One strange thing you may have noticed in IGPT/DALL-E on openAI.com is "how in
the world does it know to predict reflections, and how". In fact my AI already
has such ability. Things in the recent history of data it seen predict
themselves more heavily the more they are seen and if more recently. Hence
birds repeat unto in the water beneath them, which we call reflections. Simply
when you have a influence condition like liquid lines features or an red arrow
saying "Same Image But Upside Down" to the right of the feature, the
combination of it predicting itself and the condition predicting the same thing
make it get a much higher score. Ex. my cat, and so I brought the.......?......
'brought the' already predicts cat or similar, so cat seen recently boosts
ontop this weight already existing. Predicting itself is really the only way
your going to get the same damn 3 birds right beneath it :) on zero shot.
------------------------------------------
Artificial General Intelligence List: AGI
Permalink:
https://agi.topicbox.com/groups/agi/Td410a9b65c4d95da-Me8635c0e0a50290cc56f2138
Delivery options: https://agi.topicbox.com/groups/agi/subscription