[andrewjl]

I think that, given the change we expect to see in human-computer interaction over the next few decades, you can't say never. The majority of the obstacles you outlined are not fundamental but rather implementation problems that can be solved by motivated parties with time and effort. Of course, saying this doesn't mean someone will solve them, just that someone can.

As an example, the censorship and immoral content dilemma is easily solved by letting the user "censor" their own feeds of content they themselves find objectionable. This is already trivially done via block lists and word filters, in the future thanks to ML there can be a much more sophisticated level of control exposed including fuzzy sentiment metrics. I wouldn't be surprised if tools like this are already being built. The key idea here being that the control is with the user rather than with some 3rd-party whom may or may not be taking the users own preferences into account.