Reality, imitation and reasoning with generated artefacts

A short post this time but I have been reminded me of something. A few months ago, news came through of the recreation of John F Kennedy’s lost speech (BBC news link) using voice synthesis. Small sections of voice with different pronunciations are algorithmically stitched together to present the whole voice as if it was real.

That was my original concern. The recreation of the voice, useful for representing the possible, needs a for of critical thinking. The digital has created an imitated object but, potentially, hidden its artifice.

I was reminded of this in a recent TechCrunch article about the use of Generative Adversarial Networks (GAN), This tech (scarily) lets video change reality. I see the use in films and updating an image but it has other uses, such as deepfakes. Though the posting of deepfakes is banned on some sites. As the Forbes link argues, some of this is through curiosity rather than malice but that doesn’t preclude malicious intent. Moving from pornography and the recreation, could a digital Hitler Diaries come about and how might it be discovered?

The questions at the back of my mind is how this technique can be thought about from the cultural or Humanities perspective. How does one begin to critically think about these things or discover them? Does this mean that Machine Learning might need to be added to a critical toolkit? What does it even mean to think about these to find them?