You should check out my answer here to your second question.
For your first question, which is a special case, the answer is it might be one of the best fields to study!
Academic and industrial studies agree that working in a job that requires a lot of social interaction, particularly a job that involves caring for others, reduces automation risk. Among the safest sectors are education and personal care assistants, for example.
There are lots of possible reasons for this:
- Even if a chat bot can replace your therapist, it probably cannot replace the sensation that another person cares about your problems, precisely because it is not an actual person. For people in dire emotional straits, having someone care might be worth a great deal.
- Having a human tutor or therapist might be seen as a status symbol, even if an AI system is equally competent. If Harvard offered a degree where you studied on campus, taught by humans, and a degree where you studied online, taught by chat bots, they would probably not be perceived as equally prestigious by employers and peers.
- AI systems like chat bots tend to fail in unpredictable ways when they enter unfamiliar situations. When dealing with an emotionally unstable person in therapy, or with a small child in education, unfamiliar situations might happy often, and the costs for failure might be high. Even if AI can handle most of the job, we might not be willing to trust it to an AI system because of these risks.
To address your third point, Q & A seems closer to AGI than to existing AI. If a program can give compelling answers to free-form questions, then it can almost certainly pass the Turing Test reliably. While you might see simple bots that analyze text and post a related wikipedia link, I suspect you won't see a high quality Q&A bot for many years (or decades) because of this.