I agree. There's a world of difference between an intelligent user interacting with a computer; a computer learning something in its subject (Bayesian statistics and neural nets all seem very clever (the Spark gender test is great - [
test3.thespark.com] ); and a genuinely intelligent computer.
IMHO, we're years away from a computer which will pass the Turing test. As I tell my students, give me any computer programme and I'll find a way to break it...