Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

33 Questions

github.com

Post reply on HN
Print view View original

51–60 of 169 posts

Re: 33 Questions

#51

Post by Houshalter »

powrtoch wrote: post #33

I don't think this problem is solvable in any elegant form, but it is solvable. You'll just end up with massively conjunctive questions that you can't even hold in your head at once, like "27: Are you a non-practicing Catholic with exactly three children, or an asian owner of a minivan produced between 1998 and 2004 that isn't green, or a licensed boat mechanic with astigmatism, or..." and so on for the next 6 pages.…

Those are bad examples because those questions don't split the population in two. Very few people are non-practicing Catholics with exactly three children. If you want to limit the number of questions to just 33 then you have to choose your questions very carefully.

Re: 33 Questions

#52

Post by mcphilip »

I think first you have to show a question exists that effectively separates identical twins before you spend much time working on broad questions like gender and geography.

Re: 33 Questions

#53

Post by scraplab »

ZirconCode wrote: post #11

Just use: - Birthday (19~ bits) - Rough Location (remaining bits) And base the questions around those two, for example, where you born on a 1-15th, does the city you were born in start with the letter's a-k. This part would be an exercise in statistics, I would think. edit: And one bit for if you were the first to be born of two identical twins =p

Lots of people don't know their birthday, or if they do, it was written down as 1st Jan (or another memorable date), because their parents/carer didn't know when they came to register them, if they ever registered them. There will be bias towards those memorable dates.

Re: 33 Questions

#55

Post by fat0wl »

Laremere wrote: post #42

Assuming that the person doesn't necessarily need to know their answer (which is important for babies anyways) the answer is trivial. The first question would be "Given that we ordered all humans in order of the time of their birth, would the 1st bit of your position in the ordering be 1?", continue the other 32 questions with the remaining 32 bits.

lol too much

Re: 33 Questions

#56

Post by vinceguidry »

anilshanbhag wrote: post #38

33 is not the constraint. If we increase the limit to 50 and these 50 questions can fingerprint an individual then that will be really interesting. Some hard problems :- 1. Distinguish twins 2. Using characters in names as some like Chinese use non-ascii names.

You could distinguish twins by first identifying the twins, then by asking which one's name starts earlier in the alphabet, or whatever analogue exists in languages without an alphabet. Encode appropriately.

Re: 33 Questions

#57

Post by jld »

Houshalter wrote: post #51
powrtoch wrote: post #33

I don't think this problem is solvable in any elegant form, but it is solvable. You'll just end up with massively conjunctive questions that you can't even hold in your head at once, like "27: Are you a non-practicing Catholic with exactly three children, or an asian owner of a minivan produced between 1998 and 2004 that isn't green, or a licensed boat mechanic with astigmatism, or..." and so on for the next 6 pages.…

Those are bad examples because those questions don't split the population in two. Very few people are non-practicing Catholics with exactly three children. If you want to limit the number of questions to just 33 then you have to choose your questions very carefully.

The example was one question with a bunch of OR operators that when combined, would equal exactly 50%.

Re: 33 Questions

#58

Post by fat0wl »

The 33-question issue is a tough one for sure.

I'm instead left wondering how many extra questions (35 bits? 36 bits?) it would have to be expanded to in order to produce unique results but without having to be particularly clever in producing the questions. I bet it wouldn't take as many extra as one might be inclined to think.

Post reply on HN