Who "wins" and who "loses" depends on when you measure it.

Wayne Gretzky's talent doesn't come from studying everything he's experienced in hockey and making long-term game plans. It comes from constantly taking in all the data that's happening in the moment on the ice, and instantly generating constant predictions based on super-efficient mental models he's built in his head. Technology has to work more like Gretzky.

Some of what we're learning suggests a balance between exposing children to new things, yet giving them a chance to repeatedly experience something they enjoy, which builds "chunks" of information in their brains.

High fidelity is a rich experience, and you'll put up with terrible convenience to get it - maybe it's high cost, waiting in line, jumping through hoops. High convenience is the opposite - it's a commodity, but it's cheap and easy and ubiquitous. A great exclusive boutique shop is high fidelity; Wal-Mart is high convenience. Both are hard to establish in their own way. The thing to remember about sustaining either is that you can't sit still. Some other entity will always find a way to challenge your fidelity position or your convenience position.

Now companies tend to mine gigantic databases for insights into what might happen six months from now. That might always be valuable, but there's a different kind of value - and a competitive edge - in processing ongoing streams of data through a software model that can quickly and constantly make predictions about, say, whether a certain customer is going to defect, or an aircraft is going to run into trouble.

Our brains seem to have the power to do one or the other - record and remember every detail, or chunk it to higher level concepts and forget the details. We can't seem to do both. The fact that you could not fly over a city and remember every detail is not something to worry about.

Artificial intelligence uses a complex set of rules - algorithms - to get to a conclusion. A computer has to calculate its way through all those rules, and that takes a lot of processing. So AI works best when a small computer is using it on a small problem - your car's anti-lock brakes are based on AI. Or you need to use a giant computer on a big problem - like IBM using a room-size machine to compete against humans on Jeopardy in 2011.