Not all 'integral' objects can be recognized from visual edge filters. Some
years ago I gave myself a few days to make an edge detector myself and I got
amazing results where a person or an solid object was a different color and
contrasted against the background. But when the pixels were mixed up and the
same bright variations of pixels existed in the 'object' and the background the
process fell apart. (If I tried it today I would use slightly larger squares
and average the colors and so on.) There is no 'predicting' the pixels in the
area but you would be interested in similar colors for most solid objects. So
patterns would be useful but it is not as simple nor as straight forward as
when you might look for patterns in huge samples of text. The use of
vector-like data, which can be used with varying sizes because the magnitude
can be easily changed for the whole sample is interesting, but you also have to
account for rotations and even rotations in the third dimension.
Notice that you needed to expand your explanation of what you were saying and
you chose to expand it into more discretized dimensions. "convert", "predict
the next pixels", "distance pattern recognizer", then an example, "recognition
of various sizes", "brightness" and so on.
------------------------------------------
Artificial General Intelligence List: AGI
Permalink:
https://agi.topicbox.com/groups/agi/Td08f1cb9cdd5e5d9-M0853c521f7f8f09c0f78e6c9
Delivery options: https://agi.topicbox.com/groups/agi/subscription
