There has always been a ton of hype and little substance in the history of AI. Really, the biggest gains have been in things like speech recognition. Literally billions of dollars have been spent on it and it is pretty good but it still has issues especially with background noise.

40+ years ago, people were talking about sentient AI, not many talk about it any more. In fact, AI isn't used as a technical term anymore because of the wasted money and hype from the 50's to the 80's when it became downright toxic.

Instead, they use soft computing as a general term or use terms to describe the algorithms, such as fuzzy logic, neural networks and evolutionary computing. Yes, much of the modern AI mimic human processes like neural networks or biological concepts like evolution or immune systems or a combination of them in their design. Almost all require human training and are tightly focused.

Pattern recognition is probably the biggest advances in that field of computer science, speech and things like facial recognition. Computers can be trained fairly easily to find patterns, they still need human help in their training. Neither of which have a negative impact on many jobs.

People mistake me as a tech over everything person, and despite my education I am a bit of a Luddite. For example, I hate my "smart" phone and really hate things like Twitter and Facebook. They are a pox on humanity, but at least Facebook can help you reconnect with long lost friends, so it has some positive value. Twitter? A cesspool, imo.

Computers assist in art and scientific discovery but the software to not only do it independently, but also have that software know it is good art or an actual scientific discovery is not in our near future. Don't believe the hype.

Have you noticed how bad programs like Word are at a lot of grammar correction(their, there, they're, for example)? Now imagine Word trying to judge the quality of your writing.

There is a longstanding question in computer science, in fact if you can prove or disprove it there is a $1 million reward and you would likely win all sorts of awards and lots of fame.

It is P vs NP. I won't bore you with the details but P problems can be quickly solved by a computer. NP are problems that are solvable but not quickly, and usually not for a generalized version of the problem, but its solution can be verified quickly like P problems. Many of the things holding back AI, and lots of other things, is that we don't know if P=NP. There are also intractable problems, mathematically proven to not have a solution.

I suspect a lot of the advances you fear are in NP or intractable. NP problems on a small set can be solved fairly quickly, but once that set grows the runtime characteristics, which are exponential, factorial or worse. becomes too long to run. Like years, decades, centuries, the life of the sun, etc.

The Internet and then the World Wide Web certainly have had a lot of negative(and positive) effects on society as a whole, that is undeniable. The law of unintended consequences is all over both of those.

If your fears play out, I don't think that people will lay down and die. People are driven by purpose, the current status quo is finding jobs to pay for life, but they also start side business, self-study, raise families, etc. That won't change. There are a lot of people on this site who can't work, including me, that still find a purpose in life, as hard as it can be sometimes. I am fortunate in that my hobby is my college major. But I have other hobbies and I help raise my grandkids.

People are pretty flexible when they need to be and there are economic models being explored to mitigate any massive loss of jobs if that does eventually happen. Humans are great in being able to adapt, even when we don't want to.

Imagine what a robot would have to "know" and be able to do to replace, say hotel maids. The reach, dexterity and strength to clean showers and toilets or make the beds.

What if the guest is still staying there and they left a book on the table or bed that has an empty candy wrapper on it. The AI needed to be able to figure out that the wrapper is garbage but not the book is staggering. What if what was on the book wasn't garbage? Just identifying that the garbage and book are two separate objects wouldn't be all that easy. These are trivial things for people to figure out by themselves. For software, that one scenario is fiendishly complex and there would be so many different things it would have to be able to figure out just to be able to perform that one job.

What if someone smashed the TV, could it report that it is damaged? Okay, I will stop there but hopefully this will help you stress less.

Honestly, the biggest threat to many jobs(not service industry though) is outsourcing not automation.