The Singularity is Near! (Internet 20 questions)

It couldn’t get the Space Biker (the character that Gordon Freeman replaced from Half Life). I ended up adding the question “Did your character make it into the final version of the work they’re related with?” I’m not sure if that’s fair or not, but I figured I’d try.

It gets its data from its users. That’s also why the questions are repeated a lot: different people ask the question different ways. And people make mistakes, so it has to try and make up for that. I’ve put definite NOs and been wrong.

Failed to guess Jethro Tull (the agriculturalist, not the band) even though he was in the database.

It just guessed the blonde chick from the Evony ad. I kid you not.

OK, now I’m impressed. It got William of Baskerville right at 20. It had “English detective” by about question 10, then floundered a bit, then asked if he wore glasses on question 19.

Hari Selden took 16, and it didn’t look like it was getting close until it nailed it. You wouldn’t think “scientist who’d been into space in a novel” would be worth much, but apparently it was.

Isaac Newton took 20, and right after asking “Is he a scientist”, it asked “Has he represented his country in his sport”.

Sir Gawain took about 25, but I had a lot of noncommittal answers on him.

For Barack Obama, it got as far as “dark skinned politician who’s lived in the white house”, then needed about five more questions (no, he doesn’t use a sword, and he doesn’t have fur).

Voltron took 15 guesses.

Aule was a new character for it, though it did have a few other folks from the Silmarillion.

It just got:
The Warden/main character from Dragon Age
The Protoss Executor from Starcraft.

It found Nefertiti fairly quickly, but it couldn’t find Major Major Major Major (a character from Catch-22).

It’s not even close to sentient. It’s a fairly simple example of a “decision tree”, which is used fairly commonly in computer programming.

Here’s how it’s done. Your tree has two things, “branches” and “leaves”. A leaf has a name. Branches are where a decision has to be made. You start off with a single leaf in your tree. Let’s say, we pick “George Washington.”

So the first person comes along to play the game. They are thinking of someone, say, Napoleon. The game doesn’t have any branches in it yet, so it says “George Washington!” The person says no. So now a new branch is made. It has George Washington on one leaf and Napoleon on the other. So now you need a question for your branch. Some of these programs let the people playing the game put in a question. Others let you submit the name and whoever manages the tree picks the questions. Either way, someone enters a question. Let’s say we pick “Is your character American?” If they say yes, they go to George Washington. If they say no, they go to Napoleon.

Now someone comes along and enters Betty Boop. The game gets to the first branch, which is “is your character American?” and answers yes. This takes them to George Washington. This again isn’t right, so now we have to make another branch at this point. “Is your character male?” will work. So now our “tree” has the top branch, which asks if your character is American. If yes, it goes to a second branch, which asks “is your character male”, or if no it goes right to the end leaf of Napoleon.

Setting up the tree is a bit tedious, but for each person you enter you only have to enter one branch question. The more evenly spaced your tree is (which depends on how well you choose your questions) the more quickly on average it will reach an end leaf. For a perfectly balanced tree you can differentiate 65,536 people in only 16 yes or no questions. A real world tree isn’t going to be perfectly balanced at all, so some branches may end fairly quickly (under 10 questions for example) and others may go on for quite a bit (40 questions or more for example). Whoever manages the tree can look for long branches and can intentionally split the branch to more evenly balance the tree.

With only a handful of names in it, the game tends to suck. But as more and more names are entered into it, the game quickly gets “smart”. If you are setting one of these things up you can let your family, relatives, and friends all play it and keep putting names into it until it gets to the “smart” level, and then you can release it to the public. The more people who play it, the more names will get entered into it, especially as more and more people try to fool it with obscure names. Every time someone guesses an obscure name, it gets entered into the tree, so eventually even the most obscure names tend to be in there somewhere.

This particular game is a bit more advanced than some that I’ve seen in that it has more than just a yes/no decision at each branch. With 5 choices (yes/probably/I don’t know/probably not/no) at each branch you more quickly get down through the data to unique leaves. But then also, the “I don’t know” option can mean that you have the same data in multiple leaves.

As for its “sentience”, this is dirt simple for the computer. It doesn’t know or care what the text says. All it does is move a pointer to the first branch. Then, based on your input, you tell it what branch to go to next. When it gets to a leaf it prints out the data in that leaf and says that it is done. Any first year computer science student should be able to program something like this.

My first one was Mickey Mouse - it got that one easy. Then I thought, let’s see it guess Sir Robin The Not-Quite-So-Brave-As-Sir-Lancelot. 15 questions and it got it! Then it got Raistlin Majere in 20 - I thought it would be thrown off by answering Yes to being part of a cartoon.

Stumped it with Pete Puma.

I have been amazed by the hand-held version for a few years now. As in, it sometimes freaks me out. I love the on-line version. I know there’s a science behind it, but it’s pure enjoyment for me!

It doesn’t really have a sense of what questions are important or what answers are incompatible with other answers either.

Here’s an example of the flailing about it does while failing to guess Jakaya Kikwete, President of Tanzania:

Does your character have children? yes
Is your character into politics? yes
Has your character died ? no
Is your character dark-skinned? yes
Has your character ever lived in the White House? no
Is your character African? yes
Has your character been arrested? no
Does your character have a famous son? no
Does your character live in America? no
Is your character Muslim? yes
Does your character wear a hat (cap, crown,…)? no
Is your character black? yes
Is your character of Arabic origins? no
Is your character bad? no
Is your character a politician? yes
Has your character been the president of a country? yes
Can your character be seen on television? no
Has your character dicovered something? no
Is your character a woman ? no
Is your character often considered as evil? no
Guess: Thabo Mbeki
Has your character been in jail? no
Is your character real? yes
Does your character have Asian origins? no
Is your character a citizen of the United-States? no
Did your character have trouble with the law? no
Does your character forget easily? no
Is your character European? no
Has your character governed his country? yes
Is your character a girl? no
Does your character have grand-children? probably no (a mistake - I meant to say probably yes)
Guess: Abdelaziz Bouteflika
Is your character Egyptian? no
Is your character a member of the current government? yes
Does your character have white skin? no
Has your character killed humans? no
Is your character linked with sports? no
Is your character American ? no
Is your character Asian? no
Has your character been elected president at least twice? no
Does your character belong to a Faction? probably yes
Has your character run for president? yes
Guess: Morgan Tsvangirai

It takes till question 18 to ask “male or female” and 22 to ask “real or fictional” - pretty fundamental questions. After having multiple confirmations that the character is an African politician it’s still asking questions like “Is your character a girl”,“Is your character Asian” and “Is your character linked with sports”. A human, after about question 10, would be quartering Africa for presidents, but of course this program can’t do that, because it doesn’t really know what “Africa” or a “President” is.

I think you’re easily impressed. It’s a pretty basic website to throw together with a half decent database of popular culture. It’s not that great, it couldn’t figure out Nero Wolfe, instead it suggested MacGyver followed by Hurley from Lost.

Yes, it was one of the first things I wrote in college (in pascal even.) Mine created a decision tree and added to the tree as answers were forth-coming. Today’s are better than mine was, but it wasn’t bad for a novice programmer.

I kicked his ass 3 times on the trot with Lucas Davenport, Dave Robichoux and Lurch.

Freaky enough when it works. Penfold from Danger Mouse in 13.

I picked Bobby Darin, which I thought would be easy enough, but it answered Nick Lachey on its 20th try. Haha! I continued, and ended up stumping him.

Utterly failed on Slartibartfast from Hitchhiker’s Guide, but it got Shaft in 14.

Niel Armstrong - 19 Questions
Grover Cleveland - 17 Questions
Mike Collins - Stumped

Getafix (from Asterix) in 17 questions; impressive.

Rincewind - 14 guesses
Harry Dresden - 15 Guesses
Radagast - Stumped.