"Human programmers will disappear completely within five years." This was the prediction made by former Stability AI founder Emad Mostaque last summer. In addition, Baidu founder Robin Li also said that there will be no such profession as "programmers" in the future. . Although another programmer-turned-tycoon Zhou Hongyi does not believe that AI will replace programmers, programmers seem to have really developed a Terminator that may "optimize" themselves.
Just after Baidu announced that up to 27% of the code added every day is automatically generated by the Comate intelligent code assistant, Alibaba's No. 1 AI employee was hired, and AI programmer Tongyi Lingma will be responsible for producing 20% of Alibaba Cloud code in the future. , overseas Google has also involved AI in code writing. In the 2024 third quarter earnings conference call recently held by Google parent company Alphabet, CEO Sundar Pichai revealed that more than 1/4 of Google's new code is now generated by AI, which is then reviewed and reviewed by engineers. acceptance.
Sundar Pichai said, "This is an important milestone and marks the importance of artificial intelligence to the company." However, you must know that just a year ago, when AI programming first came out, the mainstream voice in the online community, whether it was domestic Zhihu or overseas Reddit, was that AI programming tools such as Github Copilot could not impact the status of programmers, and believed that The so-called "programmers will be replaced by AI" is nonsense. When Robin Li and Zhou Hongyi were "debating" about this, the voices supporting the "red leader" also showed an overwhelming advantage.
However, one year later, the reality is that major companies such as Alibaba, Baidu, and Google have deeply involved AI programming in code writing work. In fact, this is not that Zhou Hongyi and the big Vs in the programming field of Zhihu and Reddit were slapped in the face. They did not misjudge the situation, but were just fans of the authorities. Nowadays, the impact of AI-generated code is not on programmers who are willing to share and discuss on Zhihu and Reddit, but on low-level coders, just like Google Translate eliminated basic translation positions in the early years.
Because the current AI programming tools show the same characteristics, that is, the cost-effectiveness ratio is low. If a programmer can accurately describe the requirements he wants to achieve to AI, then the time it takes for him to describe the code to AI is usually much longer than the time it takes to write the code himself. For serious programmers, current AI programming tools are only "auxiliaries" rather than "replacements", but at the enterprise level, this level of AI programming is enough.
In fact, the nature of the work of programmers is somewhat similar to that of translators, except that translators are responsible for communication between people, while programmers are responsible for communication between humans and computers. Simply put, the programmer's job is to translate real-world needs into source code (SourceCode) through computer language, then convert it through a compiler into machine code (NativeCode) that the CPU can directly process, and finally form an executable program. and maintain the operation of the program in the future.
Obviously not all programmers are online. To be precise, the difference between programmers is even greater than that between programmers and non-programmers. In addition to high-level experts who can quickly come up with iterable minimum viable products and understand why, low-level coders are almost simply porters of code.
Thanks to the emergence of the Internet, the explosion of information has given low-level programmers the opportunity to make up their jobs. At the same time, the rapid development of the Internet industry for three decades has also created a massive demand for programmers. So much so that programming for search engines and programming for GitHub have become famous in the programmer circle. The actual work of low-level coders every day is often to transfer code to GitHub. If you can use Ctrl+C and Ctrl+V, you can become a programmer. The ridicule is not groundless.
Take Alibaba's AI programmer "Tongyi Lingma" as an example. Alibaba introduces it like this. Under the traditional development model, programmers need to spend a lot of energy writing repetitive code, debugging and optimizing, and writing code comments every day. These significantly Squeezing time for core business code writing. Since Tongyi Lingma "joined", he has served as a "code assistant" in various development links within Alibaba Cloud, including assisting engineers in the entire process of adding, deleting, checking, and modifying.
The real reality today is that a high-level programmer paired with AI programming tools will eliminate ten junior coders who can only copy and paste. Kent Beck, an advocate of agile software development, once posted on a social platform, "I reluctantly tried to write code with AI, only to find that it can replace 90% of my skills, but it can Let the remaining 10% of my skills be amplified a thousand times.”
In fact, AI programming tools do not completely replace the ecological niche of programmers, but liberate them from the tedious work of writing codes, allowing them to focus on more creative work, such as designing more efficient algorithms and solving complex problems. technical issues and developing new products. As a result, AI programming tools have quickly become a powerful tool for large manufacturers to reduce costs and increase efficiency, and more than a quarter of Google's new code is generated by AI, which means that they will have higher flexibility in terms of labor costs. sex.
Then the problem arises, because AI will evolve. Although the current Github Copilot can only be used as an "auxiliary", users need to clearly know the details and meaning of each line of code generated by AI before the relevant code can be considered usable, but If companies such as Microsoft, Alibaba, and Google continue to iterate their own AI programming tools, it will be possible for AI to generate and compile a complete software code end-to-end in the future.
At this stage, the "bastards" who can only copy and paste from GitHub are beginning to be squeezed by AI programming tools. As AI becomes more and more powerful, the competitive pressure on programmers will naturally increase. The group that has reaped the biggest development dividends in the Internet era will inevitably have a difficult time in the future. Of course, the devaluation of knowledge and technology itself is an inevitable result of its continuous development. This phenomenon has been reflected in several fields before.
In 2024, "everything can be turned into a computer" is no longer a universal option. Even if the original "low code" failed to squeeze the living space of programmers, AI programming can now truly make it possible for everyone to The coders in the factory feel the cold.