Hi again, I now got to reading Ben's essay and would like to place my comments here.
On 9/10/06, Ben Goertzel <[EMAIL PROTECTED]> wrote:
Not to be left out, I also wrote down some of my own thoughts following our interesting chat in that Genova cafe' (which of course followed up a long series of email chats on similar themes), which you may find here: http://www.goertzel.org/papers/LimitationsOnFriendliness.pdf
Note to Ben: I have recently (~4 week ago) ordered your book on AGI (http://www.amazon.de/gp/product/354023733X/) on Amazon but it has not shpped yet. Are you in the loop when it may be available - on Amazon it said August 2006. Sorry to bother you ;-) On Ben's essay: Ben is arguing that due to incomputable complexity 'friendliness' can only be guranteed under unsatisfactory narrow circumstances. Independent of one agrees or not, it would follow that if this is the case then substituting friendliness with one or all of the alternative goals proposed by Ben namely "compassion", "growth" and "choice" would not make a difference as the same incomputability problem applies to all goals. Reading the essay I got reminded of Marc Hutter's work on the Universal Algorithmic Agent AIXI (http://www.idsia.ch/~marcus/ai/aixigentle.htm) From the abstract: "Decision theory formally solves the problem of rational agents in uncertain worlds if the true environmental prior probability distribution is known. Solomonoff's theory of universal induction formally solves the problem of sequence prediction for unknown prior distribution. We combine both ideas and get a parameterless theory of universal Artificial Intelligence. We give strong arguments that the resulting AIXI model is the most intelligent unbiased agent possible. We outline for a number of problem classes, including sequence prediction, strategic games, function minimization, reinforcement and supervised learning, how the AIXI model can formally solve them. The major drawback of the AIXI model is that it is uncomputable. To overcome this problem, we construct a modified algorithm AIXItl, which is still effectively more intelligent than any other time t and space l bounded agent. The computation time of AIXItl is of the order t·2l. Other discussed topics are formal definitions of intelligence order relations, the horizon problem and relations of the AIXI theory to other AI approaches." It follows that the AIXItl algoritm applied to friendliness would be effectively more friendly than any other time t and space bounded agent. Personally I find that satisfying in the sense that once "compassion", "growth" and "choice" or the classical "friendliness" has been defined an optimal algorythm will be available to achive the goal. Best regards, Stefan -- Stefan Pernar App. 1-6-I, Piao Home No. 19 Jiang Tai Xi Lu 100016 Beijing China Mobil: +86 1391 009 1931 Skype: Stefan.Pernar ------- AGIRI.org hosts two discussion lists: http://www.agiri.org/email [singularity] = more general, [agi] = more technical To unsubscribe or change your options, please go to: http://v2.listbox.com/member/[EMAIL PROTECTED]
