New Scientist: Neural network ‘in-jokes’ could pass secrets
This reminds me of a discussion, years ago on Extropians, of “Schelling points”: for under-constrained problems there may be a cultural preference for particular solutions. For example, if I ask you to meet me in Paris on a given day (but have no time to say more), and you’ve never been there before, you’ll go to the Eiffel Tower — but if we have previously met in Paris, you’ll go to wherever we met before. At least, that’s how I’ll bet.
These are examples of Schelling points, as I (mis)understand the notion.
But.
I know two people who have various things in common.
Suppose I tell each to meet the other in Chicago on a given day.
If they know nothing about each other, they’ll go to some generic Chicago landmark.
If I tell each that the other studied physics at UChi, they’ll go to the site of Fermi’s reactor, I guess.
But if I tell Bruce only that David is a science fiction fan, and tell David only that Bruce teaches physics, they’ll probably go to two different places.
I’m not sure what this thought-experiment tells us, if anything. 😉