AI Doth Make Managers of Us All

AI is a software program. 
A software program is created by a group of people.
A program is a recipe to solve a problem, in this case, a recipe created by a bunch of people. 
The only advantage AI has over people is this: it has a perfect memory. Unfortunately, it has no accurate idea of how to statistically weigh the various facts it remembers. Which facts matter, which facts do not? That's why different AI programs provide different answers to the same question - the people who created it weight the same facts in different ways.

So, AI is really just a group of people who have very good memories, but aren't necessarily great at connecting the facts into a coherent, accurate narrative.

As I was explaining to someone yesterday, a person who doesn't know much about a subject, but interacts with an AI to discover more, is a lot like a five-year old interacting with a parent. The five-year old doesn't know as much as the parent, but that doesn't mean the parent is always correct. As long as the person isn't informed on the subject, that person won't be able to tell when the AI is incorrectly weighting the facts.

But, once you learn enough about a subject, you can now interact with the AI as a teen or young adult interacts with a parent. Run questions past the AI, consider the opinion, then determine on your own whether it has correctly pieced things together.

Elon Musk doesn't design rocket engines or electric cars himself, he hires teams of people to do that for him. Elon solves the problems that his engineering teams are stuck on. He suggests new approaches, alternate avenues of exploration. Then he turns loose his technicians to have another go at it. 

AI turns everyone into Elon Musk - you now have a team of highly skilled, well-trained people working for you, but sometimes that team, the AI, goes off the rails. They have problems they cannot solve. You have to manage them back into solving the problems they know how to solve. And keep in mind, there exists an entire class of problems that algorithmic computer simply cannot solve. 

This is the famous "Halting Problem". Kurt Godel demonstrated that pure logic demonstrably cannot resolve every class of problem. Alan Turing showed one example of this: the problem of the infinite loop. Can a computer be programmed to detect that it is in an infinite loop, so that the computer can break out of that infinite loop on its own? It turns out that this cannot be done, not even theoretically. Human beings can see that the infinite loop exists and we can force the computer to break out of that loop, but the computer can't see that it has trapped itself. It cannot break out on its own. Since AI is nothing but a computer program built on pure logic, there are questions we can answer, but it simply cannot answer. (for popular renditions of how the Halting Problem works, see Star Trek's "The Changeling" (S1E7, 1967), "I, Mudd" (S2E8, 1967).

As long as AI is built on pure logic, AI will always be less capable of resolving problems than people are. AI is a superb technician, often better and faster than any human being could ever be, but there are things we can do that AI can never do. For instance, we can see and break out of infinite loops. AI can't.

So this is the heart of the problem: prior to AI, the great bulk of people were technicians - they knew how to do their own job very well. But insofar as any job can be reduced to a recipe, a series of discrete, well-defined steps, then AI will be able to do that job better and faster than people can. However, if that recipe carries the possibility of degrading into the equivalent of an infinite loop, then AI will trap itself  inside the routine and will never be able to extricate itself. Never. People will be able to see the AI's problem, but the AI can't.

After flying an experimental plane, the X-1A, Chuck Yeager remarked, "The X-1A will fly itself into a corner you can’t fly out of." The same is true of AI. It can think its way into a corner it can't think out of. People will have to help the AI out of those corners.

Thus, AI turns everyone into managers. Managers don't know the deep details of how ANY job works, but they have a vision of how the various pieces and processes need to come together in order to attain a larger goal. AI is now the team of technicians, the human person interacting with it is the manager. AI turns everyone into Elon Musk.

Most colleges have built their reputations around training technicians, but that practice is now moribund. Wherever AI is active, we cannot afford to teach people to be technicians, because AI will always be a better technician. People must be trained to be managers. We know "conscience doth make cowards of us all", now AI doth make managers of us all. 

Not everyone is cut out to be a manager. The technicians are out of a job.