ChatGPT, Ensign Data running errands

Hello, my name is Lisa, I’m a web developer in the marketing department ispmanager. I will share my experience here from GPT chat – a neural network that facilitates writing code, writes posts and gives short answers to any questions.

The superpower of GPT is that it remains in the context of the request: tasks can be supplemented or clarified without repeating what has already been said from scratch. I wanted to test how chat would handle a request to create a complete page with markup, styles, and script, and not just a separate function. The service does not work in Russia, but there are already enough articles on the Internet that tell how to get through to the popular neural network. Requests are best formulated in English, with which GPT works best. In addition, it will help to practice building phrases in English.

First, I asked in a chat what kind of web development tasks GPT can help me with.

I asked GPT what web development tasks chat can help me with, and the neural network listed a number of possibilities

GPT promises to help with the design and development of the site from scratch, which is reassuring. Perhaps even too much. Debugging, feature extension, API integration, optimization, SEO… And the chat will also tell you how to do everything right. Now someone resourceful will decide that it’s time to fire all web developers and use the services of a neural network: she doesn’t need to feed her children, well, and pay a mortgage too, hmm.

I looked at what requests the authors of articles and videos had already tested: there were tasks to write a specific function, use the API and display the result on a web page. All this in steps, with gradual complication, and everything looked very smooth. I decided to ask for everything at once, namely: a web page with scripts and styles. Such is an attempt to replace a web developer with a neural network on simple tasks. Just like in articles and videos with clickbait headlines: “Programmers are no longer needed.” Here we will check.

Jarvis, get to work

Let’s create a page with an image and a video start button. On click, the image and the button should be hidden, and in their place, the video should be displayed and launched.

I formulated a request in English and gave it to my new free web developer – now he will quickly make a one-page website. At the first step, I asked to create an image, on top of which there will be a video start button.

In response to a request to generate an HTML page, I received markup and styles - all in one file — In response to a request to generate an HTML page, I received markup and styles – all in one file

And indeed, I get an HTML page and a styled button. It remains only to add an image file next to it or ask the neural network to pick up some specific one.

I decided to find the picture myself, so as not to explain for a long time what exactly I need. So far, a regular search engine is more familiar and familiar to me than a chat, but I admit that soon everything will be different. Perhaps in a couple of years, neural networks and voice assistants will give users answers more often than the search string we are used to.

I copied the generated HTML code into a file, added an image next to it and opened it in the browser. That's what happened — I copied the generated HTML code into a file, added an image next to it and opened it in the browser. That’s what happened

The intermediate result is quite ready: there is both an image and a button with basic styles, all the code is collected in one file. A little later I’ll try to ask you to split HTML, CSS and JS into separate files. But now I’m moving on to the second iteration and I ask you to make sure that the video starts on click.

In the second step, I ask you to add a video and launch it by clicking on the button. GPT added functions to the JavaScript file and included the YouTube Player API library

GPT chat comments that you should use the YouTube API, and completes the code: the necessary video is written here, and JavaScript is in place. It remains to be seen how it works. And here the resourceful comrade who fired the web developers would start to get nervous: what happened was not what the person who set the task expected, but what the neural network understood.

The image is there, the start button too. But the video is just under the picture, not hidden. And by clicking on the button, the video starts playing. But that’s not exactly what I wanted. Now you have to explain what needs to be fixed, or manually edit the code so that the task is completed completely.

Result in the second step. The video is below the image. The button is clicked, but the image is not hidden

Here there is a feeling that the neural network is such an Ensign Data, who tries very hard and skillfully uses the stackoverflow search, but is in such a hurry to complete the task that he does not delve into it very much.

I try to explain that the image needs to be hidden when the video starts, but GPT does not understand me and hides the image after loading the page using JavaScript. To be honest, here all the explanations began to infuriate me, and it was easier to fix the code on my own and finally achieve the desired result.

My face when trying to explain what I need to GPT fails

It remained to check whether the neural network would help separate HTML, CSS and JavaScript into files. Here the job is done perfectly. GPT does a great job with clearly defined tasks and questions: one action or feature at a time.

GPT split the page code into three separate files and commented on how to link them together

Logbook entry

The GPT service can become hands that type code and add to it, but the mind of the developer is needed to refine and check the code, formulate the task correctly and think through the logic.

It is worth using GPT-chat as an assistant. But don’t expect Jarvis level from him. This tool will not replace a professional, although it can reduce the amount of routine. In any case, such novelties delight my geeky nature, let there be more of these things! The dreamer inside of me hopes that this brings us closer to a world where people are creative and machines solve mundane tasks.