Hello, I'm Wang Yuquan, here is Wang Yuquan's important news comment.
Amazon seems to have been left behind by Microsoft and Google in the AI wave, but has recently made a big move.
On September 25, Amazon announced that it would invest 4 billion US dollars, or about 29.2 billion yuan, in AI startup Anthropic to obtain a minority stake in the company. Anthropic's business is similar to OpenAI, and they have developed their own large-model product, Claude, which is hailed as the main competitor of OpenAI and Google in the generative AI space.
Specifically, Amazon will invest an initial $1.25 billion, with the option to increase to $4 billion later.
In the future, Amazon AWS will become the main cloud provider for Anthropic, which will train and deploy future large models on Amazon's Trainium and Inferentia chips, which are Amazon's customized AI chips, and the two companies will also collaborate to develop AI chip technology.
Customers on the Amazon AWS cloud platform can easily call Anthropic large models, and can also customize and fine-tune models.
Simply put, Amazon wants to use Anthropic, a "foreign aid" to improve the AI service capabilities of the cloud platform. Why would they do that?
I once proposed in the science and technology training camp that the cloud platform that supports artificial intelligence training will be the mainstream in the future, and the growth space of cloud computing will be impacted by AI. Whether it is the training, iteration or application of large models, it requires the huge computing power of the cloud platform, and the two complement each other.
Although Amazon still has the largest market share of cloud services, this market position is not unbreakable, and even somewhat precarious.
You know, the growth rate of Amazon's cloud business is slowing, with only 12% growth in the second quarter of this year, while Microsoft increased by 26% and Google increased by 31%. Not only that, the market share is also declining, last year the Amazon cloud business market share was still 40%, in the second quarter of this year has dropped to 30%, the second place Microsoft rose to 26%, is very close.
With its own growth slowing down and competitors catching up, Amazon naturally tries its best to grab the emerging market of cloud services, and the most promising new market is undoubtedly generative AI.
However, Microsoft is deeply tied to OpenAI, has a powerful GPT-4 model, and Google also has its own Bard model, although slightly inferior to GPT-4, but also better than many models on the market.
Amazon only launched its own big model, the Titan, in April, but it has been tepid. The problem is that competitors aren't waiting for Amazon to catch up.
For example, Google Cloud CEO Thomas Kurian said that more than 50% of AI startups and more than 70% of generative AI unicorns in the industry are Google Cloud customers; Microsoft's Azure OpenAI service is also doing well, winning big customers such as IKEA, Volvo and Zurich Insurance.
Amazon may be aware of the crisis, but to keep up with Microsoft and Google, it will have to find strong "outside help" in the field of large models, and OpenAI has become difficult to intervene, and Anthropic has become an alternative.
Anthropic is not as well-known as OpenAI, and those unfamiliar with the AI industry may not have heard of it, so why do they claim to be OpenAI's main competitors?
In fact, many of Anthropic's founding team came out of OpenAI, such as Dario Amodei, former vice president of research at OpenAI, Tom Brown, the first author of the GPT-3 paper, and many of the other founding members are core employees of OpenAI who have participated in the development of GPT-3.
However, fame is not enough, the most important thing is that they have the ability to quickly develop large models. Three months after OpenAI released GPT-3.5, Anthropic quickly launched its competitor, Claude, while Google's Bard was still testing, and four months after GPT-4 launched, they launched Claude 2.
Although the Claude 2 does not surpass the GPT-4 in terms of performance, it has been able to meet the needs of some enterprise customers. For example, travel publisher Lonely Planet reduced the cost of creating itineraries by 80 percent after deploying Claude 2, and asset manager Bridgewater Associates is using Claude 2 to develop an investment analyst assistant to generate charts, calculate financial metrics and create summaries of results.
With their scarce top teams and good model results, they attracted $1.45 billion in investment, making them the second most valued AI unicorn outside of OpenAI.
To compete with OpenAI, they plan to spend $5 billion on the next big model, Claude-Next, but the problem is that one of their killer features for attracting users is the free-to-play model, so it's obviously hard to do it on their own.
And they are precisely serving customers on the Amazon cloud platform, Anthropic needs computing power and funds, Amazon needs high-quality large models, and the cooperation between the two sides is logical.
In fact, the cooperation between Amazon and Anthropic is similar to the combination of Microsoft and OpenAI, both AI startups with strong technical capabilities and well-funded traditional technology giants, essentially complementary advantages of computing power and models, and then use the giant's user base to jointly expand the AI service market.
So, what kind of industrial impact does Amazon's association with Anthropic have?
Originally, GPT-4 can be said to be a standout, but from the perspective of industrial development, this is not a good thing, it is easy to form a monopoly, and it is easy to gradually lose the vitality of innovation. After Meta launched the open source model Llama2, it created an opportunity for more people to enter the big model.
However, the model alone is not enough, if it can be deeply combined with computing power, it is possible to tap greater potential. It can be seen that Amazon and Anthropic not only jointly provide cloud large model services, but also cooperate in the research and development of AI chips, which may become a new force in the field of generative AI in the future.
This means that the field of AI computing power and large models will form a new industrial competition pattern. The game and mutual checks and balances between giants are very good for industrial evolution, not only will promote the update and iteration of models, but also want to enter the application of large models enterprises also have more choice space, is likely to bring new vitality to the AI industry, AI competition between giants has just begun, the future is very worth looking forward to.
The above is today's content, more detailed industrial analysis and underlying logic, I will share in the science and technology training camp. Welcome to pay attention to the global wind outlet WeChat account and sign up!