[ExI] ai class at stanford

[Kelly Anderson]

On Tue, Aug 30, 2011 at 10:13 AM, Adrian Tymes <atymes at gmail.com> wrote:
> On Tue, Aug 30, 2011 at 5:26 AM, Kelly Anderson <kellycoinguy at gmail.com> wrote:
>> So, if we pass the Turing test, for example, without understanding
>> 100% how humans do it, then we understand how humans talk "well
>> enough" to be useful and reproducible. So in this "engineering sense",
>> the Turing test says we understand "intelligence" to a particular
>> measurable level.
>
> Standard terms (such as "well enough") in quotes to flag similes is a
> warning flag.  You give the impression that you do not understand that
> "well enough" means "well enough" in all senses, and that you might
> not be able to distinguish between your use of quotes-as-similes and
> this paragraph's use of quotes-as-designators.  This suggests that
> rational conversation may be impossible.

I can live with well enough losing the quotes. Let's go with Alan
Turing's standard. Well enough in this case means that the simulation
would be indisinguishable from you by people who know you well. That
is, the Adrian upload would have to be able to fool your friends into
thinking it is really you for some period of time to pass the test. Is
that a little less fuzzy?

> Put another way, consider the difference in intended meaning between
> "we understand how humans talk 'well enough' to be useful" and  "we
> understand how humans talk well enough to be useful".

I meant it in the sense that we understand physics well enough to
build a bridge and have it stand up most of the time.

>> So rather than calling this the "Turing" test, we'll call this the
>> "Adrian" test.
>
> ...and there's the ad hominem.  The expected value of the rest of your
> post is low enough that I'm not even going to read it.  I will, however,
> post this just in case those were honest mistakes, so you can reply
> without making yourself appear not worth talking to in the future.

I'm not sure why I appear not worth talking to... but what I'm
proposing is that the test for a successful upload would be that your
friends could not distinguish between you and the upload for some
period of time, to be determined. I think that is sufficiently
rigorous for at least some purposes.

-Kelly