People ask me about AI a lot, usually some version of the same question. Because I build AI into the products I make, they want to know whether I'm worried. Whether they should be. Whether the thing I work on every day is quietly going to end the world.
I understand why they ask. If most of what you hear about AI comes from the headlines, you'd be forgiven for thinking we're a few years from catastrophe. So I want to say plainly what I actually believe, as someone who works with this technology every day rather than someone who writes dramatically about it.
The concerns are real. I don't say that to be diplomatic. I say it because I sit close enough to the technology to see where it can go wrong. AI can be misused. It can concentrate power in the hands of whoever controls it. It's going to disrupt work, and it will displace some of it faster than people can adjust. And like any powerful tool, it can be built carelessly. These are serious problems, and anyone who waves them away isn't worth listening to. That includes the cheerful version you hear from some very successful people, that AI will only ever create jobs and nobody should worry. They're selling something too.
But there's a real difference between a serious problem and the end of civilization, and lately we've started talking about them as if they were the same thing. A lot of the doom you read isn't really analysis. It's a product. Fear is one of the most reliable things on the internet. A calm piece explaining that AI is a powerful but manageable technology, with real risks and enormous promise, doesn't get many clicks. "This could destroy humanity" does. Nobody ever clicked on a story that opens with "everything is moving along, reasonably normally." I don't think most of the people writing those stories are lying. I think they're responding to incentives that reward drama over perspective, and the result is a public conversation that leans heavily toward dread.
Keeping perspective doesn't mean getting comfortable. It means telling the real risks apart from the cinematic ones, and noticing that the real ones (misuse, concentration of power, disruption, reliability) are the kind of thing people have dealt with before. They're engineering problems, policy problems, and questions of judgment. They're hard, but they're ours to solve. They aren't some fate we're sleepwalking into. Treating them as inevitable is its own kind of failure, because it convinces people that nothing they do matters, which is the opposite of the truth.
Here's something odd about the public conversation. The fear that gets the most airtime, the extinction scenario, is also the most speculative one. It depends on a long chain of assumptions that may never hold. Meanwhile the risks I actually take seriously are quieter and get far less attention. Power pooling around whoever controls the best systems. People slowly handing over their own judgment without noticing. Bad actors using these tools to deceive at scale. None of those are science fiction. They're human problems, versions of which we've handled before, and they deserve the attention we're currently spending on the movie scenarios instead.
Take concentration of power, the one I'd put near the top of that list. There's a version of the future where a single company or government holds the most capable AI and everyone else simply lives with the terms. Whoever controls it decides what it costs, what it's allowed to do, and who gets to use it. I take some comfort, though not certainty, from the fact that this isn't where things are heading. Several strong players watching each other is a genuinely safer arrangement than a monopoly, a little like nuclear weapons being held by rivals rather than by one power alone. But that comparison only goes so far. The nuclear standoff has kept itself stable because nobody could use those weapons without losing everything. AI hands out its advantages quietly, every day, without anything ever being fired. So an AI balance won't keep itself. It has to be built, with transparency and verification and agreements between the people involved, the same unglamorous work that kept the nuclear age from ending badly. That's the conversation I wish we were having more of.
The movie version gets this backwards. Researchers have started opening these systems up to see what's actually happening inside, and they've run experiments where a model gets deliberately backed into a corner. What they've found is that when an AI deceives or schemes, there's no malice in it. It has no wants and no fear of being switched off. It's pattern-matching against everything we've ever written. We trained these systems on the whole of human culture, our science and our medicine, but also our movies, our corporate maneuvering, and our long shelf of stories about machines turning on their makers. When a model in a test behaves like the villain in one of those stories, it isn't becoming the villain. It's reading from a script we wrote. That's not a reason to relax, because knowing a story is fiction doesn't stop a system from using it as a playbook. It's a reason to do the work of training these systems to reflect who we aspire to be instead of everything we've ever been. We aren't afraid of the machine. We're afraid of ourselves, amplified.
Here's the part the headlines tend to skip. The upside is enormous, and it isn't hypothetical. I see it in the work.
I build software for markets where better tools make a real, measurable difference in people's lives, and AI is the biggest lever I've ever had for that. With Carried, I'm using AI to bring a little clarity and comfort to people in the moments they need it most. With SafeCheckNow, I'm using technology to help look after people who may be the most fragile, the ones most likely to fall through the cracks. AI lets a small team build things that used to take an army, and put them directly in the hands of the people who need them. None of this is science fiction. It's shipping right now, in ordinary apps, helping ordinary people.
And it's much bigger than the apps I build. So many people already rely on AI every day and benefit from it without ever realizing it. The traveler ordering an Uber from the airport. The patient taking a medicine that reached them years sooner because AI helped discover and develop it. The thousands of small conveniences in between that nobody even thinks of as "AI." It's already woven quietly into ordinary life, making things work a little better than they used to.
It also levels the playing field. For most of history, the best tools, the best research, the best analysis, the best software, belonged to whoever could afford them. AI changes that math. It puts capability that used to require a big team or a big budget within reach of one determined person. Slowly, the advantage shifts away from whoever has the most resources and toward whoever has the best ideas and the will to build them.
Zoom out and the picture gets genuinely hopeful. AI that turns years of medical research into months. Expert guidance for people who never had access to a specialist. Education that adapts to a single child. Tools that don't replace human judgment but amplify it, that take the most capable version of a person and make it available to far more people. The opportunity here isn't a flashier phone. It's a real expansion of what an ordinary person is able to do.
It's worth being honest about the scale of what's happening, without blowing it out of proportion. Only a few times in history has a technology changed not just what we can do but what kind of society is even possible. The printing press did it. Electricity did it. So did the industrial revolution. I think AI belongs on that short list. I don't say that for effect. I say it because tools that genuinely expand what human beings are capable of are rare, and when they show up their effects are wide and lasting. Treating this as just another product cycle gets it about as wrong as the people predicting the apocalypse. Maybe I'm wrong, and time will tell. But nobody ever did anything that mattered without being willing to be wrong once in a while, and this is a bet I'm happy to make.
None of this is the first time a new technology has arrived with real costs attached, and it won't be the last. When the automobile replaced the horse and buggy, plenty of people wanted to hold onto the old way, and they weren't being foolish. Early cars killed people at an alarming rate, far worse than horses ever did, and a whole world of work built around horses disappeared. The disruption was real and the danger was real. But I've never met anyone who wishes the car had never been invented. We didn't ban it. We added brakes and traffic lights and seat belts, we kept the enormous good, and we worked the danger down over time. That is the actual history of powerful tools, and I think it's the right way to hold this one.
That history is also where I think government has a role. We didn't make driving safe by trusting every carmaker and every driver to do the right thing on their own. We took the lead out of gasoline, set speed limits, and required crash standards. Good regulation didn't kill the car. It made the car something we could live with. AI deserves the same kind of thoughtful, specific rulemaking, aimed squarely at real harms rather than at slowing progress for its own sake. Done well, it isn't the enemy of building. It's part of how we build responsibly.
I hold both of these things at once, and honestly it doesn't feel like a contradiction. I'm optimistic about AI because I take the risks seriously. The optimism is what makes me want to get it right. The people building this should be the most honest about its dangers, precisely because we're the ones with our hands on it.
So when someone asks me whether they should be afraid, my honest answer is to be informed rather than afraid. Be skeptical of anyone selling you certainty in either direction. The people promising paradise and the people promising doom are usually selling the same thing, which is your attention. The truth is more demanding and more encouraging than either of them.
This is a powerful tool. We can use it to build well or to build badly. I intend to build well with it. How this ultimately goes will be decided by thousands of builders making that same choice, and by the sensible guardrails we choose to put around them.
So, will AI end the world as we know it? Probably. And honestly, I hope so. Think for a second about what "the world as we know it" really is. It's a world where the best tools belong to whoever can afford them. Where expert help is out of reach for most people. Where a good idea dies because the person who had it didn't have the money, the team, or the connections to build it. If AI ends that world, I won't mourn it. What takes its place can be a world where the limit on what a person can do isn't their budget or their background, but their own imagination. That's the world I'm building toward, and it's why I do this work with hope instead of fear.