Do you have working code? Please explain exactly how your AGI theory predicts.

BTW If you meant combining sound and image predictions, then yes, that is AGI 
too. Though the merging is associative, clap sound translates to clap video. 
This increases prediction accuracy. Not storing some (non-episodically) ex. 
just store sound of clap is merging, ex. image clap>crowd, sound clap>chants, 
we know chants=crowd and so we can just not store the vision version, store 
only the sound version, and inflate the specific weights that vision shared. Of 
course if asked on the task of vision, you need vision, so we can't just throw 
away other sensories, same for semantics ex. word2vec, we can remember both cat 
and dog. So what we can do is keep all the sensories, each having the merge of 
each other. So if sound a>b over c it is based on vision etc without needing to 
touch the visual cortex! I suppose same sensory ex. words can do this too, we 
keep both dog and cat sounds stored, and when we hear cat we get all dog 
predictions + cat's, not because the energy leaks through shared patheways but 
because of the representation 'cat' which is just really 'cat' but is an 
improved tree simply! Yes, shared paths between cat and dog were necessary to 
discover the relationship BUT costs more compute/time, it's a one time fix. IF 
they DO leak by shared paths, then well that's amazingly parallel.

I wonder, dog=cat is based on shared contexts in the net, but what if 
triggering them together affects it more? Like you tell your friend "yea bro, 
cat=dog, they both eat, lick, etc see! : friend: oh, now i see, thanks for 
showing me, even though i knew all those". Simply hearing them in time works 
better so we can strengthen them more. ? Hearing "cats lick, dogs lick", is 
more connecting them than "cats lick. My mom is old. Dogs lick.". Not because 
just cat and dog are closer, but because we are explicitly suggesting they are 
the same (so long as you trust the writer! He's telling you, clearly).

...have to think on both still... unsure still

About layers of sensories, ya, remember my Christmas hierarchy image? You could 
link sound clap to image dog, then link that to smell, then link that to some 
long sentence. Building bigger features.
------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/Td2287eb2b142cb0b-M1d5e21fb0b4dfcf58e4ce359
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to