Joho the Blog
An Entry from the Archives

The way these things usually work is that if you're thinking of something that it doesn't know, it will ask you to give it a question to distinguish the thing you're thinking of from the thing that it guessed. It builds a tree of questions and goes down that tree. The reason this one appears to be so good is that lots of people have already entered items and questions.

Hope that made sense. This is a programming problem frequently given in intro classes, so anyone with a CS degree should be able to explain it.

Right. I saw the aglorithm many years ago in a programming class. Here's one reference on the web:

I am also impressed. I thought of avocado. It got it in less than ten questions. I thought of worm. It got it in less than 20.

Hmm, the URL got eaten:

http://cs.colgate.edu/faculty/stina/courses/cosc/102/f03/labs/Lab10/Lab10.html

Its amazing. He states that it is online since 1994, if i remember right. The database has been growing since. It "guessed" Gravity, money, cloud, Moon, Wind, song, music all within 28 tries. Interestingly money can have 4 corners AND be of metal. What would confuse a person makes it even easier for an app.

I was impressed with the comments above (except flashlights) so I gave it a try. I thought of a black lab. Not only did it not get it (it guessed "yellow lab" at question 18, after I had answered YES to "Is it black?"), but it then asked me what I was thinking of. I told it a black lab and it compared my answers to the way it would have answered the question. Examples:

Is it outside? You said Sometimes, I say No.
Is it known for its ability to dig? You said Yes, I say No.
Does it like sugar? You said No, I say Yes.
Is it originally from Europe? You said Yes, I say No.
Is it a predator? You said No, I say Probably.
Does it contain a liquid? You said Sometimes, I say No.

Then it gives me "Uncommon Knowledge about black labs"

Is it a common household object? I say Probably.
Do you find it in the sky? I say Probably.
Does it squeak? I say Probably.
Does it have a long neck? I say Yes.
Does it have a horn? I say Probably.
Is it red when ripe? I say Probably.
Do you use it at work? I say Probably.
Do you clean it regularly? I say No.
Does it display information? I say Yes.
Does it use electricity? I say Probably.
Do cats like to chase it? I say Yes.
Is it an article of clothing? I say Probably.
Is it lifeless? I say Probably.
Is it smart? I say Doubtful.
Is it organic? I say Doubtful.
Does it eat grass? I say Yes.

I'm a bit worried about my black lab. She never displays information. Maybe I need to plug her into the nearest outlet!!!

Uncommon Knowledge about a clitoris
Would you pay to use it? I say Yes.
Can you mash it? I say Yes.
Does it have buttons? I say Yes.
Do you need batteries to use it? I say Probably.
Does it come in different colors? I say Yes.
Does it change colours? I say Yes.
Does it sweat? I say No.
Is it electronic? I say Yes.
Is it red? I say Yes.
Is it originally from the Mediterranian? I say Probably.
Can it be played? I say Yes.
Does it have a good smell? I say Yes.
Does it contain a liquid? I say No.
Does it smell good? I say Yes.
Does it stink? I say Yes.
Is it worn on the human body? I say No.

20q is a great game it is so cool it is like it can read your mind it amazes people

I dont know if any of you tried it, but if you puzzle it with "penis" it will indeed guess dik dik. Go ahead and try it. Give it a whiz.

20Q works like a condom. You put it in, you pull it off when you are done. THROW IT AWAY, you shouldnt reuse it. FUCK THAT.

Yea, I LOOOVE this stuff!!!

I like electronic handheld gadgets and came across some reviews on Amazon about the 20Q. 20Q is the handheld version of the website version. It's has much fewer supossed 'synapses' and it doesn't grow 'smarter' as more users use it. It kind of reminds me of a reverse definition look-up.

Since the handheld model is static, I got to thinking about something. You can answer, Unknown, No, Yes, or Sometimes, to each of the 20 questions. Well, the 1st question is more like, 'You must choose a category: Animal, Vegetable, Mineral, Other, Unknown. After you say 'Yes' to one of these it moves on to the inquiries. This is where my thinking starts. ;) If it were to ask you 19 questions and you have 4 possible responses then everytime you repeat the same pattern of responses you'll get the same guess by 20Q. Why? Because 20Q's questions aren't random but elicit the next question. If it asks, "Is is flat?', does that mean the next question will be, 'Is it brown?' Not necessarily. It depends upon the preceeding answers. If its 4th question is, 'Is it flat?' and the preceeding 3 questions were answered the same then yes, it asked, 'Is it flat' because of the pattern of responses you entered elicited that question. But if you get that 'Is is flat' question on the 17th question and at another time it'd asked it on the 4th question then the 'flat' question wasn't elicited based on the same patterns being repeated but was elicited based on another pattern that elicits that same question BUT may not mean that from there on the questions will be the same because it takes into consideration the ENTIRE history of your responses.

Now to my point and to make this a bit simpler. Let's assume there's no 'category' question (animal, vegetable, etc) but that there's 20 questions and each question allows for 1 of 4 responses. Let's also assume that anytime you have a pattern of response that is different than another pattern that this will elicit of different end-of-the-game guess by 20Q. If this were true then what's the extent of 20Q's database? I mean, how many different answers are available?

To give you an example of what I mean by, 'pattern of responses', let's pretend these are the responses I gave to 20Q, in order (U=Unknown; N=No; Y=Yes; S=Sometimes):
Y,Y,Y,Y,N,U,S,S,N,N,S,Y,Y,Y,S,Y,N,U,N,Y.

OK, now EVERYTIME I enter that 'pattern of responses' 20Q will make the same end-of-game guess EVERYTIME. So, the handheld version isn't really an active 'intelligence' but is really a collection of data taken from the website responses. It seems as though some sort of logic or intelligence had to be used to generate the data but once the data was generated and put inside this toy it then really became a database. 'You put in this pattern of responses and I will yiled this 'guess'. Now for the BIG question!

Can someone tell me how many possible 'patterns of response' are possible? I know this isnt the case but I'm going to assume that everytime you enter a unique 'pattern of responses' 20Q's end-of-game guess will be unique and different from all its other guess. If this were so, how many different guesses or terms are inside that little brain of his? Does anyone know how to figure up how many 'patterns of response' are possbile, based on 20 questions and 1 of 4 possible responses for each question? At one time I was thinking that it would be 20^4 (20 to the 4th power) or 160,000. But I did some smaller scale investigations and all of those seemed to indicate that I have my numbers reveresed. Instead of 20^4 it should be 4^20! If that's the case that would mean there are 1,099,511,627,776 unique patterns of response possible! That's approximately 1.1 TRILLION! My little tests seem to say that's right but my commonsense says...NO WAY! IF that 1.1 trillion number IS accurate we all know that 20Q couldn't possible hold that many unique answers. Remember, I'm talking about the $10-$15 handheld version. In fact, I don't think there's that many words/items in the dictionary. Also, I KNOW if you had a handheld electronic gizmo that was capable of holding 1.1 trillion words it'd cost a lot more than $10! I have an electronic Franklin Dictionary (speaking) and it has 500,000 synonyms and about 250,000 definitions (I think thats right. Its the Merriam-Webster Collegiate). If each definition averaged 50 words that'd be approx. 12.5 million words in the dictionary and about .5 million in the synonym. Let's just say there's a total of 20 million words in the whole she-bang. If you purchase it minus the speaking technology I think its about $60. If my math serves me well, that means the little $10 20Q with its fascinating mind-reading ability, has about 55,000 times more words in it than the $60 Franklin! If these items were priced on a per-word basis and you kept in line with Franklin's per-word rate, the 20Q's cost would be 3.3 million dollars! At Wal-Mart maybe 2.7-2.8, but nonethless, a significant boost beyond the original $10-$15.

OK, so maybe there AREN'T that many answers within the cortex of my much beloved 20Q but does that necessarily mean that the total number of unique 'patterns of response' is any less? I mean, don't these numbers exist independent of 20Q?

Can anyone help me, PLEASE???

"everytime you repeat the same pattern of responses you'll get the same guess by 20Q"

This is not the way the game works. Try it! It does *not* ask you exactly the same question each time you play even if you answer the questions the same way each time.

this sucks

why u people so damn stupid?

Nice grasp of English and sentence structure.

poker me up