[keghn]

All-memristive neuromorphic computing with level-tuned neurons:

[yotamarker]

kind of hard to understand over her snob accent but the algorithm seems to be :

trained to recog image A
trained to recognize image B
visualized image fusion
recog image fusion

she didn't say anything about how they filtered out the noise ?

[LOCKSUIT]

Fuse 50 images and eventually the image will be black or a mess of colors!

Every time I hear "trained to recognize a cat" I think "all it is is saving the image, searching memory, and selecting the memory that is like the one searching, no duh".

Strengthening both memories is straight-forward.

[yotamarker]

but its not 50 its 2 so

[LOCKSUIT]

but its not 50 its 2 so

Come over to mah house and we'll draw it out, TRUST ME.

[yotamarker]

draw what ? dude I really starting to think your a chatbot

[LOCKSUIT]

No 2 images are combined like that unless overlayered as crops (not actually blended together but covering), cus that makes it look stupid, that is nooooot incorporated in us trust me. The only way your'll see such is by drawing it~ If external and internal input are both seen then it's not helpful really, unless you want to imagine tape or dust overtop say your table speakers hmm...but can I? No. Yes? Hmm....rrr......RRR.................Omg I can do it with 2 internal memories too....
..................................Ok I accept, adding this to my transcript

There, that was easy:
They can be combined, to acquire even more. For example, remembering two memories, such as a music video, and a song, and attention goes to only the most ranked or bright or energized vision and song when only both return from searching memory. Another example, is remembering a clip of a bunny jumping, or rotating chair, which gets cropped out by object visual differences, and overlayered on top your external, or internal memory vision of a desk or walking around a room, which will automatically play new sets into storage for a bit depending on its energy, which is overlayered your external vision, which the external or internal underlayer is seen more if less opaque or more weakly matched ex. tape over a phone or memory, and then the creation searches memory.

[yotamarker]

I think I figured out how the dots were filtered :
they are stand alone, they are not connected to the near pixel
she probably didn't say how it was done because they used a ready made filter