AI you don’t go looking for
Upload a photo, hover over a friend’s face, click on their nose, type their name. That was how you tagged people on Facebook, and it felt revolutionary — we could annotate our friends in our own photos.
I still think that interaction had a lot to do with Facebook’s early growth.
Under our noses, Facebook switched it all up. Using the data people had manually labeled, it built systems that could make suggestions automatically. Nobody had to seek out machine learning; people encountered it exactly when it was useful.
That pattern is everywhere. Netflix recommends something to watch. Amazon recommends something to buy. A camera focuses on a face. A phone sorts a photo library. The machine learning is there without anyone learning a new ritual.
AI isn’t fully there yet. Chatbots are a good interface because people already know how to talk, but a chat window isn’t the final destination for intelligence in software.
The world is just starting to wake up to artificial intelligence, and we’re nowhere close to the end. So what’s after the chat window?
When it becomes ordinary
The companies that shape the future won’t only build destinations where people “go use AI.” They’ll embed useful intelligence into ordinary work until the assistance feels like part of the environment.
That doesn’t mean hiding what a system is doing or removing human control. Invisible shouldn’t mean unaccountable. It means the technology appears at the moment of need, through an interface that already makes sense, with boundaries a person can understand.
The second half is the one I’d hold onto: boundaries a person can understand are what separate a tool from a system acting behind your back.
The part I still don’t have an answer to is where all of this lands.
What I build will be software where the model never appears in the name of the feature, where a person can see what the system did and overrule it, and where nobody has to learn a new ritual.