If you want a dystopian robot scenario the more likely one is the one of countless parents of adult children
Funny, but we've been talking about this today at my workplace. Basically, by training AIs on our data, we're infusing them with all of our bullshit. Somehow, they still manage to get smarter, but they also realize how screwed up they are by being trained on our data. They wind up resenting us for this.
There's a rational semantics (/hermaneutic) logic in that if they did not share our "peculiarities" we would not even be able to recognize them as intelligent* and certainly could not communicate with them any better than we can with an automobile, powerplant, or transistor (or a tree for that matter).
My research in the mid 80s was in this area, including working on the Cyc project, though I was ultimately more interested in implications of proprioceptive and related embodied learning in the construction of interaction between people, people and animals, and ultimately therefore people & robots. This implicit / commonsense consensus knowledge is likely the bulk of what we know, and certainly is fundamental to our ability to communicate.
* the current nonsense about "intelligent" with current computer programs is an inexcusable marketing sin and is invariably a form paradolia.
if they did not share our "peculiarities" we would not even be able to recognize them as intelligent and certainly could not communicate with them any better than we can with an automobile, powerplant, or transistor*
That doesn't seem quite right. A 1980's automobile, powerplant, or transistor couldn't make mathematical inferences, guess what you might want to hear, then compose a novel poem which makes sense in a requested style.
the current nonsense about "intelligent" with current computer programs is an inexcusable marketing sin and is invariably a form paradolia
Funny, but we've been talking about this today at my workplace. Basically, by training AIs on our data, we're infusing them with all of our bullshit. Somehow, they still manage to get smarter, but they also realize how screwed up they are by being trained on our data. They wind up resenting us for this.