[Sigia-l] Findability

[Karl Fast]

> And yet, in a few years, a refrigerator may 'understand' more about
> food preservation and detection of spoilage better than you ever
> could.


A lot of what we've been talking about boils down to how we define
'understanding' and 'intelligence.'

The natural way we define these things is in a human-centric way.

We say things like "a machine will understand something when it
understands things the way a human does." The implication then is
that the machine must, internally, either function like a human
functions or achieve an extroardinary capacity for mimicing human
abilities.

But is this a good way to define 'understanding' and 'intelligence?'

Perhaps not.

Let me argue why it isn't (and thus undermine most of what I've
written on this thread).

For many years it was argued that apes did not have the capacity for
language. Then along came Alan & Trixie Gardner who said "Wait a
minute...you're defining language as spoken language. But apes are
physically incapable of making sounds like we do. What about sign
language?"

So they got a chimpanzee, named her Washoe, and taught her sign
language. And now we know that apes have the capacity for language.
A rudimentary capacity, but it's there.

The problem wasn't with apes, it was that we defined language in a
very narrow way that required apes to have certain physical
characteristics that they didn't have. That is, we defined language
in a way that was very human-centric.

So perhaps we are discussing intelligence and understanding in
pretty much the same way (or at least I am).



--karl