How much does it cost to develop the application layer of large AI models?

Labor costs are still a cost that cannot be ignored, especially for high-end talents. Just look at the US$10 million signing fee Microsoft issued to OpenAI employees two days ago. In addition to intellectual costs, the rest is the infrastructure cost of software and hardware. I will just talk about this today.

Large model cloud service

For the majority of small and medium-sized teams, using large model cloud basic services to build products is a relatively convenient technical route. MaaS (Model as a Service) is a relatively new term, and the Chinese explanation is model as a service.

8d5dd0cf1a8e38e5ce56534c1bf8c562.png

Domestic MaaS platform

Many domestic cloud platforms provide open source large model services to the outside world, such as Baidu Cloud, Tencent Cloud, Alibaba Cloud, etc.e6129b9cd9948d9d6b76eecac8e260b1.png8fc7143b80ae6d43ce33472071c155d8.png488b795761fe70546e5dd0861d067717.png

Overseas large model cloud service

In terms of certain capabilities, the capabilities of overseas models are still slightly better, and sometimes we have to adopt a survival curve method to rely on the power of overseas models. 5efbe599e2ba9640daba27867b8e9682.pngThe cost of the OpenAI interface provided by Microsoft Azure cloud is similar to that of OpenAI, but the services are two systems and are not interoperable.

Private cloud deployment

Sometimes, in order to have exclusive service capabilities and use models with maximum freedom, privatized deployment is also a business strategy that has to be adopted to respond to customer needs, such as vertically small-sized large models.e70cfd27f8765c7768e8981ec2ac8a39.png8945fb27fa6cbabdac2cbe019ea89e4d.png9f879f7225413056e6720bde70d09564.png

buy card

Due to some international trade reasons, some cards are not available for purchase if you want, and the prices are as high as the tide. The prices of some GPU cards that can be traded are also high, which has led to many teams preparing to develop their own chips, such as Microsoft, Amazon, OpenAI, Huawei, etc. If they don’t want to be stuck, they must take the technology into their own hands. Of course, whether you can break through depends on your strength.44987bc764ebfb053b2b1a918418b18b.png

In short, AI is a money-burning job and a hard technology. It is different from previous business models, such as e-commerce, social networking, video, etc. You can also get a rough idea from the scale of financing, which is often carried out in units of hundreds of millions.

Fortunately, if you are also on the AI ​​voyage, I wish you smooth sailing.

—Extended reading—

A book that can affect your (child's) life, not to be missed

This is the best translation software I have ever used, don’t miss it

Overcoming the 35-year-old middle-aged crisis

A big AI model with a big name but no useful purpose. It doesn’t live up to its name.

Having trouble learning English? Let me introduce you to some AI masters to accompany you in your studies (2)

Guess you like

Origin blog.csdn.net/hero272285642/article/details/134978208