With 34 years under my belt in this crazy industry, it's becoming harder and harder to ignore certain ... shall we say patterns?

As I look over my incoming feed I see an increasing polarization of opinion around AI:

What it can and can't do
What it should and shouldn't do
And last but not least, is it ethnical and should you use it?

I'm here to tell you where I stand in very simple terms. Whenever a controversial topic comes around in technology, the first thing you want to do is totally ignore the extremes.

Sure, are there plenty of really smart experts telling us that AI is a bad idea all around, represents theft, and should be avoided at all costs and under all circumstances? Yes there are.

But they're just plain wrong.

The set of algorithms and ideas that the vague marketing term "AI" represents including some INCREDIBLY powerful tools that only a fool would totally ignore.

I myself use Github Copilot constantly and it basically acts as a glorified spelling checker and code nanny. But here's the thing: It saves me HOURS a week, and I can prove it very simply.

I can just hear someone out there saying "I don't need such things! I write my code by hand as god intended" but you might consider checking your privilege and hearing what I have to say.

I am partially blind and fine/gross motor impaired. My words per minute is a very respectable 85, but my error rate is INCREDIBLY high. This leads to numerous typos in code that, again with my bad vision, can be VERY hard to catch. It's very helpful to have something sitting there telling me "Hey you didn't actually mean to type WonkasAmazingGibStopper, you meant WonkasAmazingGobStopper instead. As you well know, programming is a VERY harsh god. Typos like this can lead to hour after hour lost reading over source code to find that single mistaken letter.

So that's my argument FOR, now let's talk about the dangers. There are very real dangers, and you must be aware of them if you even think about using these tools.

In short. AI lies through its teeth, and it does that because of the way large language models are built. They consider your input one character at a time and decide what the most likely next choice will be.

Comments