**cough**
As to what you're worried about:
There is zero evidence that this will happen. We do have many examples in which technology advances make certain jobs and industries obsolete while simultaneously creating more skilled or specialized jobs. We used to have armies of ditch diggers, typists, farmers, drafters, telephone operators, human calculators, etc. Hundreds of millions of jobs have been made obsolete. We've also significantly added to the size of the workforce and increased worker productivity. Yet we don't see any long term rise in unemployment. Individual skills (especially the lack thereof) are becoming obsolete, but the human workforce isn't. Automation is a tool that augments human effort, it does not replace it.
Machine learning is the methodology used to to train Neural Networks, as a result the terms are used mostly interchangeably (although you can build a NN without ML and visa versa). Neural networks are not simulated neurons. Machine learning is not akin to human learning. It's meta optimization. These algorithms and techniques are powerful, especially in image and speech recognition. Unfortunately many people talk about them as techno-magic solutions to all of our problems. And before you throw the just add more input data, performance of cnn's is logarithmic at best wrt #samples. They're also "flat" and have significant difficulty representing any kind of abstraction. Furthermore, while they're fairly capable of interpolation, they tend to be catastrophically bad at extrapolation and have difficulty differentiating between the two.
eg.
What ML has done is made it far easier to make convincing demonstrations. These demonstrations are akin to a trick-shot highlight video. You see videos of amazing capabilities as all of the failures are edited out. And in many cases, developing these techniques is as likely to solve the underlying problem as practicing trickshots will prepare you for the NBA. Automation is HARD.