Kevin Cramer wrote:
I tested this and it is very very poor at invariant recognition. I am
surprised they released this given how bad it actually is. As an example I
drew a small "A" in the bottom left corner of their draw area. The program
returns the top 5 guesses on what you drew. The letter A was not even in
the top 5, much less being the first best guess...
Back to the drawing board for this fundamental problem that no one has
solved...including anyone on this list. And I can say with certainty that
until it is that AGI will not come to pass.
I agree that any reasonably powerful AGI that has been given visual
sensors since its childhood
will be able to solve this kind of visual invariant recognition problem
easily.
However, I wouldn't say that this is a prerequisite for human-level AGI:
some AGI's could simply
not be aware of visual stimuli, existing e.g. in a world of mathematics
or quantum-level data, etc.
Novamente for example doesn't deal with low-level vision....
I would certainly expect that a mature Novamente system would be able to
easily solve this kind of
invariant recognition problem. However, just because a human toddler
can solve this sort of problem easily, doesn't
mean a "toddler" level AGI should be able to solve it equally easily.
Different specific modalities will
come more naturally to different intelligences, and humans are
particularly visual in focus...
-- Ben
-----
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/?list_id=303