A Developer’s Thoughts on OpenAI’s ChatGPT

by

If you’ve been living under a rock lately, it’s been hard to ignore the conversations online about 🔗ChatGPT. It’s been everywhere and it’s (rightfully) blowing people’s minds. Today I would like to share a developer’s thoughts on OpenAI’s ChatGPT.

First off, what is ChatGPT? It’s an AI chatbot by 🔗OpenAI, who have trained an AI model (🔗read more about AI models here) called ChatGPT. What’s amazing about it, is that it interacts in a eerily exceptional conversational way. The kicker is that it can write code for you too.

Why is it’s ability to write code impressing developers worldwide? Because it’s doing it impressively effortlessly and quickly (although sometimes with questionable accuracy depending on your inputted wording). This is a very big deal for obvious reasons. It’s also a momentous milestone in just how far AI has come.

During this research preview period, OpenAI has opened up the ChatGPT tool 🔗for free to get users’ feedback and learn about its strengths and weaknesses. I’ve been playing around with it myself, and the results have been blowing me away.

Here is an example of a request I had for it: “write a php script that converts a CSV file to JSON and uploads it to a s3 bucket”

Within seconds a fairly decent PHP script was generated. In my books, fairly decent is a good starting point for something that took just seconds to create.

I then wanted to improve this output by adding some basic error handling and thus modified my statement to; “write a php script that converts a CSV file to JSON and uploads it to a s3 bucket with error handling“

A few seconds later an improved version of the script was generated for me. Colour me extremely impressed! To be clear, I would still make a few improvements to it, but as starting point and as a developer, this saves a massive amount of time.

I decided to take a look around on Reddit to get other developer’s sentiments towards ChatGPT at the moment and the 🔗sentiment seem shared – it’s blowing developer’s minds.

People have been able to get it to do some pretty wild things already like 🔗building and running a virtual machine inside of ChatGPT or 🔗giving it full access to a terminal on their computer. It’s all very exciting and it certainly seems like we’re on the precipice of a revolution akin to something of that like of the mass-adoption of the internet.

In fact, ChatGPT has become so popular, developer help site 🔗Stack Overflow has had to start banning people posting AI-generated replies citing questionable quality of responses. This is a fair concern by the site and brings me to my next point; accuracy of ChatGPT.

At the moment, while certainly impressive, ChatGPT is not perfect. I would even say it’s far from it as it currently stands. It makes mistakes and it still needs a human developer to read the code, implement and test it. The model has been trained on a sub-set of biased data which may or may not have been correct to begin with and there lies the inherent problem with AI at the moment; biases and only being as good as the data it is fed.

Despite that flaw (for now), thoughts can easily run wild with what the future may look like with such a tool at our fingertips as developers, especially as it grows more powerful and refined.

‘Is this going to replace developers one day? Is this going to put developers out of a job soon?’ – those would be the natural thoughts of any intelligent or sane mind. My honest answer to that would be that I doubt it.

Reasons being; I foresee developers working side-by-side with whatever tools ultimately adopt or integrate the ChatGPT API, augmenting our jobs to help with the small to medium tedious tasks, or even assisting on micro-pieces of a much larger task.

At the end of the day you will still need a human developer to ultimately perform empathetic, moral and intuitional decisions that only human consciousness can make.

Resources:

Leave a comment