Machine learning has cracked any problem put in it's path. AI hasn't been done yet because no one's actually tried. There's the Loebner Prize but it's too easy (needs to be multi-month and with difficult challenges like write an essay or computer program) and there's no heavyweight turing tests with heavyweight teams (in the ballpark "heavweightness" are search engines and ibm watson).
"AI hasn't been done yet because no one's actually tried."
Researchers have been trying to create some kind of framework for general intelligence since the 1950s with multiple governments (notably the US, EU and Japanese) throwing billions at it, with relatively little results - resulting in things like the great AI Winter.
We have lots of clever focused systems now that are made feasible by the enormous capacity of modern hardware, but I'm not convinced that we have made much fundamental progress on real human level general intelligence.
Note that I do think that general AI is possible - it's what motivated me to get postgrad AI research, I just think that it is a long way off. YMMV.
Yeah, but they haven't put it in the framework of an actual benchmarked test like machine learning problems. They futzed around with prolog and other stuff rather than have any kind of metrics.
You know, I hope you're dead wrong here. Because unless its generality somehow stops at AI research, or we somehow find a way to guarantee it will do what we ultimately want…
The AI does not hate you, nor does it love you, but you are made out of atoms which it can use for something else.
