If you’ve been paying attention to the headlines lately, you’ve probably seen the usual flood of fearmongering about AI.

“It’s going to take over everything.”
“It’ll become sentient.”
“We’re building our own doom.”

And so it goes.

But none of this is even remotely close to reality. I should know. I’ve built my own AI. Literally. I wrote the code, wired up the tools, and watched it work. So I have a very clear picture of what it can and can’t do. It’s not a god. It’s not a villain. It’s not even a thing with intentions. It’s just a natural language interface for software. That’s it.

AIs don’t actually do anything. It just talks. Whenever you see an AI “doing” something, what’s really happening is that it’s been hooked up to a piece of software, that’s not part of the AI, that performs a specific task. The AI’s role is to generate a string of text that triggers that software. So, when we say the AI is browsing the web that’s not actually what it’s doing. It just responded in a way that that causes a piece of software, that can do nothing but perform a simple web search, to run. It can’t write emails, make phone calls, or purchase stuff on Amazon. Because it can’t do anything but talk.

So, when we hear something like, “In 5 years we will be unable to tell when an AI is scheming against us”, you need to stop and ask, “Does it have the ability to scheme?” And if someone gave it the ability to scheme, that’s all it can do. It can scheme… And? Remember, everything an AI does is software, not sorcery.