I think we concentrate more on a one to one conversation - in fact I would suggest that reading counts is "listening to a public pronouncement" - something that is in the environment and not specific to us.

I suspect that twitter counts the same. It's something we don't apply much attention to because it does not look like a human being talking to us.

In fact I suspect that a whatsapp that records one human speaking and then plays it back will have a different attention spike than the text based idea

Edit: of course I wrote the above on autopilot.

This raises an interesting issue: if an interrogator "on autopilot" was fooled, is that really a valid Turing test that was passed?

Fortunately, the solution is simple: to have a Turing test of the interrogator.

You'd have to test interrogators' interrogators ad infinitum then.

thatsthejoke.jpg

Just a random associated though: I wonder if the game of Mafia [1] is somewhat better way to discern intelligences, than the Turing test.

E.g. imagine a game where mafia (AIs) can eliminate actual humans from the game by convincing their fellow humans, that eliminated people are actually the mafia (e.g. AIs).

[1] https://en.wikipedia.org/wiki/Mafia_%28party_game%29

So-so. But it makes emotion a lot more involved. And faking emotions might be hard for humans, it is trivial for AI, and also easy to appeal to our empathy.

The point of the game would be to eliminate AIs, not mafia. No in-game "nights", just a free for all discussion with periodic elimination rounds.

There's ##werewolf@freenode that would provide you the opportunity to see how hard it is to convince others even as humans, there's even a gamemode where the game bot becomes the wolf and people detect that by just some heuristics and gut feeling.