The most human human, Turing test

Discusses a book titles “the most human human” by Brian Christian. Every year their is a contest by chat bot programmers to write a computer program that can fool judges in a 5 minute conversation, letting them think it is a human. Parallel is a prize for the human who is least judged as a computer. The Loebner Prize.

This entry was posted in Turing Test and tagged , , , . Bookmark the permalink.

3 Responses to The most human human, Turing test

  1. mitsukuchatbot says:

    The Turing Test states that each judge has 5 minutes talking to the machines and the computer must fool 30% of the judges. The Loebner Prize is actually a lot more difficult, as the judges get 25 minutes with each entrant and the computer must fool 2 out of the 4 judges (50%). I think the Turing test will be beaten quite soon but the Loebner Prize will take a lot longer, as the conditions for winning are much harder,

  2. modelmark says:

    I think the openings just come from a database, so a pseudo random start position already gives advantage to the human. The amount of different games that can come on the chess board is still larger than the amount of atoms in the universe, I think. Maybe Go has not been tried sufficiently. Computers will likely pass the Turing test the coming decade, while you would think a conversation is way to complicated to automate as well.

  3. gustavonarez says:

    Interesting. Yes, the best chess engine now (Houdini) can beat the best human player, tho in the japense game of Go this doesn’t happen. I wonder what’s the ‘quality’ that a Go master has that can beat a powerful computer in this game. Obvioulsy in chess the computers win by brute force on calculations, but Go is so complex that it can’t beat the human that way. So if it’s not calculation the ‘edge’ of the human in Go, what is it?

Leave a Reply