Using ChatGPT to Rethink Your Code

Using ChatGPT to Rethink Your Code

With the advancement of artificial intelligence technology has come the promise of quick, simple, and competent AI tools and assistants. While advanced AI certainly comes with a mixed bag of feelings for many, there is still excitement behind the power. Also, the potential AI brings to automate and streamline many tedious processes that take up valuable time and energy in day-to-day tasks.

One discipline in which this optimizing power is particularly exciting is programming. Any programmer knows that, despite making for a cool movie scene, most projects involve a lot more research and debugging than just sitting in front of the terminal and cranking out line after line of perfect code. With the artificial intelligence aid, however, this process might become much less tedious.

Assisted Research and Development

True in more applications beyond programming, research is a process whose length and difficulty often depends on a bit of luck. While you’re looking for solutions, how long your search will take depends largely on whether you’re using the right keywords and clicking on the right links. A huge portion of the development process is finding the right resources from which to learn and implement more concepts and ideas. Having an AI assistant to shift through countless sources and provide you with quick, easy to understand summaries and explanations has the potential to decimate that lengthy process.

Further, an AI chat like ChatGPT doesn’t just find sources for you, but it can break down and present content for you in a variety of different ways, including tonal preferences in GPT-4. If you don’t understand a certain concept, ask ChatGPT to explain it in different, more digestible ways. In a big sense, this adds a fundamental pedagogical element to a traditionally difficult self-teaching space.

Limitations and Cautions

Of course, AI chatbots like ChatGPT don’t just trivialize the work that needs to be done. There are plenty of limitations to keep in mind. One big one, across any ChatGPT application, is that the algorithm is trained on limited data. Specifically, the chat will have no idea about information created after a certain point in time. Programming is a discipline that changes and grows each and every day. It is perfectly plausible that the correct or most efficient package to be used was released after ChatGPT’s training cutoff, which means that over-reliance on the AI responses might produce suboptimal results.

Another limitation is ChatGPT’s current questionable accuracy. One of the main improvements announced in the newest iteration, GPT-4, is that the AI is significantly more likely to produce factual information, but this iteration is still not available to the public (at the time this article was written). There are a lot of reasons why an AI chat might produce inaccurate information, including errors in training data and difficulty with language interpretation, but the moral to learn is the same: always proceed with caution! ChatGPT can produce great code and save you the time digging through GitHub. It is still your responsibility as a programmer to make sure that the code runs correctly at all, and also that it is appropriately implemented in your specific context. Clean, consistent, and uniform code might not be a deciding factor during compilation, but it saves lives during necessary and inevitable revisions and changes.

Staying Behind the Wheel

For the same limitations mentioned above and more, it’s important to mention also that these impressive and groundbreaking advances in artificial intelligence are not removing humans from the process. For one, AI can help optimize and accelerate many processes, but apps like ChatGPT are not autonomous. Plus, fully automated processes already exist, and we humans have not been edged out of the labor force quite yet. ChatGPT won’t be this deciding factor, if one even exists.

More importantly, as ChatGPT and other AI apps continue to improve, there is still plenty of information that they aren’t able to pick up on. Any developer who has coded in C++ across different devices and platforms might have run into annoying issues. Slight differences in installed versions and packages that make the same code runnable on some devices and entirely broken on other devices, to name a couple. As mentioned before, even if the chat can produce perfectly functioning and legible code, your project shouldn’t just look like a patchwork of different generated lines in different styles and formats.

These are small but crucial details specific to not only your personal coding style, but the culture of your workplace and your context. Even if ChatGPT can do everything else perfectly—which it cannot—contextualizing the content appropriately and cleanly is something that is 100% in your hands. At the end of the day, ChatGPT is a tool and, as with any tool, it is only as useful as the hand that wields it!