Hacker News Reader

Comment by temp8964 | Hacker News Reader

This can be really useful in criminal investigation.

This is generating a fantasy face. To use this kind of AI-generated content as if it was factual would be terrifying.

Same with human memory recall, and that is used in trials as evidence

There's a wide gap between showing a police portrait, which is clearly drawn by hand, vs. what looks like (but is not) a crystal-clear photograph.

Not much. What it's doing was quite well summed up by jeffparsons in a different comment thread - it's "fabricating detail based on a training set".

This system makes an image that looks better than a low res image, but it doesn't necessarily make it look more like the original image.

How? We don't trust machine learning to match faces, why would we trust it to recreate faces? Genuine question.

It’s probably a tool that could be used for generating a likely appearance for helping to find a person to interrogate, similar to eye witness sketches. I would hope it’s never allowed at trial though.

Prosecutors would need to be able to defend every element they use to gather evidence in court. Nothing downstream from inadmissible evidence can be admissible in court. “Fruit of the poisonous tree.”

No. Actually, that’s not true. If you upscaled a picture of a drug deal and then used that to narrow your search set and then posted a watch on that set to catch them dealing drugs, you’ve got a useful case. Totally admissible.

This tech is probably less error prone than the mentioned eye witness sketches

It is almost certainly as bias trained. Such that if it is not stupid iterative with the witness, it is probably less likely to converge.

Only a fool would believe these reconstructions were real. Unfortunately fools are more common than we would hope.

It is genuinely alarming to me how many people in this thread are saying that this will be a boon for crime investigation, medical imaging or the like.

To spell it out: This is not re-creating what was actually present in the original image. That is, and will always be, impossible due to fundamental limits of information in the source seed. What it is doing is using AI Hallucinations to create a believable-looking fake.

For identifying persons of interest from bad footage perhaps. Not so much for positive identification.

Need to be careful what data the AI model is trained on. i.e. it could bias the features of a particular race, causing those people to be held as suspects more often.

I could not help but be reminded of the "enhance it!" trope common in crime shows: https://www.youtube.com/watch?v=Vxq9yj2pVWk

:waits patiently for all the people that said this was impossible to recant:

Oh but it IS impossible, that hasn't changed. We're just getting better at generating believable fakes.

i'll let the jury be the judge of this

Reply on news.ycombinator.com