More complex a system gets, more issues it has. Never less.
Consider seriously, how many AIs could match a generalist's ability and knowledge and would have a job after a few months such as if they got the job of research assistant. I can say I would have fired those AIs I've encountered thus far that work in the public arena - either as use for populating wikis or a blog or offering search answers - too many pitiful errors ... a couple of honest ones is forgivable, a handful is a problem that needs sorting out, correct only 95% of the time, bye.
What for example some laypersons seem to be learning lately by way of web, is a good indication of poor AI adoption. AI is wonderful for patterns, statistics and endless fruitful possibilities. But with information a single mistake might be overlooked, it happens -- what doesn't pass is that type of bloke we've all met somewhere ... sometime in our lives, who seems to know everything, a wealth of information, right up until they stumble onto something we ourselves are specialists in, and we come to slowly realise they are no more than a skilful BS artist - even not half of what they are talking about turns out to be BS ... 10% or 5% BS is still infuriating as it takes more effort to correct everyone else who's been hanging on their words.
That'll work out well in the long term.