Google's Bard chatbot had 10 million page views in July.
In China, large models such as Wenxin Yiyan, Tongyi Qianwen, and Xunfei Xinghuo have sprung up. Benefiting from the promotion of capital, the news of AI-related corporate financing has also continued one after another. For example, recently, Facewall Intelligence has completed a new round of financing of hundreds of millions of yuan, the content business AIGC video application platform "Chopsticks Technology" has completed nearly 50 million yuan in B1 round of financing, and Mobvoi, which focuses on generative AI and voice interaction technology, has officially launched its IPO and plans to be officially listed on the main board of the Canton Stock Exchange on April 24 this year.
The discussion of artificial intelligence in 2023 is not as good as when Lee Sedol defeated Alpha Dog. However, according to the latest report from the Stanford Institute for Artificial Intelligence (HAI), global AI investment will decline for the second consecutive year in 2023, with AI-related mergers and acquisitions falling from $117.16 billion in 2022 to $80.61 billion in 2023, a decrease of 31.2%; Private investment fell from $103.4 billion to $95.99 billion. Overall, total investment in AI fell to $189.2 billion last year, down 20% from 2022.
Image Source: AI Index Report 2024
It is worth mentioning that generative AI attracted $25.2 billion in investment in 2023, almost 9 times more than in 2022 and 30 times as much as in 2019. Moreover, generative AI accounted for more than a quarter of all AI-related private investments in 2023.
Image Source: AI Index Report 2024
1. Domestic AI investment computing power is the most popular
The world's 5G looks at China, and so does artificial intelligence, and China's rise in the field of artificial intelligence should not be underestimated.
According to the data, in 2023, in terms of private investment in artificial intelligence, the United States occupies the first place with 67.2 billion US dollars, followed by China, with 7.8 billion US dollars of investment and 17.8 times that of the United Kingdom's 3.8 billion US dollars, in terms of machine learning models, the United States leads with 61 well-known machine learning models, followed by China with 15, and China accounts for 61.1% and the United States accounts for 20.9% of artificial intelligence patents.
Returning to the domestic market, AI investment has gone through several waves of ebb and flow.
Combined with the statistics of China's AI investment and financing from 2018 to November 2023 and the AI investment and financing data in each month of 2023, 2018 and 2021 were the relative highs of the amount and number of investments in the AI field in the past, and the amount of AI investment and the number of shots in 2023 are actually lower than those in 2022.
Image source: China Business Intelligence Network
Yi Lijun, founding partner of Venture Capital, told DoNews that 2023 will be a wave of enthusiasm brought by generative AI (ChatGPT), but most of the institutions that actually sell are concentrated in industrial capital and US dollar funds, and the investment is mainly focused on the investment of large models and computing power. AI+ vertical models and industry applications are just in the initial exploration stage, so the discussion is hot, and the actual investment is low.
Specifically, AI computing infrastructure is the most important for capital in the domestic market, because the United States has leading companies such as Nvidia and AMD, while domestic computing power only accounts for 5% of the world, but China's demand for computing power accounts for one-third of the world's total, and the demand for domestic substitution is urgent, so the amount of domestic investment in computing infrastructure is twice that of the United States.
Large AI models are second. The reason is that domestic start-ups and large factories are basically on the same starting line, and start-ups are pre-trained using open-source models, coupled with the data advantages of large manufacturers, the capital required is also huge.
In the field of AI applications, from medical diagnosis and financial services to autonomous driving and intelligent manufacturing, the application scope of AI technology is constantly expanding. Capital markets tend to invest in AI applications that solve real problems, improve efficiency, and create new business models. These applications not only bring direct economic benefits, but also drive innovation and development across the industry.
Whether there will be a rebound in 2024 is a topic that AI companies and investors are concerned about, and will AI be as short-lived as the VR/AR popularity after the emergence of the metaverse concept?
"The large model generalizes a general basic ability and turns it into a brain, which can be applied in all walks of life, so that the machine has more and more ways of thinking, creating, and adapting to real humans. Therefore, the scale of the entire AI track and the application field are far better than AR/VR. Yi Lijun said. Although the whole industry has small cycle fluctuations, it is still a track of sustainable development, industrial growth and long-term investment attention.
But it should not be ignored that at some point in time, everyone is on it, but in the future, when new technology comes out, it may become a kind of technical debt.
Yi Lijun said frankly that the "technical debt" is actually determined by the attribute and scale of the institution's source of funds, and when the attribute of the fund is to prefer short-term financial returns, rather than long-term value, it will choose to "gamble" on the outlet. "For us, the previous investment will not become our liability but an asset, always said that investment is the realization of cognition, many times the change of technology seems to be disorderly, but in fact there are traces to follow, we need to judge by the experience of previous cases. ”
Second, AI technology innovation TOB first
"The hustle and bustle of the capital market does not necessarily mean the maturity and practical application of technology. Meng Xiaofan, President of Deloitte China Consulting Group, said in an interview with the media.
Although Wensheng pictures and videos have attracted a lot of attention, the AI solutions of enterprises have not yet reached the mature stage, and they are just superimposing AI functions on the windows of the original products. Zhou Lingkun, President of Deloitte China Consulting Enterprise Technology & Performance Practice Group, agrees. He believes that according to the technology maturity model, generative AI is currently in a period of expectation inflation.
Of course, there is an expectation to have the motivation to innovate. Nowadays, AI is not only integrated into mobile phones, but also into automobiles and multiple verticals.
Which tracks will AI make "explode" first?
Yi Lijun said that technological innovation is generally TOB first, mainly because enterprise users have an urgent need to improve efficiency, reduce costs and enhance competitiveness, and AI technology can effectively solve these problems. With the advancement of technology and the increasing expectations of consumers for intelligent services, the C-end can only explode when the cost drops to the range that consumers can afford.
Taking OpenAI as an example, the products he started with are also TOB, for example, Copliot applied in Microsoft is used to assist enterprises in coding, and GPTStore is also used by enterprises to expand the functions of applications with ChatGPT plug-ins, with the purpose of creating an ecosystem like Apple store, making profits for developers, and promoting the development of AI.
Yi Lijun also pointed out: "Now both domestic and foreign AI companies are facing a problem - how to make profits. "According to Sequoia, in 2023, the global investment in generative AI is $50 billion, but the revenue is only $3 billion. The willingness of domestic users to pay is relatively low compared with foreign countries, and few SaaS companies are profitable so far, so the domestic landing scenario must be on the TOB side first.
"AI products will explode on the C-side, but the biggest problem right now is demand. Yi Lijun said that the needs of the C-end are not clear, and the C-end users have no obvious application scenarios for the functions that have emerged from generative AI, nor is it a pain point. "The problem of clothing, food, housing and transportation has been met by existing products, so I think AI needs to combine robots to break through the framework of the digital world to the physical world and solve the daily needs of C-end users. ”
As an important medium for AI landing, whether it is TOB or TOC, Yi Lijun said that there have been two major trends.
First, more powerful edge computing capabilities: With the development of edge computing technology, AI terminals will have more powerful local processing capabilities. Taking AIPC as an example, with the optimization of AI algorithms and the improvement of hardware performance, more and more AI computing tasks can be completed locally on personal computing devices, which will reduce the dependence on cloud computing and improve the speed and privacy of data processing.
Second, more intelligent services and interaction capabilities: With the continuous progress of machine learning and deep learning technology, AI terminals will be able to provide more intelligent services, including more accurate speech recognition, image processing, natural language understanding and other functions, making the interaction between users and AI terminals more natural and efficient. For example, Xiaomi Su7, after the user calls Xiao Ai, he can directly control the device and the application through voice recognition and dialogue.
Therefore, AI terminals need to be closer to user needs, which is a combination of AI capabilities + products and processes that meet user needs.
Third, the growth of AI start-ups is full of thorns
The allure of generative AI is a competition for companies.
According to the Fortune/Deloitte CEO Survey Insights' 2023 survey, 80% of respondents believe that generative AI will improve business efficiency, 55% say that enterprises are evaluating or experimenting with generative AI, and 52% believe that generative AI will enhance business growth opportunities.
Among them, there are many start-ups. According to survey data from market research firm Pitchbook, the total funding of generative AI-related startups will reach $27 billion in 2023, of which about $18 billion will come from technology giants such as Microsoft, Google, Amazon, and Nvidia.
This year, the actions of these big factories are even more continuous. Amazon said it would invest an additional $2.75 billion in AI startup Anthropic to close a deal struck last year, investors close to Elon Musk are in talks to help raise $3 billion for his AI startup xAI, and OpenAI co-founder and CEO Sam Altman and former Apple chief designer Jonathan Yves are seeking $1 billion in funding for a new undisclosed AI project......
It seems that the start-up has unlimited scenery, but it also bears a lot of hardships behind it.
For example, GPT3.5 requires 10,000 A100 chips to train, but the money to buy the card and the operating costs are not something that startups can get into. Yi Lijun said bluntly, "Everyone thinks that there will not be so many players with large models in the market, and there are currently more than 100 companies on the Chinese list, and I think there will only be a few big factories left in 3-5 years." ”
Therefore, it will take time to prove whether AI start-ups can withstand the big waves in the next few years.
In addition, Meng Xiaofan put forward three suggestions for enterprises to use AI technology.
The first is to form its own AI strategy at the enterprise level as soon as possible, and at the same time, it is necessary to form an AI construction structure for the rapidly evolving AGI market, including the scenario level, process level, AI level, and data level, and consider what kind of value AI can achieve in the enterprise.
The second is to set the goal of becoming an AI-driven organization. AI should provide employees with tools to help them improve and enhance their productivity, knowledge and creativity, and continuously promote enterprise innovation, rather than trying to replace employees with AI or forcibly adding AI to the company's processes.
The third is to pay attention to the importance of data. In order to establish their own competitive advantage and comparative advantage, enterprises will inevitably build up some resources, but after AI, if the data problem is not handled well, it will not be easy to trace the source, and it may become a coffee wing.
Based on the long-term development of technology, Zhou Lingkun said: "I hope that the future technology system will no longer be a 'headache doctor, foot pain doctor' system replacement architecture upgrade and adjustment, but to establish a framework system for health examination." ”
end
It is not difficult to see that in the noisy market environment, rational investment is still the main theme.
For the future development of AI, AI must not only realize its own technological evolution, but also integrate and innovate with technologies such as 5G and cloud computing and various industries to achieve commercialization. It can be said that the road is long.
It has to be mentioned that there are still some crisis awareness of the arrival of AI - whether artificial intelligence will replace people. Recently, in the process of interviewing and participating in activities, many experts have a positive and optimistic attitude, which also gives the author a reassurance.
In Yi Lijun's view, AI is just a tool, and the two are more of a coexistence relationship. The role of AI is more to assist humans in production or creation. First, AI may be more efficient and precise than humans at specific tasks, especially in scenarios where large amounts of data are processed, repetitive tasks are performed, or fast calculations are required. But AI currently lacks the creativity, emotional understanding, and moral judgment capabilities of humans, which are also important for solving complex problems and making decisions.
Fu Changwen, President of CCID Consulting, predicted at the 2024 IT Market Annual Conference that "AI will be eliminated and added in manufacturing, education, construction, medical care, and finance, and there will be a net increase of more than 2 million jobs in the manufacturing industry, which fully shows that the manufacturing industry and AI are in a positive relationship." ”