And that may be the most human thing we have ever done.
After roughly 40 years around computers, I’m finding it unusually hard to stop building things and go to bed.
I can have an idea, start working through it with AI, and a few hours later have a working prototype. Sometimes I’m still shaking my head at what just happened. I know how much time and help I would have needed to attempt the same thing on my own.
I’m handing over work that I would once have been proud to do myself, and getting more ambitious about what I want to attempt. There’s an uncomfortable implication in there for a species accustomed to being the smartest thing around.
We got used to being the smartest
Human beings have spent a long time overcoming physical limitations through intelligence. We built shelter, learned to cross oceans, figured out flight. We accumulated knowledge and passed it to people who could take it further. Our position in the world owed a great deal to our ability to work things out together.
It was easy to start treating that advantage as permanent. Everything else might change, but surely we would remain the ones doing the thinking.
We are losing our position as the uncontested apex intelligence. That’s the shift I see underway, and I don’t think we need to wait for a machine to outperform every person at every task before acknowledging it.
Current AI systems are uneven. Anyone building with them has reasons to be impressed and reasons to get frustrated, sometimes within the same conversation. I still have to test what I build and check answers that sound convincing. Knowing where a system falls short is part of learning to use it well.
But I struggle with the insistence that whatever a machine achieves must, by definition, fall outside intelligence. If you decide in advance that no result could change your mind, what exactly are we debating?
“Show me that it works” is a reasonable challenge. We should be willing to accept the answer when it does.
Of course we built it
We have never been particularly willing to let biology set the terms.
Writing let knowledge survive the person who held it. Engines gave us access to power our muscles could never produce. Computers took calculation far beyond what anyone could manage in their head. A great deal of human progress has come from recognizing a limitation and getting impatient enough to do something about it.
Intelligence was always going to attract that same impatience. We have problems we want solved, people we want to help, and far less time than we’d like. An individual can only learn so much, consider so many possibilities, and hold so many interacting variables in mind. Put a team together and you gain capability, but you also have to coordinate it.
Why would we assume those limits were the ones we would finally agree to live with?
Our own intelligence became a constraint, so we started manufacturing more of it.
The comparison with earlier tools has a limit, of course. An engine supplies power; AI can contribute to deciding what to do and how to do it. The more responsibility we give it, the more consequential its mistakes and objectives become. Having built it does not automatically mean we can control everything it does.
Still, the impulse is familiar. We found another boundary and started working on a way through.
Directing the work changes the work
When I’m building with AI, the exciting part is how much more I can attempt. I can explore an idea while I’m still curious about it. I can try an approach, discover why it fails, and try another without every experiment becoming a major commitment.
That changes what I spend my time thinking about. I still need to understand what I’m building, but I’m spending more time deciding what it should accomplish and examining whether it actually does.
Human-directed intelligence is a useful way to describe that relationship. A person can bring far more capability to a problem than they could supply through their own effort alone.
I don’t think that requires us to declare creativity, judgment, or imagination permanently off limits to machines. That would be an odd argument to make after complaining about people moving the goalposts. I expect systems to keep improving at work we currently regard as distinctly human.
Our stake in the outcome remains. People will live with what gets built. Someone has to care whether it helps them, whether it deserves their trust, and whether the objective was worth pursuing in the first place. Those responsibilities won’t disappear because producing an answer gets easier.
I don’t need to remain smarter than the system to have a reason for using it.
The adjustment is personal
It’s easy to discuss this as a grand transition for humanity. It feels different when the capability being automated is the one you spent twenty years developing.
If your expertise pays the mortgage, someone enthusiastically announcing another leap in AI capability can sound like a threat. That reaction makes sense. So does wanting evidence before trusting a system with work where a mistake matters.
I’m excited about what I can build, but my excitement doesn’t settle someone else’s concerns about their livelihood. Greater capability for humanity does not guarantee a good outcome for every person. Access matters. So do the decisions employers make about how to use it.
And there’s very little breathing room. We’re trying to make sense of the implications while the capabilities keep changing.
Evolution did not ship release notes. We now get regular invitations to reconsider what we thought required a human being. Even when you welcome the progress, that is a lot to absorb.
The denial is real. So are the consequences people are worried about. We’ll have a better conversation if we can acknowledge both.
What I want us to build
This is the possibility behind what I’ve been calling The Age of Imagineering. As more intelligence becomes available to do the work, more people can attempt something that would previously have stayed an idea.
I find that ridiculously exciting. Someone who understands a problem through years of dealing with it may finally have the means to build a useful solution. I want to see what people do with that opportunity.
Getting there takes more than access to a capable model. We need ways to test what it produces, enough understanding to challenge it, and people who remain accountable for decisions made with it. We also need to keep asking who gets to participate. A future in which more people can build is worth working toward; it won’t happen just because the technology exists.
Losing our exclusive claim to the highest intelligence can expand what human beings are able to accomplish.
That’s the possibility I keep coming back to when I’m working on a prototype long after I should have called it a night. I’m watching a limitation become something I can work around, and wondering what else I’ve been accepting as fixed.
We’re going to have to get comfortable with being challenged by something we created. I suspect that will bruise our pride more than we expect.
But if the result is that we can solve problems we’ve struggled with for generations, I’m willing to live with the bruised ego.
I want to see what we can do with the help.
AI engineer, builder, and author of The Age of Imagineering.