> I couldn't give it an image of a narwhal and an image of a horse and expect it to give me a unicorn.

You absolutely could with a generative adversarial network and the right training dataset. They can interpolate between images in a semantic way, and add and subtract components of the semantic representation. While it's not narwhals and horses, here's an example of using images of men with and without glasses along with a woman without glasses to generate a woman with glasses: https://raw.githubusercontent.com/Newmu/dcgan_code/master/im...

More results here: https://github.com/Newmu/dcgan_code And note that this field is advancing so rapidly that these results have already been improved on despite being only six months old.

To be fair, you haven't defined awareness in any meaningful way. Concepts are just coded by specific patterns of neural firings - there is no reason a neural net can't theoretically also code concepts meaningfully (of course, understanding that process and getting it work perfectly is decades away, IMO).

I don't buy the "understanding is hard". Also you'd have to characterize precisely what is so special about a unicorn. There are animals with 2, 3 .. things on their forehead. That doesn't make them special. What about pegasus. Or other invented beings of 50/50 blends (minautor, egyptian gods). We just have one particular model or animal qualities ingrained and 'enjoy' toying with a maximizing function of having both our traits and other animal traits. I'd love to see a cockroachman god figure.

I'd bet a thousands dollars that the underlying principles Deep Dreams uses to infer fractal and continuous images are the same we have in our brains.

> Meh, not really, the computer doesn't understand the images it's mixing,

That's 100% wrong. It does create abstract representations and it can imagine back. For example, there is this neural net that can do "woman + glasses = woman wearing glasses" on images, without any tags. It can combine concepts and generate new images. There is also another one that can "dream" images of bedrooms. It can even interpolate 10 intermediary bedrooms between any two images, and they look very good.