ChatGPT: A Tool for Coding and Debugging - Discover My Top Trick

One of the more fascinating aspects of ChatGPT is its ability to generate code that's actually quite good. I first put this to the test in 2023 when I tasked it with creating a WordPress plugin for my wife's website. The result was impressive, though it was a relatively simple project.

So, how can you incorporate ChatGPT into your daily coding routine? Here's a quick rundown:

1. ChatGPT can produce both useful and unusable code. For the best outcomes, provide clear and detailed prompts.
2. It shines when assisting with specific coding tasks or routines, rather than building entire applications from scratch.
3. Use ChatGPT to discover and select the right coding libraries for specific purposes, engaging in a back-and-forth discussion to refine your choices.
4. Be mindful of who owns AI-generated code and always verify its reliability. Don't just take the output at face value.
5. Treat your interactions with ChatGPT like a conversation. Refine your queries based on the AI's responses to get closer to your desired outcome.

Now, let's dive deeper into using ChatGPT for coding.

What types of coding can ChatGPT do well?

There are two key points to understand about ChatGPT and coding. First, it can indeed generate useful code. But second, it can also get completely sidetracked, producing code that's practically unusable.

I learned this the hard way after successfully creating that WordPress plugin for my wife. Curious to see how far ChatGPT could go, I crafted a detailed prompt for a Mac application, specifying everything from user interface elements to settings interactions. The AI responded with a flood of text and code, but it stopped abruptly mid-code. When I prompted it to continue, it churned out even more code and text, but none of it was usable. It failed to organize the code properly, omitted crucial operations I'd requested, and left placeholders like "program logic goes here."

Through repeated trials, it became clear that asking ChatGPT to build a complete application from scratch is a no-go. If you're new to coding and expect ChatGPT to do all the heavy lifting, you'll likely be disappointed.

Where ChatGPT truly excels is in assisting coders who already know their way around. It's great for crafting specific routines or tackling particular tasks. Instead of asking for a full app, you might request a routine to add a menu to a menu bar, which you can then integrate into your project. The tool will handle that well.

However, remember that while ChatGPT might seem knowledgeable, it lacks the nuanced wisdom that comes from deep, practical experience. It can write code, but it won't capture the subtleties needed for complex problems.

Use ChatGPT to demo techniques, write small algorithms, and create subroutines. You can even ask it to help break down a larger project into manageable chunks and then assist with coding those sections.

How to use ChatGPT to write code

1. Narrow down and sharpen up your request

Before you even ask ChatGPT anything, decide what you want your function or routine to do, what parameters you'll pass in, and what output you expect. Think of it like hiring a human programmer: Are you giving them enough detail to work with, or are you being too vague, leaving room for misinterpretation?

For instance, if you want to summarize web pages, decide that you'll input a URL and expect a summary text block as output. Plan out your request carefully before you proceed.

2. Use ChatGPT to explore libraries and resources

Continuing with our web page summary example, instead of manually extracting text from HTML, you could use an AI library for intelligent extraction and summarization. ChatGPT is excellent at finding and suggesting libraries and resources.

Let's say you want to do this in PHP without paying for API access. Start by asking ChatGPT about open-source AI libraries that can summarize web content. You might ask something like, "Describe ten different open-source AI libraries (and the languages they work with) that I can use to generate a summary of the main core contents of any web page, ignoring any ads or embedded materials."

ChatGPT will provide you with options, and you can refine your search by asking follow-up questions, like, "Are any of these free?" or "Which of these libraries have no usage limits and don't require any additional payment or licensing?"

If you're looking for PHP compatibility, you might ask, "Of those 8 libraries, can I use any with PHP?" Then, to understand the differences between the suggested libraries, ask, "What's the difference between Sumy, Gensim, and NLTK?"

Finally, to pinpoint the best choice for your needs, you could ask, "If I want to create summaries of web page news articles, which library would work better?" Once you've decided on a library, ask how to integrate it, like, "Can you explain how to use Sumy from PHP?"

This conversational approach is a form of programming. It's about integrating resources and understanding how to communicate with different components of your solution. ChatGPT helped analyze and plan the integration, which is a crucial part of coding.

3. Ask ChatGPT to write example code

When we talk about using ChatGPT to "write code," we're really talking about getting it to write example code. It's not about producing your final, polished code. Even with a small function, like the line sorter/randomizer I made for my wife, ChatGPT can't maintain or modify the code it's generated. If you need fresh code, you'll have to ask for it anew, as the AI struggles to make tweaks to existing code.

For example, I asked ChatGPT to create a PHP function called "summarize_article" that takes a URL as input and returns a 50-word summary of a news article. The function would use the Sumy library to extract and summarize the main body of the article, ignoring ads and embedded materials.

The resulting code was straightforward, using the Goose library to retrieve the article contents and passing it to Sumy for summarization. But remember, this is just a starting point. You'll need to refine, customize, and integrate it into your project, which is a typical part of any coding process.

4. Debug and refine the generated code

ChatGPT, like all programmers, can make mistakes. But you can use other AI sessions or different chatbots to help debug the code. For instance, you could feed the generated code into another ChatGPT session and ask, "What's wrong with this code?"

The AI might point out potential edge cases, error checks, or situations that could break the code under certain conditions. You can then adjust your code to handle these scenarios, making it more robust and reliable.

FAQs

Does ChatGPT replace programmers?

Not yet. ChatGPT codes at the level of a talented first-year programming student, but it's lazy. It might reduce the need for entry-level programmers, but currently, it's more of a time-saver for existing programmers. It can help with writing code and looking up information, but it can't handle many programming tasks independently.

How do I get coding answers in ChatGPT?

Simply ask it. Use an interactive dialogue to narrow down your answers. Don't expect one question to magically do all your work. Treat the AI as a helper and resource, and it will provide a lot of useful information. Just remember to test that information, as AI can sometimes fabricate things.

Is the code generated by ChatGPT guaranteed to be error-free?

Absolutely not. No code, whether written by humans or AI, is guaranteed to be error-free. All code needs thorough testing, alpha and beta testing, and even then, bugs can still slip through. Always verify the code generated by ChatGPT.

What do I do if the code I get back is wrong?

Treat ChatGPT like a slightly uncooperative student or employee. If the code doesn't work, tell it, "That didn't work. Please try again." Often, the AI will provide different variations on the same problem. If it can't get it right after a few tries, you might need to code it yourself. But usually, especially with basic coding tasks, it gets the job done.

How detailed should my description of a programming issue be when asking ChatGPT?

Be detailed. The more specific your instructions, the better the AI can help. Think of it as assigning a task to a student or employee. Give enough detail so they can create a first draft without needing to ask too many questions. The more guidance you provide, the more likely ChatGPT will produce something useful.

If I use ChatGPT to write my code, who owns it?

The legal landscape around AI-generated code is still evolving. In the US, Canada, and the UK, copyright requires human authorship, so AI-generated code may not be copyrightable. There are also concerns about liability based on the training data and how the code is used. For more detailed insights, refer to ZDNET's articles on this topic.

What programming languages does ChatGPT know?

ChatGPT knows most languages, from modern ones like PHP, Python, and Java to older ones like COBOL and Fortran. I've even tested it with obscure languages like LISP and IBM/360 assembly language. It's quite versatile.

Can ChatGPT help me with data analysis and visualization tasks?

Yes, it can handle a lot of data analysis and visualization without code. For instance, you can ask it to create charts and tables directly. If you need code, you can specify the language and data source, like asking for a bar chart in Swift using online population data.

How does ChatGPT handle differences between dialects and implementations?

While we don't have exact details from OpenAI, we know that ChatGPT's training data may not always keep up with the latest changes in programming languages. More recent changes and less popular languages might be less well-represented. Keep these points in mind: the more recent the change, the less likely ChatGPT knows about it, and the more popular the language, the more accurate it's likely to be.

The bottom line? ChatGPT can be a valuable tool in your coding arsenal. Just don't expect it to perform miracles. Yet.