So you mean compression, right? When you say find the structure in s, you mean a compressed representation that decompresses to s, right?
On Wed, Apr 17, 2013 at 11:09 AM, Alan Grimes <[email protected]> wrote: > On Wed, Apr 17, 2013 at 2:48 AM, Sergio Donal <[email protected]> wrote: >>> >>> Please, could you explain it a little bit further? >>> I do not understand it. > > > Okay, that's fair. > >>> The premise is that the input is structured, why is structure important? > > > Because that's where the useful information is. > >>> What is 'f'? >>> Do you mean that 'f' is the intelligence-function, so intelligence is >>> about finding the structure in the input? > > > Yes. > >>> Why does 's' have to be a 'proper subset'? This point confuses me. > > > Well, the AI is not omnicient. Furthermore the input might be ambiguous. So > therefore 's' is the structure that is recoverable from the input. > >>> And also, what happens if the intelligence has some kind of structure >>> itself (like memory from other derivations) so it is able to make sense of I >>> and make it useful for solving a problem, even when I was just random? > > > If I is random then 's' is null because there is no useful information. > > The function can trivially be made iterative by: > > f(I,s) --> s' > > Which is obviously an improvement... > > Now, how do you get AI? > > First you need input. There are two ways that this can be done, both should > be applied. > > 1. Generate a structure at random and then generate an image or sound from > it. (this way you have a known "gold standard" of the structure). > > 2. Use sound/pictures/videos from your CD/DVD collection and from around the > web. > > Second, you need a way to feed this into f... I'm not exactly sure how this > would be done. > > > Third, you need to develop f... > > > Fourth you need to take s' and compare it with the gold standard s too see > if they're equivalent. > > Or, if you were using stock photos/videos/sounds etc, then you will have to > use the recovered structure to attempt to re-generate the original. The > success of your algorithm can then be measured by how closely the re-created > data matches the input. > >>> I believe that I am missing the subject and object of the statement, like >>> is the intelligence a property of someone who finds the structure in I, or >>> something like that. > > >>>> Given some structured input I, with structure S >>>> >>>> AI is defined as >>>> >>>> f(I) --> s >>>> >>>> where s is a proper subset of S that can be derived from I with minimal >>>> ambiguity. >>>> >>>> This definition is partial but covers all forms of perception. > > > > -- > NOTICE: NEW E-MAIL ADDRESS, SEE ABOVE > > Powers are not rights. > > > > ------------------------------------------- > AGI > Archives: https://www.listbox.com/member/archive/303/=now > RSS Feed: https://www.listbox.com/member/archive/rss/303/3701026-786a0853 > Modify Your Subscription: > https://www.listbox.com/member/?&; > Powered by Listbox: http://www.listbox.com -- -- Matt Mahoney, [email protected] ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
