Following this point, ChatGPT seemed to have deviated from the intended course. I encountered difficulties in incorporating any additional features into the original plugin, as ChatGPT primarily seemed to generate the shortcode version repeatedly. It is evident that code maintenance is not a core strength of ChatGPT.
After this, ChatGPT seemed to lose focus. It failed to incorporate additional features into the original plugin, often repeating the generation of the shortcode version. Clearly, code maintenance is not a strong suit for ChatGPT.
Undoubtedly, this is a simple plugin. Its size is comparable to the assignments I used to give my first-year programming students when I taught at UC Berkeley. However, it falls far short in complexity compared to the much larger open-source WordPress plugins I maintain, which serve over 50,000 users.
In addition: Should we be seeking answers in different places with AI chatbots?
When I initially utilized ChatGPT to develop a quick plugin for my wife and shared it on my social media channels, I received some pushback from my followers. They expressed various concerns:
Is it potentially malicious? One person worried that the AI may incorporate malicious or promotional information into the plugin, and non-programmer users would be completely unaware. However, the truth is that even if you hire another human coder and can't read code yourself, you still won't know what's inside. Ultimately, it comes down to trusting reputation or having someone you trust review the code. In the case of the code generated during my experiment, it was clean.
Is it the best it can be? Once again, individuals held the AI's code to a higher standard compared to human coders. While I am a skilled programmer, my code contains bugs and isn't necessarily the epitome of perfection from the outset. However, with humans, code improves over time. I'm uncertain if an AI can take existing code, enhance it, debug it, and make it better. Nevertheless, certain projects simply need to be functional. It doesn't always matter if they're the absolute best. Often, good enough is, well, good enough. ChatGPT generated code that met this standard.
Will it reduce the demand for human programming jobs? Most likely. Witnessing a machine take someone's livelihood is always disheartening (especially if it's a job I might desire in the future). AI is undoubtedly a viable option for quick and straightforward projects, whether they involve writing or programming. People's job security will undoubtedly be impacted.
I have been studying AI for decades and even completed a thesis on the subject. However, this current surge in high-quality generative AI has amazed me. Nevertheless, what we are currently witnessing is a level of productivity that is "good enough." Unfortunately, that is precisely what many clients demand.
However, for the time being, ChatGPT and similar models are unable to produce in-depth analytical articles like this one, as they rely on opinions, anecdotes, and years of experience. While they may be able to create subroutines and functions for larger coding projects, they are simply unable to produce entire major coding projects.
In addition: Google's competing ChatGPT demo experiences a mishap.
For instance, over the past year, I have had to attend 20-30 meetings with a major API provider in order to integrate their service into mine. Most of these meetings involved a mix of business matters, technical requirements, and licensing policies. I challenge any AI to endure, comprehend the nuances of, and fulfill the partner requirements within such a bureaucratic environment, while also producing code modules that satisfy everyone involved.
So yes, I was quite impressed with the plugin I "created" for my wife. However, AI still has a long way to go before it can replace experienced developers and writers, as long as clients want the type of work that these seasoned professionals deliver. And that, my friends, is something that genuinely concerns me.