I think you vastly overestimate what AI will be capable of in the near future.
It seems that quite a few well-educated people in that particular field also do.
Near future? Probably not in a few years. 20-30 years? There'll be a lot more "untouchable" or "safe" jobs finding themselves automated.
Hell, I can do the work of a room full of draftsmen. Want to make a few revisions to an engineering drawing? No need to scrap half the set and start from scratch.
Or I can skip the draftsman stage for the most part and email a solid model directly to a machine shop in a few seconds, and have finished parts in a week.
I still have a job, but that room of drafters doesn't. I'm sure at one time, human computers were also working "safe" jobs that couldn't ever be automated. Now we can fit a few billion transistors into one or two cubic feet of space and sell it to a consumer.
They're doing that now, it was a 60 Minutes feature. It's not super intelligence, it's taking advantage of the stock market by getting a split second advantage in seeing price swings before everyone else does. If you (a machine!) can do that it can place trades to skim profits off the internet of equity markets. There's more crooked than you can imagine in this world. My advice, keep your nose clean, the law of karma says that gives you the best odds of being happy and healthy by and by.
That's always struck me as a brute-force approach. I'm thinking more of a genuinely intelligent system that actively learns how to recognize new patterns, and to improve itself on the fly, and to make correlations between data and sophisticated predictions that a human would likely initially perceive as a series of errors.
It simply reacts faster than a human can react. Yes, I think it's a bullshit cheat of the system. ("Let's spend a hundred million dollars to move our business
closer to the physical exchange so that we can get a better ping time." It's online gaming tactics, except you can skim billions off of the market with your ping times, and by DOS-attacking competitors with quote requests, all in a fast game of Hot Potato.) My understanding is that stocks are a way of raising funds in the present by selling partial of the company. Then it turned into executive compensation, incentivizing them not to increase the value of the company, but to increase the perceived value of the stock itself. And then high-speed trading decoupled stocks even further from the true value of the company, into a ping war with a high cost of entry.
I don't see where he said "near future". That's an indistinct term anyway. I don't consider 100 years to be an especially long time, and I think AI/robotics will be quite advanced by then. 100 years past, the Victrola was the hot new shit.
That's where my mind's at.
Go back 150 years and explain GPS to anyone.
"Special long-wavelength light shining from clocks 12,000 miles away in space will permit a small handheld device to speak instructions to you so that you can pilot a special carriage along a vast interstate highway system that spans the coast. The carriage is powered by small explosions that occur many hundreds of times per second. The clocks in space are special because they're moving fast, and time slows down when you move really fast, but it's alright because they compensate for that."
That would sound worse than someone today saying that computers will exceed the capabilities of the human brain in under 50 years.
(Yes, I said something about 20-30 years earlier. You don't need human-level intelligence to automate jobs. Pattern recognition can be extremely powerful, even if it's not an intelligent system, though that skill is a very important part of intelligence as we understand it. Patterns are useful ways of effectively compressing data to manage it more easily.)