The Akinator is sometimes almost spooky.
Check out this game report. This is a list of its questions and my answers, in order. Doesn’t seem to me to have narrowed it down nearly enough when it suddenly asks me about modes of transportation.
I’m just sayin’, if you and I played “20 Questions,” I doubt I could have gotten it in 12.
Doesn’t seem right that it would expect a yes answer to “is your character linked to a form of transportation”. Actually, every answer you gave was the answer it expected- that suggests Amelie Earhart is the first expected character to every game!
That just means that those are the answers it expected for Amelia Earhart. If you had said that your character was male, but then eventually it guessed Earhart, it would have labeled that answer as unexpected. It doesn’t have any expectation at all about the answers before it figures it out-- If it did, it wouldn’t ask those questions.
I’m not sure - EVENTUALLY but before it reaches a final conclusion, it clearly has narrowed the pool down quite a lot and starts to ask a couple of questions (Such as the transportation one) that are clearly designed to work that pool down to just 1.
Ah ok, that makes sense that the expectations it lists on the log are in hindsight.but that aside, it should have expectations for every question,unless the answer pool it’s using is exactly 50/50- in the beginning those expectations are a hair over 50%, so it’s just not a confident enough expectation to make an assumption.
It always tries to ask 50-50 questions. This obviously isn’t always possible, but it’ll try to pick the question that’s closest to 50-50. Even when it’s narrowed down the scope quite a bit, it’ll still try to pick questions that evenly divide that narrow scope. Thus, for instance, if it’s narrowed it down to the four main characters of The Big Bang Theory, it’s more likely to ask “Do you live with a roommate?” than “Were you born in the US?”, or “Do you have a PhD?”, since the first question would divide them 2 to 2, while the others would divide them 3 to 1.
It took two sets of guesses to get Shamgar, which amused me since I originally entered Shamgar into its database years ago.
It completely failed to guess (Illinois Attorney General) Lisa Madigan.
I just stumped it twice. But it won one.
The subject was mystery novels. Akinator is not really up on mystery novels and should go attend Left Coast Crime or Bouchercon or something.
I just did Major Major Major Major from Catch-22. It’s first guess was Yossarian from that book! Second guess was right. I was pretty impressed.
It got Chingachgook from Last of the Mohicans in one. It’s surprisingly good in a lot of areas.
It got Paksnarrion (character from self-same fantasy novels) in one. That impressed me a bit since the questions seemed to be more general.
I’ve tried several of these types over the years, and so far this iteration seems to be pretty good, especially considering the non-specific questions. It got Susan Calvin in the first guess, but Jonathan E took a whole lot of questions and one wrong guess. The game report is interesting for that character - the last 12 questions did not have an expected answer! How does that help??
What that means is that it’s a character for which Akinator doesn’t yet know very much (not played very often, presumably). After a while, though, it eventually rules out a bunch of characters for which it does know those answers, and is left with one that it might as well guess by default. The next time someone plays Jonathan E, Akinator will know a lot more about him.
That makes sense, but it still seems weird.
I stumped it after I posted here. I see a problem with this kind of program. Like, I stumped it with Harry O, the San Diego based PI played by David Janssen whose full name is Harry Orwell. People that liked him called him Harry, those that didn’t called him Orwell, but the show was “Harry O”. So if the question was “do people refer to your character by his last name” (a question to get people like Quincy and Columbo, who don’t officially have first names) what’s the correct answer?
There is also no way to distinguish non-defining characteristics. Does he smoke? Not all the time, not every episode. Does he wear a suit? Sometimes, and in the case of Jim Rockford, is that even a suit at all?
And how do you answer the suit one for someone in Baywatch?
Damn, he’s good! He got Harry Dresden (Dresden Files novels), Stephanie Plum (One for the Money, etc novels), John Taylor (Nightside novels) and Captain Chaos (Cannonball Run) all in under 20 questions. I never thought he’d guess…HIM!
As far as I am concerned, the answer to “Does he smoke?” in that case is “Yes” and the same for “Does he wear a suit.”
Whatever they are calling the middle option these days. Those choices between yes and no do have relevance.
And if Akinator doesn’t know a character, it’s because his users haven’t put him in yet. I hope you guys are putting in these new characters so he’ll have a chance to learn. He’s not Watson–he can’t learn on his own.
If a character is fairly new, he’ll also need you to put in new questions to isolate them. Again, he learns from you, not from some outside database. That’s why he’s only gotten better over the years.
I added the new character. But I was wondering about the process. It doesn’t feel right stumping it if I don’t know how to answer the questions the way they want.
The MOR answers were “don’t know”, “probably” and “probably not”. To me, none of them are equivalent to “sometimes”. I’ll try again with ‘don’t know’ when the answer is ‘sometimes’.
I’m not trying to be a smartass to the game. I’m trying to play fair. I thought Susan Calvin was going to be difficult, but that’s because I forgot the movie ‘version’ of I, Robot. But it got that in about 20 questions. It got Peter Tork (of the Monkees), but it surprised me it took more questions after determining that he played bass guitar in a 4 man band that had a TV show in the 60s. How many others are there that match that?
It took 30 tries for it to get Chode from Tripping the Rift.
Supposedly I was the 298th person to play him.
