/ic/ - Artwork/Critique

>>6858541
>If the AI cannot take one image and learn from it to make variations of it without additional data, but can only replicate, then it does not learn, it just copies.
yes, it can only do that if it has additional data showing it the range of possibilities.
but this does not mean what you think it does.

if you show a human who has never seen a rhinoceros beetle and doesn't know anything about it something like pic related. would they be able to draw this beetle in other situations, from other angles?

the answer is yes, maybe not accurately, but they can try. but how would they do that? they would probably refer to their knowledge of other beetles, or insects in general.

but what do you call this process. what is the human using here to fill in the gaps?
the answer is: OTHER. DATA.

which brings me back to what i said before: but if the AI has only one image, its entire existence is that one image.

make sense so far? and do you understand what i'm trying to say here?