I explain it fully in my AGI handbook though :D.

earth
e a r t h
e  a  r  t  h
htrae
e
a
r
t
h

are all Earth

In vision, a tiny boat is recognized as a big boat if you see a tiny boat for 
the first time and only know in memory a big boat. And you can fill in the rest 
of the image if only see half a tiny boat, to recognize/complete a tiny boat 
(I'll show for Large here:) you simply predict the pattern that if the e and a 
and r and t are all have a large expected distance error, then h will too, so 
you predict the next letter/ pixel but with a extra distance off like the 
others. Very straightforward now to recognize all sorts of features in vision, 
it's just like text prediction! It may use this mechanism more but really it's 
all the same stuffs and uses many of the things text prediction uses, mostly 
the same really.
------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/Td08f1cb9cdd5e5d9-Mfec689befdca072d9e3ae6bc
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to