The Rise of Artificial Intelligence

For many years, I've had a general inclination to follow tech news. I've always been a nerd, so this sort of thing usually excites me.

Lately, though, I've been hearing a lot about a particular topic: artificial intelligence. ChatGPT. Copilot. It seems to be everywhere these days.

"It's amazing. You MUST try it out."
"It's coming for your job!"
"AI? That's just another fad. They're trying to trick you into giving them more money."

Well, it seems to be trendy, so I might as well cash in on those clicks. Here's how I've reacted to AI over the last couple of years.

Resistance

For a long time, I thought AI was a load of old rubbish. That's why I end my blog entries with the following closing line:

No AI tools were involved in writing this blog entry.

This line is a statement of pride in my content. It is here to let you know that whatever bot-generated content you might find on other websites, the stuff here was written by a human.

So, why did I want absolutely nothing to do with modern AI?

It can't think

I thought everyone was giving AI too much credit. I thought that perhaps everyone was forgetting a rather import fact: computers can't think.

Yeah, computers are pretty neat. Given a sequence of instructions, a computer can automate many tasks for you, and it can complete these tasks many times faster than you. But it can't think.

This limitation also applies to ChatGPT. Ask it a question, and it'll attempt to generate an answer that makes sense to you. But make no mistake: it's a computer program, and it can't think either.

This means it might answer your question incorrectly. It probably will, a lot of the time. And it won't know the answer is incorrect - you'll have to go and find out for yourself.

So, why would I want to waste my time, messing about with a chat bot that pretends to know what it's talking about, when I still have to research everything for myself anyway?

It runs in "the cloud™"

I am not keen on proprietary applications that I can't run on my own computer. ChatGPT is one of those applications.

I can't sudo apt install chatgpt. In order to use it, I have to go to the OpenAI website, create an account, and rely on a connection to some server on the internet.

This means something like the following could happen:

All of the scenarios above have been known to happen with many proprietary online services, and any of the above would have a serious impact on my use of them.

At the end of the day, I don't want to become reliant on an online service today, knowing that I might not be able to use it tomorrow.

It's coming for my job

Daryl from South Park, shouting: "THEY TOOK OUR JOBS!"

Perhaps I knew all along that AI wasn't a load of old rubbish. Perhaps I saw the merits of AI, and was simply scared of what it might mean for me.

I have a flat. I have bills and living expenses. I need money to pay for it all, so I need a job. And I have one. So, it's all good, right?

But then, some chat bot comes along, and says that it can be used to produce code. Bro, what?!

Now, I didn't really believe there was a threat to my job. What I do is still too sophisticated for large language models, in their current state, to replace me. But I can't say I wasn't a little worried.

So, for about a year, I just chose to ignore this weird AI trend, hoping it would blow over. So far, though, it hasn't.

Giving it a go

At some point, I had to write a date range filter in a query language that I was still learning.

Whenever I have to read the documentation for this language, it tends to feel like a lot of work. More so than with the documentation for other languages. I just don't get along with it very well.

This time, though, I decided to ask ChatGPT how to solve my current problem. In response, it gave me a query string that looked about right. I pasted it into my code editor, and it worked.

So, AI isn't a load of old rubbish. Actually, it's pretty neat! I guess I should start using it more.

Cautious adoption

I decided to get a licence for GitHub Copilot. Now, when I'm coding, Copilot just sits there, in the background, offering up code suggestions. I can take them or leave them.

If I'm writing something relatively simple, I'll often pause for a moment, allowing Copilot to make a suggestion. Often, its suggestion will be correct, and I'll hit the tab key.

If I'm writing something a little more complicated, I'll probably be writing a lot more of the code myself. AI may be useful, but it has its limitations, and I need to account for those appropriately.

And in any case, I'm not going to commit AI-generated code without reviewing it first. I need to make sure that I understand it, and I need to make sure it's correct.

Use on Durrant.tech

Yes. I must confess, my website has been infected by the virus of artificial intelligence. In fact, just by reading this entry on my website, you've experienced AI-generated code.

So, I decided to ditch the monospace typeface for headings and homepage links on my website. Instead, I wanted to use Atkinson Hyperlegible as this website's typeface.

The problem is, this typeface might not be present on your computer. This means I have to load it in as a web font. The process is relatively straightforward, but also a little tedious.

Instead of doing it all myself, I decided to get ChatGPT to do it. And it did a fine job.

Perhaps I'll use AI here and there for this website's code base. I have a lot of ideas in my head, and if I can get AI to do some of the boring bits, then sure, I'll take it.

When it comes to my blog entries, though, AI will have no involvement. This is a personal project, and AI-generated writing has no place here.

No AI tools were involved in writing this blog entry. The CSS is dirty, but the Markdown file is clean.

Published 2024-05-28.