That isn't "We are on shaky ground with new, unproven tech." That's "We know it doesn't work. We are comfortable lying about it because money."
All of the industry AI labs follow the same pattern, even (perhaps especially) the en vogue ones (well, I guess depending on wtf "solve intelligence" is supposed to mean...). Watson is just old enough that it's now obvious the advertising agency was given too much leash.
Ditto for self-driving.
Ditto, in fact, for basically any basic or applied research happening in the research arm of any for-profit organization. And at a lot of universities besides! In fact, as far as I can tell, positive PR is some non-trivial aspect of the ROI for industry research labs.
BUT... there's a HUGE GAAP ( ;-) ) between marketing hype and lying in financial disclosures. One is eye-rolly and should be called out on internet message boards and during sales pitches. The other is criminal.
IBM's Watson comes immediately to mind.
The difference is whether or not you can come through with some percentage of your promises. The ones who can eek out some kind of repeatable traction long enough to exit are successes, the ones who never get to scale are frauds.