On December 1st, 2022, the day after the public release of ChatGPT, Hacker News had a post that introduced me to the service. I had been following the development of artificial intelligence for some time, being somewhat of a casual futurist. But up to that point my use of AI had been mostly hidden behind Google searches and ad targeting. That changed when I was faced with a provocative text input and the promise of free natural language output.
Two transformative months later and my job will never be the same, the way I learn has changed completely, and my plans to teach web development are back to the drawing board.
From Science Fiction to Stable Diffusion
My father, Mitchell Harding (aka John Henry Thong) co-hosted a notable radio show in the ’70s called ‘Hour 25‘ on Pacifica Radio’s KPFK in Los Angeles. I grew up with Science Fiction. Harlan Ellison was a regular presence. Ray Bradbury was a familiar face. I’ve read many of the classics where the promise and perils of AI are common themes.
Growing up during the ’80s it was hard to miss movies about artificial intelligence like Blade Runner, The Terminator, War Games, Short Circuit, and Tron. The ‘Information Age‘ was upon us and the collective imagination about the implications of computing ran wild. Commercial excitement mixed with cautionary fear.
For many decades now (even centuries) we’ve been anticipating some kind of real-world impact of artificial intelligence on our lives. And yet, for the past 30 years we haven’t seen much progress as advancements in AI have largely impacted us in indirect ways.
As I mentioned above, the most notable impact of AI on our lives is likely natural language search in Google or maybe a simple customer service chatbot. And some of us know that AI is used to turn the data gathered about us into targeted ads. But these uses don’t really feel like the future we’ve been anticipating. Where is my Hal9000 or Jarvis?
More recently, AI has made a big splash with image generation. Websites like NightCafe allow users to generate novel images based on a text prompt or seed image and have garnered a great deal of attention for being both fun to use and also ethically questionable. But while thought provoking and entertaining, AI image generation still didn’t quite feel like a path to our long-held AI fantasies.
AI’s ‘iPhone Moment’
On November 30th of 2022 ChatGPT came out of private beta and debuted with a free-to-use public interface. Within 5 days the service had 1 million users. And just two months later it has over 100 million users, or more than Instagram. No online service has ever been adopted this quickly.
Many are talking about why the service has been so popular. To understand that you need to have some idea about how it works and what it is able to do. (Trying it yourself is a great way to get a sense of this.)
Very briefly and roughly, ChatGPT is machine learning software that has been trained on a huge amount of data scraped from the internet allowing it to provide and synthesize that information through a natural language interface. In other words, you ask it questions and it responds in a way that doesn’t sound computerish.
The service is very good at some tasks like coding and not good at others like giving personal advice. And so it’s important to understand that it is not actually thinking, no matter how smart, or even neurotic, it sounds. It is simply a ‘large language model’ that can spit out information and knows language well enough to make it sound human.
However, in spite of its limitations it has captured the collective imagination. It is fun to play with, it seems to have endless uses, and it finally gives the public a sense that our dreams of AI have arrived.
The Day My Job Changed
One thing ChatGPT is particularly good at is coding. The software was specifically trained on coding and in many ways transforms that task. The day after the debut of ChatGPT I had the pleasure of discovering this on my own.
When you first sign up and log on your are met with a text input and three sample prompts. A prompt is the question you ask to get a response. One of the samples is “How do I make an HTTP request in Javascript.” So I knew it could answer questions about code.
My first prompt was asking ChatGPT to generate the html and css for a snowflake. It not only printed out the code, but it explained the code and suggested that making an intricate snowflake could be a complex task. The resulting snowflake was not impressive. But the code and language output certainly was.
Next I asked it to create a typing program that put a different letter on screen every second which you would have to click to keep playing. I got a bunch of code, but it didn’t work. I then asked it to debug the code and got a lot closer to the result I wanted. ChatGPT wasn’t really able to write a large chunk of Javascript, but I immediately recognized that it would be a fantastic reference for any questions I might have about coding.
Most impressive was its ability to describe code, debug coding errors, and to suggest ways to enhance code. Within my first week, it had many times helped me to find silly errors and also to find more efficient and performant approaches.
Over that weekend it became clear to me exactly how transformative the technology would be for my profession. There really wasn’t a better reference for coding knowledge. Google searches result in a page full of ads and no specific result. You end up clicking on many links and many of those don’t help.
In contrast, ChatGPT was most often delivering the correct answer on the first try and offering additional notes for proper usage. In essence it felt like it understood my goal and was trying to be helpful. The time savings in that result can’t be understated. And the clear implication to me was that I could both do my job faster and also learn new things a lot more quickly.
Enhanced or Replaced?
To many developers the introduction of such a powerful tool suggests the possibility that AI will soon replace them. That first weekend I asked ChatGPT if that would be the case. It said no, and I have come to agree with it’s reasoning. It said that AI would be a tool that could make coding more efficient but that humans would still be needed to direct the process. In essence it is saying that coders will ultimately become project managers.
Seeing the writing on the wall, I knew I needed to be using this technology. And I knew that before long those not using it would not be competitive on the job market. So on Monday morning I reported to my supervisor that I was now using ChatGPT as a reference and assistant and that it should make me more efficient. No going back.
Many industries and jobs are being currently upended since ChatGPT debuted. Content writers become editors. Roles in finance are insecure. Many middle management positions are just going away. At the same time many new companies and job positions are being created. We now have ‘prompt engineers‘ whose sole job is to productively interact with AI.
Within web development it’s likely that companies will before long require fewer developers. And it also seems likely that the role of developer will transition from writer of code to compiler of code. With further development of the technology it may be that writing code by hand will become obsolete. But that time has not yet come.
Transforming Web Development Education
In many ways we could call this the ‘Pong stage‘ of natural language AI assistants. For all the hype and doom one can find on this topic, we really have no clear picture of the impact this will have on our future. However, one things is entirely clear to me and that is the impact even this first stage has on education.
In my view one of the biggest barriers to tech education is the fact that most educational tools are either too easy or too difficult for most beginners. They are either aimed at 4th graders or adults with some initial technical knowledge. For all the free and commercial options, very few seem to remember what it felt like to start at the beginning, without assuming the user is a young child.
In contrast, ChatGPT will offer you information at any level you request.
- Ask it for a definition. Then ask it to give you the same definition written in more basic terms.
- Ask it for code with a description of each line.
- Input some code you found and ask it to summarize the code.
- Ask it for different code that accomplishes the same goal.
- Ask it to list all the steps involved in a larger project and then ask it for code at each step.
In addition to it’s benefits for beginners I have found that I am learning new code quickly as well. The first week with ChatGPT I built a text-to-rhythm audio app, a simple video game, and a feature-rich fractal viewer. While ChatGPT could not produce the code on it’s own, I felt more confident approaching complex tasks using Javascript because I knew I had an assistant that could help me overcome obstacles.
One significant result of this new technology is that my plans to teach web development have dramatically shifted.
After-school Opportunities
For some time I’ve been interested in teaching a web development class after school. After-school programs offer a unique opportunity to reach kids as the time is voluntary for them and no teaching credential is required.
With the outline of the course and a rough lesson plan I’ve been aiming at teaching the class soon at a local youth center near my home. The premise was to share with the students what I actually use on the job in order to demystify the work and make it seem accessible.
But with the launch of ChatGPT my plans have changed quite a bit and I’m even more excited than ever to make it happen. I can now show students how AI has made my job easier while also presenting them with free services like ChatGPT or You.com Search that facilitate learning.
The pitch is compelling. I’m telling kids that not only is my job easier than ever, but that learning to do it is also easier than ever. And even as my job is changing and perhaps my specific role will become obsolete, familiarity with AI and it’s ability to make learning easier is highly beneficial.
Embracing The Future
So I’ve decided not to fear the future of my profession and embrace what may come. And to make this attitude practical I know I need to learn more about AI, use the technology, and help others to use it as well. Perhaps my future is in education. Maybe my work becomes focused on project management. But I do not believe that the transformative impact of artificial intelligence is a shrinking of opportunity, but instead an exciting transition.