The first AI programmer in China joined Alibaba Cloud: the exclusive employee number AI001, and the KPI is that one person writes 20% of the company's code
Reported by the Heart of the Machine
Author: Egg Sauce
You must have heard the saying: learning computers starts with a baby.
Learning to code used to be a fad for many years, but with the growth of generative AI technology, the tech world seems to have taken a turn for the better.
NVIDIA CEO Jensen Huang said, "In the future, programming will be left to AI, and everyone will be a software engineer in the future. Soon after, Devin, the world's first AI software engineer, shocked developers with a million points, and really sparked a heated discussion about the future of the programmer profession. Could it be that the job of the coder is really going to be taken away by AI? In fact, AI is still just a work assistant for programmers.
In fact, there is an AI programmer in China who has been working in a major Internet company for a while.
It is the new colleague that tens of thousands of Alibaba Cloud engineers have been dealing with frequently recently-"Tongyi Lingcode", with the exclusive job number "AI001".
Before officially joining the company, Tongyi Lingma first "interned" inside and outside Alibaba Cloud for a few months. Everyone's reviews are very good, and the real praise rate is over 80%.
In the traditional development model, programmers need to spend a lot of energy every day writing repetitive code, debugging and optimization, and writing code comments, which greatly squeezes the time spent on writing core business code. However, since Tongyi Lingcode joined the company, it has played the role of "code assistant" in various development links within Alibaba Cloud, including assisting engineers in the whole process of writing code, reading code, checking bugs, and optimizing code.
Moreover, the working rhythm of Tongyi Lingcode is even more extreme than that of 996: 7x24 hours on call......
After a few months of internship, Tongyi Lingma has saved a lot of "work results". Taking API development and testing as an example, Tongyi Lingcode can shorten the time spent on manual writing tests from minutes to seconds, saving more than 70% of the test code workload of human programmers.
Alibaba Cloud also announced again today that Tongyi Lingcode will be launched on the Tongyi Qianwen APP, which is free and open to the whole people, and everyone can write code, read code, and learn programming skills on their mobile phones anytime, anywhere.
AI programmers are full of skills, focusing on an all-round assistance
If Tongyi Lingma has a resume, it must be full of shining points.
Jointly developed by Alibaba Cloud and Tongyi Labs, Tongyi Lingcode has been downloaded more than 2 million times and millions of lines of code are adopted by human programmers every day.
Speaking of "professional", Tongyi Lingcode has a solid programming foundation, in-depth study of trillions of code data, and is currently proficient in more than 200 programming languages, proficient in 16 of them, including Java, Python, Go, JavaScript, TypeScript, C/C++, C# and other mainstream languages.
When it was first released at the Apsara Conference last year, many of the skills of Tongyi Lingcode were impressive, such as line-level/function-level real-time continuation, natural language code generation, unit test generation, code optimization, comment generation, code interpretation and other functions.
Give it a piece of project code and let it continue to write the code according to the context:
Even unit tests can be written quickly:
If you're tired of writing code comments, let it do the work for you:
At the same time, Tongyi Lingcode also has the ability to develop intelligent Q&A, abnormal error troubleshooting, etc., which reduces the complexity of the original investigation work:
This time, Tongyi Lingcode, which officially joined Alibaba Cloud, has added the "Code Optimization" function, and the upgraded Q&A model is also more accurate, and at the same time, it has enhanced the capabilities of cross-file completion and function completion.
Thanks to the help of Tongyi Lingcode, the work efficiency of employees in multiple technical positions such as front-end, back-end, and operation and maintenance has been greatly improved, especially in tasks such as coding unit testing, building script generation and understanding, and cross-language code conversion.
Another advantage of Tongyi Lingcode is that it supports special training of enterprise-level code base, which can better understand the code and document data of enterprises, and supports localized adaptation. In the future, Alibaba Cloud will also fully configure Tongyi Lingcode plug-ins in internal development tools such as JetBrains IDEs and Visual Studio Code for all employees to use.
Talk is cheap，Show me the code
Having said all this, whether the Tongyi spiritual code can be compared to a human programmer still depends on the actual combat effect.
In the development phase, let's take a look at the new "Code Optimization" capabilities:
There is also real-time continuation of the line/function code:
Video link: https://mp.weixin.qq.com/s/t_OJu1JGrzk7q8A2GAfA0g
In the "test" link, Tongyi Lingcode is very efficient in generating test frameworks and test cases, and human programmers need to work for a few minutes or even ten minutes, which is only a matter of seconds.
Video link: https://mp.weixin.qq.com/s/t_OJu1JGrzk7q8A2GAfA0g
Test framework build
Video link: https://mp.weixin.qq.com/s/t_OJu1JGrzk7q8A2GAfA0g
Generate test cases
In addition, the efficiency of writing O&M scripts is much higher than that of humans, and it can also help find solutions to O&M problems at any time:
Video link: https://mp.weixin.qq.com/s/t_OJu1JGrzk7q8A2GAfA0g
During the process, developers can ask any questions, such as "I want to use a database to store photos, how to do it":
Video link: https://mp.weixin.qq.com/s/t_OJu1JGrzk7q8A2GAfA0g
Another example is "How to write a web program for uploading and downloading photos through Spring-Boot":
Video link: https://mp.weixin.qq.com/s/t_OJu1JGrzk7q8A2GAfA0g
To use a buzzword to describe it, Tongyi Lingma is an "all-round employee". Whether it is development, testing, or O&M scenarios, it can show strong strength.
AI 越来越强,抢走饭碗 or 利好码农?
Of course, we are also curious, why is the level of Tongyi Spirit Code so high?
The partners within Alibaba Cloud revealed a few points to us: in essence, Tongyi Lingcode can change from a programming assistant to an all-round assistant, which is inseparable from the profound strength of the Tongyi model.
The first is "context-aware, adaptive generation of granular decisions". Tongyi Lingcode has the ability to generate correct code content (word level, line level, block level, function level, class level, etc.) according to the language type and context scenario, which can not only generate long code for developers, improve coding efficiency, but also generate short codes in a timely manner to reduce interruptions.
The second is "code generation based on cross-file automatic perception in the library". In order to solve the problem of illusion, Tongyi Lingcode analyzes the code currently being written in real time through advanced device-side real-time code semantic analysis technology, and obtains the required semantic information such as relevant code references and similar code fragments based on advanced technical methods such as code semantic analysis, code reference link tracing, dynamic language type derivation, and similar code analysis, so as to make up for the lack of simply paying attention to the current code file.
At the same time, Tongyi Lingcode can "generate code and answer questions based on search enhancement". Specifically, it obtains the private domain knowledge in the retrieval engine in real time through retrieval augmentation technology, so that the intelligent coding assistant can generate code or answer questions based on enterprise knowledge, and enterprises can update and take effect of private domain knowledge in real time, so that they can always generate or answer questions based on the latest private domain knowledge.
What makes us feel is that from Microsoft's previously launched programming assistant Copilot to AI programmer Devin, AI has brought waves of impact to programmers. This revolutionary progress took just over two years.
The official inauguration of Tongyi Lingcode in Alibaba Cloud marks that programming has taken the lead in becoming one of the fields where large-scale model technology has landed. Since then, the boundaries of programming languages have been broken, and Java programmers can easily write code in different programming languages such as Python and C. The question is, will more programmers be taken away by AI?
Perhaps there is no need to be too pessimistic: just as the advent of the automobile has left some coachmen without jobs, it has created a new profession as a driver. The relationship between human programmers and AI programmers should be one of mutual promotion and mutual learning. As Kent Beck of Software Development Methodology says, with AI tools, 10% of our skills will be improved by 1,000 times in the future.
For example, in the future, Tongyi Lingcode, which quickly integrates into the team, will take on more difficult tasks, which is definitely good news for Alibaba Cloud engineers:
"In the future, 20% of the company's code will be written by Tongyi Lingcode, but programmers are still at the core of R&D, and they will have more time to focus on system design and core business development. 」
For ordinary people, there is no need to download professional software in the future, and they can write and learn code using Tongyi Qianwen. For students who have no coding foundation, Qianwen can provide study suggestions, such as helping users develop a complete set of plans for learning Python from both theoretical and practical levels.
What are your thoughts on AI programmers?