csdn founder jiang tao: if you don’t master ai developer tools, you will lose the next strategic high ground
2024-09-25
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
the big model provides api capabilities, but jiang tao said that in order to actually use it in a product, in addition to data chain support, a tool chain is also needed, so the developer's tool chain is undergoing a huge evolution.
developers who can only write code or translate code by hand are easily replaced by tools and will face a job crisis. the crisis at the age of 35 will be greatly exacerbated. programmers who design interfaces may disappear, and aesthetic designers who create interactive forms are scarce, so positions for tool manufacturers under large models will be created.
big models drive intelligence and bring about a paradigm shift in development methods. on september 24, jiang tao, founder and chairman of the developer community csdn, said in an interview with the paper that the underlying model has evolved dramatically, evolving into a spatial model, a robot model, and a sequence model, and the developer's tool chain is also undergoing a huge evolution. the most direct experience for developers is code generation. the emergence of ai programming tools will exacerbate the 35-year-old crisis for programmers. programmers who understand structure and systems can use ai tools to improve productivity and empower more developers. creative high-level programmers have a better future, and programmers facing the business and demand sides will increase significantly.
csdn is the world's second largest developer community, with over 44 million real-name registered users, covering 95% of chinese developers in china and around the world. in the past year, 47% of csdn users were developers who paid attention to open source ai. jiang tao said that after the arrival of big models, the popularization of programming tools, open source cloud, application intelligence, and software localization and localization are all driving the innovation of developer productivity tools. if you don't master open source tools and developer tools, you will lose the next strategic high ground. to this end, csdn and huawei cloud launched the ai model community to build a chinese open source foundation.
jiang tao, founder and chairman of csdn.
talking about the developer ecosystem: the underlying model has evolved dramatically
jiang tao said that at the basic model level, there are basic models represented by transformer and other new routes, including liquid ai, a large model company incubated by mit, and fei-fei li, a spatial intelligence startup company founded by stanford university professor. the underlying model is evolving rapidly, evolving towards spatial models, robotic models, and sequence models.
data infrastructure is also changing. in the past, data infrastructure was represented by databases, but now operating systems are large-scale models, and the data layer is oriented towards large models. vector rag (retrieval enhanced generation) and data quality management have emerged.
the big model provides api capabilities, but jiang tao said that in order to really use it in products, in addition to data chain support, a tool chain is also needed, so the developer's tool chain is undergoing a huge evolution. in terms of applications, openai launched the chatbot mall gpt store. now the industry is betting on agents (intelligent bodies), which requires agent service providers and tool service providers.
the most direct experience for developers is code generation. on the front end, 80 to 90 percent of the work of writing code can be automated, and the programming capabilities of major models have improved rapidly, so the code tools have changed.
talking about ai development code: creative high-level programmers have a better future
in the past, a person could produce 50 or 100 lines of programming per day. after the emergence of ai tools, the output has increased to 500 to 1,000 lines. it is like driving a car instead of pulling a rickshaw. in the past, the amount of code generated restricted the development of digitalization. now ai has reduced the workload, liberated productivity, and expanded the boundaries and scale of digitalization.
the arrival of big models will increase the number of developers. in the future, as long as they can talk, think computationally, and have logic, everyone can write applications and everyone is a developer. however, jiang tao said that being able to write does not mean being able to write articles or books. the same is true for writing code. programmers who understand structure and systems can use ai tools to improve productivity and empower more developers. creative high-level programmers have a better future, and programmers facing the business and demand sides will increase significantly.
however, developers who can only write and translate codes are easily replaced by tools because they can neither rely on business nor systems, so they will face a job crisis, and the 35-year-old crisis will be greatly exacerbated. programmers who design interfaces may disappear, and aesthetic designers who create interactive forms are scarce, so positions for tool manufacturers under large models will be created.
however, the big model is still in the process of iteration, and some capabilities are not available. a group of temporary developers have emerged in the market. jiang tao said that these temporary developers will have to go through the test of time. just like in the early days of the pc era, the operating system capabilities were weak, and third-party software vendors helped develop related functions and gradually grew stronger, but most software vendors died out due to the disappearance of scenarios. the big model will also "eat up" some jobs during its evolution. he suggested that future college students should learn english well and master mathematics, which are the foundation of science, and also learn computational thinking.
talking about model open source: leveraging external development capabilities with small investment
since the beginning of this year, domestic big models have experienced a wave of price cuts. jiang tao said that the investment and return of big models are huge. big models solve the subscription payment model of software and bring about changes in technology and business models. we must seize this opportunity. the core issue that china's open source ecosystem needs to solve is commercialization and marketization. once it is successful, it can give play to the scale and data advantages of programmers.
at the same time, open source models and closed source models will also grow together. he said that at present, the closed source model has a first-mover advantage and has a large number of users and data. in the future, there will be a pattern of 1-2 closed source model manufacturers competing with each other in the market, supporting revenue by providing services. there is only one way for closed source systems in the future, that is, to keep all the business processes and business data of customers firmly in their own hands. the advantage of open source is that it can leverage external development capabilities with small investments, and these external developments will face the survival of the fittest. the capabilities of the open source model will approach the closed source model, just as android gradually narrows the gap with ios.
csdn and huawei cloud jointly launched the ai model community to build china's open source foundation.
the largest model gathering place at present is the english open source community platform hugging face. in order to increase the proportion of intelligent application bases of chinese developers on the csdn platform, csdn and huawei cloud launched the gitcode ai model community. relying on the user scale of csdn and huawei cloud's technical capabilities and resources in ai, it provides chinese developers with one-stop model services such as model selection and computing power call.
jiang tao said that csdn's goal is to become a tool platform for intelligent application productivity. with the popularization of large and small models, process data such as design, generation, and iteration are increasingly modeled, which promotes the upgrade of the tool chain. if you do not master open source tools and developer tools, you will lose the next strategic high ground.
the paper reporter zhang jing
(this article is from the paper. for more original information, please download the "the paper" app)
