10 years ago you had to call computational statistics "machine learning" to trick VCs without any technical knowledge to fund you. now you have to call it "AI". both are complete misnomers.
can computational statistics be used for evil? yes (mass surveillance, deep fakes, etc.)
can it be used for good? yes (improved weather forecasting, earlier disease diagnostics, etc.)
on the whole do economic incentives all-but-guarantee it will be used for evil more than good? yes.
does any of this have anything to do with cognition/intelligence/sentience/etc? not even a little