AI peaked and declined when most people here were small children.

It doesn't seem like you know very much about modern AI work. Watson was on Jeopardy two years ago.

See: http://en.wikipedia.org/wiki/Strong_AI

Watson is not AI. It can't reason. It's a very big, very fast search engine with access to a ton of documents. It's not based on any theoretical breakthroughs, just the natural advancements in computing power. It's very useful, but something that can't reason isn't going to create the sea-change people think will come out of AI, and Watson doesn't get us any closer to that goal than we were decades ago.

I'll give an example. Google's search engine has been a hugely successful application of certain AI concepts. But it is not AI. It can use the link structure of the web to help you find documents, but it could not, for example, recreate the link structure between a set of documents.

Well, you can call it whatever you like. But, I'd certainly prefer google self-driving car to a taxicab with a human driver. And I don't see how is that much different from a robot performing a routine operation by the way. So here goes these surgeon jobs, alongside with the cab drivers.

...no, you wouldn't. It is untested in plenty of environments right now, and being able to drive around Palo Alto guided without killing anyone is nice but not nearly enough.

One day, of course, it'll be amazing and safe. But AI has a history of its promising advances being perpetually pushed to the future.