Anyone interested in the development of AI should read Clayton Christiansen’s The Innovator’s Dilemma (1997), where he discusses the phenomenon of disruptive innovation— a term many people think they understand, but don’t.
To simplify a great deal, what it means is that successful new technologies often begin as less effective than current technologies— but if they’re cheaper and/or more easily available, they will tend to crowd out the old tech anyway, sometimes catching up in quality, sometimes not. He cites many examples, particularly in the area of disk drives, etc.
A very basic example: the first printed books weren’t objectively better than old books copied by scribes— manuscripts handwritten by monks were often gorgeous, much “better” than the ugly, smudged things that first came off Gutenberg’s press. The small proportion of the population who could already read scoffed at this new innovation, which they didn’t need. But the printing press brought written words to the masses, giving them new content and reasons to learn how to read. Eventually printed books took over, and printing improved. These new books were “good enough” for most people, and they widened the market for books and information. (Later on, in a smaller example, paperback books were objectively of lower quality than hardcovers… but they made books more accessible to more people at lower price points, and they now vastly outsell HCs, to the chagrin of some book lovers. Today, websites and YouTube are taking over the functions of many books, to the chagrin of many. I’ve watched hundreds of hours of YT on reefing, but purchased no books.)
This concept of “good enough” is the key to understanding disruptive innovation. I think we’re fast approaching the point when AI will be “good enough’” for many more applications than we see now. Yes, now it’s often laughable. AI can, strangely, hallucinate, and seem to lie. We’re still better at almost everything than AI is. But AI will be deployed on more and more tasks, first where the stakes are low. It will improve faster than anything has ever improved. Eventually we may refuse to allow a human doctor to operate on us, since AI will be vastly more experienced. (I’m sure there will always be a human around, like the vestigial driver in some self-driving trucks.)
So yes, I do envision a day, perhaps not very far away, when a particular AI (not necessarily any AI) will be the definitive source for reefing questions, just as another will specialize in brain surgery. I think AI will be fragmented like this, into niches, hopefully avoiding the centralized Overmind/doomsday/Skynet scenario imagined by so many SF films (Terminator, Matrix, Dune….) For now we can laugh at it, and rightly so. Soon we won’t be laughing.
The genie can’t be put back into the bottle, but I do worry about a full-on Luddite rebellion as jobs start to require technical skills that most people won’t be interested in learning.
And now back to your regularly-scheduled Christmas programming.
