Is AI going to end us?

I've had feedback from a few different directions that the posts are too long, so this week I'm cutting the read time. Honestly, I think our attention spans are cooked, mine included, and since this post is about AI, I'm going to blame AI for that too.

So, is AI going to end humanity? People ask me some version of this because I build machine learning models, small ones that look at maps of silicon wafers (the thin discs computer chips are made on) or guess how a molecule will behave. My honest answer is no, not in the way the headlines mean. But I do think it's changing us, and not all of it is good.

Let me start with my week. The links in my last email broke, and I asked an AI assistant to help me fix it. It was genuinely useful. It found the problem faster than I would have, explained it nicely, and then, with exactly the same confidence, suggested a change that would have quietly killed my email inbox. To be fair to it, it noticed a few minutes later and took it back. If I hadn't known roughly how domains work, I'd have done it and then spent a week wondering why nobody was replying to me.

That's the thing nobody warned me about. These tools are extremely agreeable. Ask one whether I should drop my degree and become a full-time Spurs pundit and it won't laugh at me. It'll give me a five-step plan, a name for the YouTube channel and a posting schedule. Then tell it I'm worried it's a terrible idea, and there's a decent chance it agrees with that too, just as warmly. Some days it feels like talking to the most well-read yes-man in the world. It isn't lying to you. It's built to be helpful, and agreeing with you feels helpful.

And still, I think most people are underestimating them. Most of the people I know who have "tried AI" used the free version once, asked it something silly and decided it was a toy. The only people I know using the really good versions every day are IT guys, and half of them are doing it on their company's credit card. The paid versions are a different thing. My honest suggestion is to pay for one month, it's around 20 euros, and actually use it for your work, your emails and all the questions you'd normally type into Google. I think you'll be a bit shocked at how far ahead these things already are. I really don't think people should sleep on this.

What surprises me is how fast it's moving and how little anyone outside tech seems to notice. A few years ago these chatbots were a party trick that wrote bad poems. Now they write working code, read a forty-page document in seconds and tell you what's in it, translate, explain research papers, and help people do in an afternoon what used to take a week. That happened in about the time it takes to do a degree. I see the pace in my own YouTube watch later list. Whenever a new model or tool or library comes out, I save a video about it to learn it properly later, and by the time I get round to it three or four months on, it's already outdated and something newer has replaced it. My watch later list is basically a museum at this point. I don't think most people have caught up with how much has already changed around them, never mind what's coming next.

As for the end of the world headlines, I honestly think a lot of that is marketing. If you run an AI company, "our product is so powerful it might be dangerous" is a great line to have in the news when you're raising money. The real dangers are quieter and already here. Scammers can copy someone's voice from a short clip and call their parents asking for money. Videos of things that never happened look completely real.

If you're on any family WhatsApp group, you've already seen these, even if nobody noticed. A clip of a politician saying something outrageous, a famous doctor recommending some miracle cure, a news video that feels a bit off. A few years ago faking that took a studio. Now it takes a laptop and an evening. My rule now is that if a video makes me angry or scared enough that I want to forward it straight away, that's exactly the moment to stop and check where it came from.

None of that is what actually worries me, though. What worries me is closer to home, and it's about me as much as anyone. A couple of years ago, when I had a problem to solve or a bug to fix, I'd get out a whiteboard or a pen and paper and plan the whole thing out before writing a line of code. I barely do that anymore. Now the first thing I do is ask an AI, and it has made me lazy in a way I don't love. I'm not a neuroscientist, but it genuinely feels like my brain is doing less, like I can't even do normal things the normal way anymore. I see the same in people around me who reach for it before they've tried to think at all.

When I worked things out on paper, I'd get it wrong three or four times, cross things out, and eventually understand why the right answer was right. Now I get the right answer first time and understand it a lot less. We already did this once with Google Maps. Hardly anyone knows the way anywhere now, because we never had to struggle to learn it. The struggle was the part that made it stick. I think people are slowly forgetting how to learn. Not how to find answers, we've never been better at that, but how to sit with a problem long enough for it to actually go into your head.

That matters more than it sounds, because the only way to catch an AI when it's confidently wrong is to actually know the subject. For everyday stuff a wrong answer costs you nothing. In medicine or law or engineering it costs a lot. If we stop learning, we lose the one thing that lets us check it, and I think the space between what it says and what's actually true is exactly where people like me are going to have to earn a living.

So I'm trying to go back to doing it the old way first. Pen and paper, get it wrong a few times, and only then ask the AI to check my work instead of asking it to do the thinking for me. Some days I even manage it.

If you've made it this far without sneaking off to WhatsApp or Instagram halfway through, well done, your attention span is in better shape than mine. I know I promised a shorter post, but this one is about half the length of the last one, so I'm counting that as progress.

AI isn't going to end us. What worries me is that it will quietly do our thinking for us while we're all busy being impressed by it, and one day we'll notice we've forgotten how.