See this question. You are presuming that an AI will not have good rapport. I suggest they can be built to be better at that than humans.

They did a study where people talked to a psychiatrist via chat. Except the psychiatrist was a (very, very primitive) chat bot. People were okay(ish) with the experience, and the patients didn't know. When they were informed that it was actually just a program, the patients were almost universally far more happy with the experience, and more willing to speak more freely about their issues.

@what I'm at work so I can't watch the video, but I'm 95% sure that the documentary (15 mins) that I linked in reply to John's answer cites the study by name (either during it, or on the credit slide). I'll try to find it later tonight if the video doesn't.

@NexTerren - I wonder if an AI will post to internet forums while being paid to produce code . ;-)

You're presuming that the AI will be (can only be) “programmed” and not feel genuine love. I maintain that such a being can have the same qualities as a human. A mind upload should have exactly the same free will as the natural brain. Once understood, constructed minds can duplicate the effect. Does the fact that the human therapist only tolerates your innane prattle about your problems because he’s being paid (and wouldn’t “like you” in a social setting) devaluate the experience?

Being a psychologist myself, i can confirm that chat bots are playing a bigger and bigger role in psychotherapy... or at least people are trying (check ELIZA or ALICE) to make them play a bigger role.

@what Actually, i think it is kinda useful.While psychotherapists in Germany (where i work) usually have long waiting lists for their patients (patient comes from patience, right?), many forms of therapy are influenced only little by the therapist himself, and more by the chosen method. At some point in the future, a well written bot has almost no limits to it's capacity, and could provide some "psychological first aid" to people who can't afford a therapist or don't get a quick appointment. As a result, therapists might have more time for the patients that need special attention...

@JDługosz, my mom is an elementary school teacher and has been for 20 years, I suggest you have no idea what goes into that job. As a programmer, who happens to program AI's among other things, what she does in the classroom is orders of magnitude harder then anything else suggested on this page. And 10's of orders of magnitude harder then anything Watson from IBM has done.

So? You seem to insist that a created (or uploaded) mind will somehow be a fancy formally-progammed piece of code like today's “apps”. In principle they can be intelligent minds in the identical manner that we are. If you are imagining having to sit down and engineer a piece of software for “elementary school teacher” or any other job, you are still missing the point. A constructed cybernetic system can in principle be just like your mom in being self-actualized and sentient, and can be a teacher.

@JDługosz, if it is self-actualized and sentient, then it isn't an automaton, it's a living being however it got there, and we might as well just of kept someone like my mom on teaching, huh?

@Ryan in my own answer I qualified that “if that’s what you mean by automation…” Some constucted cybernetic system?

@JDługosz, But what makes said system be a teacher? It's sentient and self-actualized, what if it decides it wants to be a firefighter, or an airplane pilot? If it has the self-determination to be able to replace dedicated loving educators who are we to limit it's free-will? And at that point don't we need to pay it so it can maintain itself and won't it want to enjoy entertainment like we do? So won't it want breaks as well?

Pfff, computers have been into psychotherapy for a while now ;)

@Ryan - What your mom does in the classroom is definitely orders of magnitude harder, but is that level of complexity necessary to be an teacher for most kids? Most kids will do just fine no matter who (or what) is teaching them or the methods used to teach them. There will always be "special needs" kids that will require more expertise. But whose to say that an AI of the future couldn't be better at picking out the symptoms and being able to determine the optimal solutions better than humans? Doctors are already helping develop experimental diagnosis systems with really good success.

Humans can't possibly keep up with all the latest research and papers being published. Computers can. That's what the next phase of Watson development is pursuing. They've taught it to read medical papers. Thousands of papers a month are published. No human can keep up to date. Watson can. Early testing has shown it to find exactly the same diagnosis as human doctors a really high percentage of the time (I think about 80%) and I also seem to remember about 1/3rd of the time it comes up with something new that humans didn't see but then later confirmed as a good diagnosis.

Anyways this is getting off topic.