Blade Runner - Autoencoded: 'Tears in the Rain' scene - side by side comparison

The film Blade Runner (1982) reconstructed frame by frame using a type of artificial neural network called an autoencoder. This is a side by side comparison of the 'Tears in the rain' scene.

After being trained on the whole film for 6 epochs, each frame was then reconstructed using the autoencoder and resequenced into a video.

The Blade Runner film, "Blade Runner" trademark, and many of the sounds, images and movie clips are Copyright © 1982, 1991 by the Blade Runner Partnership and/or The Ladd Company.

[rockomundo]

This looks like what it would look like to reconstruct a dream or memory from a human brain.

[MaxLohMusic]

What is the significance of this? When it all comes down to it, auto-encoding is just an advanced style of "compression" isn't it? And the fact it doesn't prioritize detail to the faces indicates (to me at least) that an AI "remembering" this movie would have quite some trouble knowing which character did what. Thoughts?

[InsolentHalo]

Could this be used to interpret the data coming from artificial eyes going to a human brain?

[geraldnapshireii]

"We present an autoencoder that leverages learned representations to better measure similarities in data space. By combining a variational autoencoder with a generative adversarial network we can use learned feature representations in the GAN discriminator as basis for the VAE reconstruction objective. Thereby, we replace element-wise errors with feature-wise errors to better capture the data distribution while offering invariance towards e.g. translation. We apply our method to images of faces and show that it outperforms VAEs with element-wise similarity measures in terms of visual fidelity. Moreover, we show that the method learns an embedding in which high-level abstract visual features (e.g. wearing glasses) can be modified using simple arithmetic." (Link in description)

[Bluebuthappy]

I'm guessing an animator could have a lot of fun with that effect.

[cwookdev]

Looks like he's wearing the one ring

[ericmichel]

It is interesting, but can you really call this learning when there is no real understanding? Seems more like just a creative version of programing and processing, perhaps I am missing the real significance of this video processing. How would this be used, or is it meant to be purely artistic?

[MCCanaryVideos]

what is a memory?

could this be qualified as a memory??

I think people are looking at this too literally and not the possible conceptual or even philosophical points this project and experiment brings up.

Does this not show an amount of understanding, by recounting the aspects of the film that it found important? By manually generating what it thought was important? Can that not be seen as a bias of some sort? Is that not what it means to have a memory, and in comparison with the ideas the film used, valuable only the one who retains the memory, and will be just as fragile as the memories held by every human known?

It is rather something to ponder, a true interesting forefront in a mentality that is beginning to manifest, regardless of what people think of it ethically or morally.

These are questions we should at least be asking, if we are going to be prepared for the rise of AI manifestation.