Computers, Internet, AI
Machine learning’s mind-blowing difference is that it programs itself. Provide it with scads of data and it will find relationships among the dots, yielding models through which one can run inputs—often quite complex—that yield sometimes surprising outputs. There is, of course, a lot of human intervention in this process, from deciding which data to feed in, to optimizing the algorithms to get useful results, to checking the outcomes to see if the system is perpetuating or even amplifying the biases inevitably expressed by the data. Still, a computer that in one sense or another can program itself is a remarkable achievement that will be noted in any future history of our time on the planet.
As we get used to the presence of machine learning and how it works, we will, I think, learn a different set of lessons. Machine learning is interesting where the models it makes for itself are more accurate than the ones we humans would program into the computer. We are thus brought to see that the world is far more complex than our way of thinking about it—with our limited memory and CPU—can recognize.
The world as revealed by machine learning is more orderly than the random creative chaos of the Internet and less knowable than the simple logic of our old computers. The inputs are far more unpredictable than our old computers demanded, but still less surprising than the human selves we encounter on the Internet.
These have been the three paradigm shifts from digital technology: one that shrank the world to the machines’ borders, one that enabled chaotic human creativity to flourish and a third that weaves ineffably complex webs out of the bits we have fed it. I cannot name any invention that has so fundamentally affected our ideas so quickly, other than language itself.