Hacker News Reader

Comment by bondarchuk | Hacker News Reader

>"(street background expanded)"

As in... expanded using generative AI? (The perspective on the lamps is really off unless they're different size lamps)

Yes, using AI as per the filename "backgroundexpandedusingai" [0], it seems to be an expanded version of this pic from Wikipedia [1] (but blurrier).

[0] https://8400e186.delivery.rocketcdn.me/articles/wp-content/u...

[1] https://en.wikipedia.org/wiki/Priscilla_Norman#/media/File:L...

Looks like the whole thing was run through AI, details on her face and scooter changed too.

And an oil drop on the street below the engine has disappeared.

I hate this. Why not show us the actual photo. Infuriating

I wonder how we'll deal with the inability to tell what's true or not in the coming years. Even without full deepfakes.. just a gradual hypothetical restoration turning subtle hallucination in many many places.

Judging by how little people care about the veracity of claims made on social media and youtube (long before AI), not much will change.

The root problem is that we don’t have very robust epistemic standards. We mostly go by vibes and what we want to be true.

the previous era of claims were not immensely large and not multimodal, now anything can be faked

Technically they do, just split into chunks interleaved with AI generated background. Original: https://en.wikipedia.org/wiki/Priscilla_Norman#/media/File:L...

>just split into chunks interleaved with AI

That would be relatively benign, the other possibility is that the whole thing was encoded and then decoded through some neural representation.

Reply on news.ycombinator.com