I disagree.
Image recognition algorithms are extremely good now, to the point where they rival humans. They will eventually surpass humans. Also, did you catch the Jeopardy games where a computer (Watson) handily beat past Jeopardy champions?
One of the major remaining hurdles for AI is the ability to be able to intelligently participate in a conversation. Once that is breached, a lot of people whose job is producing hot air can be replaced.
The other thing about AI is the ability to rate, publish and share outcomes in a globally distributed fashion. This would potentially allow it to “learn” much faster than any single human could.
Except that's not really AI; it's [a very specific] pattern recognition.
One of the major remaining hurdles for AI is the ability to be able to intelligently participate in a conversation. Once that is breached, a lot of people whose job is producing hot air can be replaced.
That's a goal [the turing test], but not as much a major hurdle as you think, in the way you think.
One of the things that AI has to do is "goal planning", say taking "build aPL/I compiler targeting MIPSV" and generating such given only the spec for MIPSV and PL/I.
The other thing about AI is the ability to rate, publish and share outcomes in a globally distributed fashion. This would potentially allow it to learn much faster than any single human could.
And?
Lots of social networking sites do this half-way w/ 'like' and 'share', that says nothing about the level of intelligence therein.