RE: Think cash
Marcel Popescu[SMTP:[EMAIL PROTECTED]] wrote: My proposal was to randomly create an image, which should be 1) easily recognizable by a human (say the image of a pet), but 2) complex enough so that no known algorithm could "reverse-engineer" this. [You need a randomly-generated image because otherwise one could build a large database of all the possible images and the correct answers.] Background information would also be very useful - see http://www.digitalblasphemy.com/userg/images/969403123.shtml - it's easy for a human being to identify the animal in the picture, but (AFAIK) impossible to write a program to do the same thing. Ideas? At 01:53 PM 10/11/00 -0400, Trei, Peter wrote: You refer the the problem of recognizing a photo of an animal. It used to be said that no computer program could reliably distinguish between a dog and a cat, but I'm not sure that's the case since the development of neural networks. Blind humans aren't always good at recognizing screen images. Neural networks are good at recognizing things. Sometimes more precisely defined algorithms are good too. Some examples of recognition systems - you can look in the archives for pointers to the UCBerkeley "Naked People Finder", which does a reasonably accurate job of distinguishing whether pictures on the internet contain naked people. The people who did the research on that also designed the "Incredible Horse Finder", which identifies horse pictures on the net. I remember that those systems did a lot of modelling; I don't remember if they also did neural nets or not. If they wanted to describe shapes of dogs and cats and differentiate between them, it would be relatively doable. There's also a company out there that does "passfaces" - they pop up 9 pictures of people's faces, and you identify which one is in the set that's you password-equivalent. They do about 4 rounds of this, with random sets of faces; it's closer to a PIN than a real passphrase in strength, because they thought that was enough for their problem space. An interesting aspect of it is that humans are very good at recognizing faces, but not usually that good at describing them, so it's hard to give somebody else your passface set. Thanks! Bill Bill Stewart, [EMAIL PROTECTED] PGP Fingerprint D454 E202 CBC8 40BF 3C85 B884 0ABE 4639
Re: Think cash
At 11:54 AM 10/12/00 -0400, James A.. Donald wrote: -- At 12:59 PM 10/11/2000 -0400, Marcel Popescu wrote: An interesting idea has surfaced on the freenet-chat list: is it possible to build a program that creates some sort of a puzzle, whose answer the generating computer knows (and can verify), but which can only be answered by a human being, not by a computer? [Additional requirement: it should be easy for the human to answer the puzzle.] Origami world. Computer generates a random 3D object out of large polygons with fairly sharp angles of contact, subject to various limits on the way in which the object is generated. Displays 2D image of 3D object. Human infers 3D object from 2D image, infers unseen portions of the image from rules by which the 3D image is generated -- for example that the object must make sense mechanically -- that it should be stable resting on a plane. You seem to be supposing that human perceptual algorithms (and the illusions they produce) are somehow unknowable or unreplicable by nonanimal machinery. This is meat chauvinism. Look into David Marr's _Vision_ for starters... or Grossburg's (of BU) stuff.. Now back to your regularly scheduled spam laced with cryptography
RE: Think cash
-- Marcel Popescu[SMTP:[EMAIL PROTECTED]] wrote: An interesting idea has surfaced on the freenet-chat list: is it possible to build a program that creates some sort of a puzzle, whose answer the generating computer knows (and can verify), but which can only be answered by a human being, not by a computer? [Additional requirement: it should be easy for the human to answer the puzzle.] My proposal was to randomly create an image, which should be 1) easily recognizable by a human (say the image of a pet), but 2) complex enough so that no known algorithm could "reverse-engineer" this. [You need a randomly-generated image because otherwise one could build a large database of all the possible images and the correct answers.] Background information would also be very useful - see http://www.digitalblasphemy.com/userg/images/969403123.shtml - it's easy for a human being to identify the animal in the picture, but (AFAIK) impossible to write a program to do the same thing. Ideas? Mark That's a really interesting question. My off-the-cuff answer would be 'no'. The constraints which say that the problem is randomly generated by a computer and the answer also evaluated by a computer are the killers. Any problem which one computer can create, and solve, can also be solved by another. Perhaps one could generate the solution, and find a problem which is solved by that solution, but finding a type of problem which humans will always solve one way, and computers another is the rub. You refer the the problem of recognizing a photo of an animal. It used to be said that no computer program could reliably distinguish between a dog and a cat, but I'm not sure that's the case since the development of neural networks. Almost any question which has a solution which is clear, unambiguous, and easy determined by a human can probably also be solved by either a regular program or a neural net. What you are really attempting to find is a reliable, fast, single-question Turing test. I'm far from sure this is possible. Peter Trei
Re: Think cash
At 12:59 PM 10/11/00 -0400, Marcel Popescu wrote: Real-To: "Marcel Popescu" [EMAIL PROTECTED] An interesting idea has surfaced on the freenet-chat list: is it possible to build a program that creates some sort of a puzzle, whose answer the generating computer knows (and can verify), but which can only be answered by a human being, not by a computer? [Additional requirement: it should be easy for the human to answer the puzzle.] My proposal was to randomly create an image, which should be 1) easily recognizable by a human (say the image of a pet), but 2) complex enough so that no known algorithm could "reverse-engineer" this. [You need a randomly-generated image because otherwise one could build a large database of all the possible images and the correct answers.] Background information would also be very useful - see http://www.digitalblasphemy.com/userg/images/969403123.shtml - it's easy for a human being to identify the animal in the picture, but (AFAIK) impossible to write a program to do the same thing. I don't follow the other list you mentioned, so I don't know what the actual problem to solve is - my guess is that this is an anti-bot protection measure, intended to make sure that only human participants can engage in a conversation. If that's the problem - or if it's similar - you'll also need to make the puzzle difficult enough that it's hard to brute-force or solve statistically - let's say you provide the other party with 20 images, 19 cats and 1 dog, and ask them to identify the dog. What keeps a bot from answering the question 20 times? Let's assume the first arms-race countermeasure prevents answering the question more than once by generating puzzles on-the-fly from known cat and dog images - so the bot just picks an answer randomly, and keeps doing that until they hit. Can God create a rock so big he can't lift it? I think you're barking up the wrong tree, thinking about "known algorithms" and such - just like with crypto, the real way in isn't to attack the strong front door, but to just go around it. This sounds like maybe it's essentially a credentialling/ID problem, where you're generating credentials on the fly based on a short-form Turing test. Can you restate the problem so that instead of a Turing test it's a more familiar multi-channel authentication process? (e.g., require new participants to have "introductions" from existing participants, track introductions, and remove the access for accounts found to be bots, or found to have introduced bots .. or similar.) -- Greg Broiles [EMAIL PROTECTED]
