Recently, a study led by the Department of Computer Science at the University of Copenhagen published a paper titled 'The Impact of Large Language Models on Programming Labor Markets' on the arXiv preprint platform. The research team analyzed public data from 150000 active GitHub developers, 50000 freelance programmers, and 500 technology companies between 2021 and 2023, and found that AI programming assistants (such as GitHub Copilot, ChatGPT, etc.) did not have a "one size fits all" impact on "coding jobs," but rather exhibited a "layered effect.

Research finds

Entry level tasks are the first to be 'eaten'. The order volume for "low complexity" requirements such as a large number of repetitive scripts, unit tests, and front-end page templates has decreased by 12% -18%, and the average unit price has dropped by 8%.

Senior positions are temporarily safe. The demand for "high complexity" involving system architecture, performance optimization, and security auditing has increased instead of decreased, with a quotation increase of about 5%.

The recruitment end of enterprises is still expanding. Despite the popularity of AI tools exceeding 60%, the number of new programmer positions added by surveyed companies in 2023 still increased by 7% year-on-year, mainly concentrated in AI infrastructure, data engineering, and algorithm security.

The developer's skill stack is being shuffled. Programmers who use AI assistance allocate 25% of their "mechanical code typing" time to requirement analysis, code review, and operational automation, resulting in a 28% increase in average delivery efficiency per person.

Industry voices

Zhou Ming (a pseudonym), the chief researcher of Microsoft Research Asia, said: "AI is not a substitute, but a way of turning 'writing code' into 'reviewing code+debugging model'. In the next five years, compound talents - who understand both business and algorithms - will be the hard currency. ”

Ma Yue, founder of Open Source China (pseudonym), believes that for small and medium-sized teams, AI reduces the cost of trial and error, and may instead stimulate more application innovation, thereby creating new job demands. ”

Regulatory and Compliance

In response to the copyright, security, and compliance risks that may arise from AI generated code, the China Academy of Information and Communications Technology has launched the development of industry standards for "Technical Requirements and Testing Methods for Intelligent Programming Tools" in August this year, and is expected to release a draft for soliciting opinions within the year. The draft proposes hard clauses such as "traceable source of generated code" and "prohibition of direct copying of GPL code snippets" to ensure that AI assisted development operates within the legal framework.

Conclusion

In the short term, AI is more like an efficiency multiplier than a job harvester. But researchers remind:

Junior developers need to transition to a "business+algorithm" composite route as soon as possible;

Universities and training institutions should include AI assisted development as a mandatory course;

Enterprises should establish a "human-machine collaboration" process to avoid AI output going live without audit.

As the author of the paper said, "Programmers will not disappear, but people who only know how to write if else may really disappear