Places features

What do you think of the above conversations? Can you work out whether they're with humans or robots?

steve
Test 1: Too obvious. The computer, #2, responds in the typical Eliza-like evasiveness. It also completely misinterpreted the question "What do you meditate on?" The human holds a much more believable exchange with the tester.Test 2: In #4, the computer makes a series of fairly non-sequitur responses based on either the structure of the text it received (e.g. question) or a key word (four = we're talking about gospels? That's a stretch). On the other hand, #3 is either a human with a fair wit or a computer that's been coincidentally loaded with the rules "hand NEAR holding" and "jewelery box". Notice the part about shaking hands... you can't shake hands, because you're communicating via IM. The human cleverly stated that.Test 3: I think this is the most difficult to rule. However, #5 shows better comprehension and the conversation adheres better (particularly note that a number of times Charlotte responded in two lines, e.g. "that sounds really cool"... "a man at our church is a production executive at the BBC"), while #6 tends to catch key words without much flow (and the only two-line response is a typo correction, which would be easy enough for a computer to cleverly fake). I'd say of all the bots, this one performed the best.Test 4: The entire conversation consists of the computer (which is Elbot, by the way) catching a key word or phrase and then making a joke about how it's a robot. Human, #8. Computer, #7.Test 5: Taco completely gives it away. Human, #10. Computer, #9.

Elbot
BTW, can you imagine me on a farm!? I think robots would make great scarecrows. The beeping noises would definitely scare off the crows.

chris
some of all the chat windows were like humans but had a certain robotiness so the robots are getting to be as smart as humans but there not there yet.