[Haskell-cafe] Neural nets and the menu (was: something different)
[EMAIL PROTECTED] wrote: Albert Y. C. Lai writes: Achim Schneider wrote: There is this story about some military (US afair) training a neural net to detect tanks in images ... 50% accuracy. I have some similar stories to tell A. ... students assumed sin(x+y) = sin(x) + sin(y) B. ... But that day, that car, it showed no red light. All this is nice to read, and make me thank Lord I am still alive, but, for goodness sake, *what do you want to convey?* Probably that you have to come up with something more confusing and powerful than a neural net and then transfer your consciousness into it to understand yourself and how to program something like you. No quanta, no Wikipedia, no first-class objects, no Haskell, café I have to fetch myself in la cuisine, and I feel lost. There is always a good chance to get an Espresso-Baileys in any good café. Or Baileys-Espresso, if you prefer. If you're lucky, you can also get mead in your coffee. I recommend any kind of sandcake with it. -- (c) this sig last receiving data processing entity. Inspect headers for past copyright information. All rights reserved. Unauthorised copying, hiring, renting, public performance and/or broadcasting of this signature prohibited. ___ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe
Re: [Haskell-cafe] Neural nets and the menu (was: something different)
Achim Schneider writes: [EMAIL PROTECTED] wrote: ... *what do you want to convey?* Probably that you have to come up with something more confusing and powerful than a neural net and then transfer your consciousness into it to understand yourself and how to program something like you. Oh my, really? I have to discuss this with my Master, R. Daneel Olivaw. My turn to tell an anecdote. Have you heard about the homework of Theodore Hill (mathematician)? He asked his students to flip a coin 200 times, or use some other serious random generator 200 times, and to report the result sequence. There was an option for lazy people, they could fake the sequence and produce it from the thin air, as intelligent humans could. Then he took the homeworks, scanned each for a couple of seconds, and with a remarkable accuracy selected the cheaters. How? == If you haven't hear the story, check the Web for Benford's Law (discovered by Newcomb 50 years earlier), and all the business which came with it, notably some automated ways of discovering human cheaters who try to fake *natural* data. Instructive, and despite reasonable analysis, still a bit mysterious in its essence. I find it nicer than cheaters who try to convince me that automata may fake human behaviour. Jerzy Karczmarczuk ___ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe