--- Panu Horsmalahti <[EMAIL PROTECTED]> wrote: > An AGI is not selected by random from all possible > "minds", it is designed > by humans, therefore you can't apply the probability > from the assumption > that most AI's are unfriendly.
True; there is likely some bias towards Friendliness in AIs designed by humans. However, the magnitude of the pre-existing anti-Friendly bias is so large that it isn't likely to help. Just look at the AI designs that have been proposed by human researchers- the vast majority of them would have killed us all for some stupid thing like computronium or iron blocks. > There are many > elements in the design of an > AGI that most researchers are likely to choose. I > think it is safe to say > that the average AGI design by a human is extremely > different from one made > by an average alien (for example). Again true. The problem with this is that the pro-human optimization inherent in human design isn't anywhere near strong enough. It's like improving on your odds of quantum leaping over the Empire State Building by climbing the first fifty floors. > Genetic programming is not stupid with adequate > processing power, and > Moore's law makes that risk closer every year. I'm not familiar at all with the specifics of genetic programming. Do you have a good introduction? - Tom > ----- > This list is sponsored by AGIRI: > http://www.agiri.org/email > To unsubscribe or change your options, please go to: > http://v2.listbox.com/member/?&; ____________________________________________________________________________________ Need Mail bonding? Go to the Yahoo! Mail Q&A for great tips from Yahoo! Answers users. http://answers.yahoo.com/dir/?link=list&sid=396546091 ----- This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=4007604&user_secret=7d7fb4d8
