AI code tools provide benefits to us, but they also bring some dangers

图片1.png

Some things you need to know

Artificial intelligence is rapidly integrating into our daily lives, bringing many exciting changes.

We can say that we have entered an era of fully utilizing AI-driven technology. Improving the efficiency and quality of software development is an important link in promoting technological innovation and application. To satisfy human nature's desire to "be lazy", the market is constantly developing new AI code tools,such as XXAI for translation and writing, Google's AI assistant for office work support, and Adobe's AI generator for image processing.

To achieve frictionless, intelligent application building, clever humans have thought of using AI tools and services to simplify the development toolchain. Major AI companies are continuously developing and upgrading AI code tools. Recently, it was learned that Anthropic has also followed the market trend by launching an AI code tool for Claude - a JavaScript sandbox tool. This functionality is currently available to all users as a preview version.

The use of AI code tools can lower the entry barriers for software development and enhance development efficiency. However, the ease of software development brought by AI code tools should also draw our attention to the importance of using AI in a regulated manner.

What can AI code tools be used for?

Some might wonder, what exactly are AI code tools useful for? What functions do they have?

图片2.png

Specific functions are as follows:

  1. Automatic code generation: Automatically generate code based on user requirements, reducing repetitive work.
  2. Code completion and suggestions: AI tools can provide intelligent completion and real-time suggestions.
  3. Error detection and debugging: Automatically detect potential errors and vulnerabilities, and provide fix suggestions.
  4. Documentation and comment generation: Automatically generate relevant documentation and comments based on the code, improving code readability and maintainability.
  5. Learning and training: For beginners, provide real-time feedback and guidance.
  6. Code refactoring: Analyze code structure, suggest optimization and refactoring solutions, improve code performance and maintainability.

Taking Claude AI as an example, Anthropic has further enhanced Claude's code writing capabilities by launching an AI code tool for Claude AI, allowing users to write JavaScript code using Claude, upload their own code to a sandbox environment, and automatically generate running results, helping programmers or business people debug code or perform precise and efficient calculations.

AI code tools - A developer's godsend

Google released the 2024 State of DevOps report, stating that over 75% of developers and IT professionals rely on artificial intelligence to handle at least one daily professional task. Google Cloud's DevOps Research and Assessment team surveyed nearly 3,000 technology professionals globally, indicating that more and more people are starting to use AI technology.

图片3.png

Specific data from the status report is as follows:

  • 76% of respondents said they rely on AI to complete tasks such as code writing, information summarization, and code explanation
  • 81% of respondents said their companies have shifted technical resources towards developing artificial intelligence
  • At least 67% of people said that artificial intelligence is helping them improve code, enhance workflow, productivity, job satisfaction, code quality, and internal documentation quality.
  • Documentation quality improved by 7.5%
  • Code quality improved by 3.4%
  • Code review speed increased by 3.1%
  • Approval speed increased by 1.3%
  • Code complexity reduced by 1.8%
  • 25% of enterprises are making dedicated strategic investments to build AI capabilities, expected to grow by over 20% in the next six months.
  • More and more enterprises tend to utilize internal data and small customized language models, focusing on efficiency rather than pure computational power.

Using AI code tools as auxiliary tools to improve productivity

图片4.png

Although AI has brought productivity improvements, research shows that AI adoption may lead to a decline in software delivery performance, with delivery throughput expected to decrease by 1.5% and delivery stability decreasing by 7.2%.

Matt Hoffman, product manager and data analyst at Uplevel, said. A GitHub survey released in August showed that 97% of software engineers, developers, and programmers reported using AI programming assistants.

"We see different studies showing people saying 'this is very helpful for our productivity', but we also see some people saying, 'you know what? I've actually become more of a (code) reviewer.'" Hoffman said.

This also serves as a warning to us. While we cannot deny the significant role AI code tools play in improving development efficiency, we must also realize that we need to use AI tools in a regulated manner, placing AI code tools in an auxiliary position, reducing our dependence on them, enhancing our independent development capabilities, and maintaining initiative.

Japanese man arrested for improper use of AI code tools

图片5.png

We must also acknowledge that there are not many restrictions on the use of AI code tools! In addition to developers being able to use AI code tools, even a person without IT professional knowledge can use AI code tools to develop programs and write code.

A "smart guy" in Japan used "unofficial versions of ChatGPT" on computers and smartphones to create program source code that encrypts specific files and makes them unusable, and demands payment in the cryptocurrency "Bitcoin". He created a computer virus similar to ransomware by modifying and merging the above code.

For misusing generative AI to create computer viruses, he was sentenced to three years in prison, suspended for four years, by the Tokyo District Court.

The defendant stated during the inquiry on the 1st of this month that it took about a month to complete the virus. He admitted that he had no IT knowledge, saying, "I don't think I could have made it without generative AI." When the defense attorney asked if it was "easy", he answered "yes".

This also shows how powerful the AI code tools on the market now are, even a novice without IT knowledge can use tools to program the applications they need. However, we do not advocate or support such illegal behavior, tools should be used for the right purposes!