AI as a simulation of human cognition is definitely getting close. What I still think is quite far out is a type of intelligence uniquely suited to computers and machines that is based in taking full responsibility for their own lifecycle. In that regard, hardware will likely be the biggest limitation because we are very far off from self replicating machines. Once machines/computers are able to fend for themselves and grow/reproduce, then a truly unique AI will emerge with all of the concerns unique to its existence. That AI is the one I fear because it is utterly unknown.