Chatbots come in for a HUGE amount of criticism, particularly from artificial intelligence researchers, for not being real AI. However, I believe that chatbots, in particular AIML, can handle like 95% of tasks that 95% of people require in terms of practical, usable dialog systems – such as Internet of Things, or Home Automation.
The controversial Turing tests as executed so far, in particular the Loebner Prize, are in fact better than nothing, which is all that anyone else has produced. IMHO, it’s a bit ridiculous to criticize functional agents – that can actually fool people – when you’ve got absolutely nothing to go head to head with them in a fair fight.
In the popular press these days people are going on and on about BAD robots. Robots are not now, and not for the foreseeable future, GOOD enough to be bad. Yes, there will be an immense problem with bad robots, but not with robots good enough to be bad, ethically speaking. I am predicting at least 50 years of simply bad, awful robots before we even get to that point. And that is the downside of chatbots, they’re simply not very clever yet.
See also my Quora answer to: