Hacker News Reader

Comment by rvz | Hacker News Reader

This is exactly why we need detectors for this sort of technology. Like with DeepFake detectors there should also be detectors for DALL-E 2, Stability AI, etc.

The whole technology (especially AI technology, Deep Learning) is a dystopian creation and nothing good comes out of it.

As for this art contest, I would just disqualify digital art all together and only allow hard copy, physical oil on canvas paintings. (And no, not a printed copy)

View on HN

Not going to happen. Already advanced AI authors combine many tools. They may use their own image/photos as input (which is non-AI), combine with AI, compose objects/layers, fine-tune, inpaint, outpaint, text inversion and that's about a single week of progress.

You can't stop this even if you tried.

> You can't stop this even if you tried.

Who said stop? I said 'detect'.

DeepFakes have detectors, there are GPT-3 detectors already being developed and the same will happen with DALL-E 2 / Stable Diffusion. I won't be surprised to see someone develop such a detector for both of these AI tools as a research project.

It depends on what you mean, detection of specifically a deep fake, or the detection of AI involvement in image creation in general?

Detection of a deepfake, for example a generated photo of a person of something that never actually happened, may be in the real of possibilities, but still difficult.

Detection of AI in an artistic image? Not a chance.

Reply on news.ycombinator.com