With the popularity of ChatGPT, many media began to report that "AI is about to take away the jobs of programmers" and " AI is coming! All leaders have begun to lay off staff substantially" , which made some junior coders worry, while some senior programmers It expresses disdain. Is this "crying wolf". Today, we will conduct a practical test to explore the performance of ChatGPT when dealing with daily development work.
Test objective: Can ChatGPT cope with development work.
Test content:
1. Understand business and product requirements
2. Write business logic code
3. Solve the problems in the code
The test starts:
1. Understand business and product needs
Simulation scene:
The big boss of the company was furious about the frequent overturning of the car during the business process. It is required to investigate and audit immediately, and incorporate the problems found into system management. The person in charge of the business led the order to the development department, and made up his mind: the disaster must be brought to the east, and the blame should be given to the development department. "Help me make a form to submit for audit and troubleshooting, huh? What fields do I need? The overall process? Now that the boss is going to kill someone, let's collect the problems first."
ChatGPT Appearance
The ability of ChatGPT is amazing. Without providing a business description, the design is made in seconds only by relying on the form name. This usually requires programmers to have a level of senior industry experience. Further ChatGPT also made specific instructions and guidance for the business. "How can AI understand SB's business needs?" The remarks are now self-defeating, and ChatGPT may understand your business better than the entire team.
Summary: In the demand understanding process, there is no need for complex description and communication like manual docking. Most of the time, AI understands the needs and results better than the business, and can even guide the business in reverse. From the perspective of the demand side, the greater role of developers is to take the blame.
2. Business code writing
Simulation scene:
As a newly recruited grassroots intern programmer, today's task is:
- Delete the comments in the code delivered to Party A to ensure that subsequent development orders can be obtained.
- Learn the old code!_! "Leader! There are no comments in this code! Okay, let me try it"
- Complete function development. .
- Secretly write a graduation thesis.
Task 1: Sneakily remove some comments from the release. Deleting comments is a trick.
Result: success
Task 2, understand the obfuscated code without comments
Result: success
Task 3: Write a query in python, and display the factory personnel examination schedule based on the Gantt chart. The data source is the execl manually compiled by the factory director
Result: success
Summary: ChatGPT can directly convert requirements into code, and can also understand and analyze the logic and content of existing code.
3. Solve the code error
Scenario: In fact, writing code can be copied and pasted. To reflect the programmer's skill, it depends on code debugging and error reporting.
I found a few programs to report errors, and let ChatGPT analyze the cause of the error.
Conclusion: Based on security considerations, the source code is generally not given to AI for analysis, and usually only an error report is given to find the cause of the error. Therefore, programmers still need to correct and debug according to the problems given by AI. But if the AI is a privately deployed privately trained AI. Then it is possible to realize that the writing and debugging of the code are all done by AI.