--- Richard Loosemore <[EMAIL PROTECTED]> wrote: > Matt Mahoney wrote: > > --- Richard Loosemore <[EMAIL PROTECTED]> wrote: > >> > >> Would an AGI with exactly my (human) intelligence be able to pass your > >> compression test? > > > > Only if your intelligence was uploaded to a deterministic machine. The > human > > brain is not deterministic. > > Then your test is surely pointless, for these reasons: > > 1) It is not at all obvious that an AGI built on non-determinstic > hardware could be "uploaded to a deterministic machine" without breaking > it, so the option of trying to upload it is a non-starter.
You were the one that assumed in the first place there was an AGI with exactly your intelligence. Aside from the technical difficulties of scanning neurons, there is no theoretical barrier. You can always simulate noise with a deterministic pseudo random number generator. > 2) That means that you would potentially be faced with an AGI that was > as intelligent as me, but it would fail your compression test. I can give you 6 billion more examples that would fail... > It would be intelligent (well, it would be as smart as me...), but your > test would say it was not. > > QED. If you implemented an AGI on a nondeterministic machine, then you are right. But, 1. Most people working on AGI are using deterministic hardware. 2. Suppose your AGI has 100 tunable parameters. How do you want to set them? a. Run thousands of experiments and test each one with a Turing test? b. Run thousands of experiments and test each one using text compression? -- Matt Mahoney, [EMAIL PROTECTED] ----- 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=231415&user_secret=fabd7936
