4 ms·

This gives me the same feeling as those ML-powered “enhanced zoom” features: where does the photograph end and the machine made-up fantasy start?

gsliepen 5y ago
Obligatory Red Dwarf reference: https://www.youtube.com/watch?v=2aINa6tg3fo https://www.youtube.com/watch?v=2aINa6tg3fo
- 14 5y ago
  Take my upvote. This has been posted to HN before but I love seeing it again. Red Dwarf was truly an amazing show way ahead of its time imo. For those who have never heard of it, it is a space based comedy I would recommend to all.
  - chris_wot 5y ago
    Definitely the best spoof of Blade Runner I've seen so far.
- sergiotapia 5y ago
  https://www.youtube.com/watch?v=T5tIDad3xxs https://www.youtube.com/watch?v=T5tIDad3xxs I love these guys so much
dathinab 5y ago
"ML-powered" enhanced zoom can't be trusted. The moment you need ML magic to magnify you are basically telling the ML "guess what is there based on what we can see and the training database you where trained on". Also due to the nature of such algorithm they do not have any form of common sense or "realistic general purpose" world view, and tend to be black box so it's hard to check if they learned something "stupid"/"ridiculous". So it's totally possible that somehow it learned to place knifes into the hyper-magnified hands if they have a specific skin color and a turquoise sleeve on a picture with a color scheme as if it's a cloudy day. Or similar arbitrary things. Sure it's supper unlikely but not impossible and there is not good way to find such "bugs". Worse multiple ML systems trained on the same data might share some mistakes and there are just that many huge comprehensive datasets.
- andbberger 5y ago
  OK sure but the distribution of natural images is highly redundant ie pixels are not independent and there is structure to exploit. It's not magic obviously but done properly it's a pretty reasonable thing
  - MauranKilom 5y ago
    > OK sure but the distribution of natural images is highly redundant ie pixels are not independent and there is structure to exploit. Sure, but that depends very strongly on the number of natural images you trained on and the specificity of the situation you present to the algorithm. Maybe one such structure in the training set just happens to be people holding knives given "a specific skin color and a turquoise sleeve on a picture with a color scheme as if it's a cloudy day".
    - andbberger 5y ago
      I said if done properly. My point is that in principle using a natural image prior to 'enhance' an image is not a totally crazy thing.
      - dathinab 5y ago
        It isn't (EDIT: isn't crazy) . But there is no good way to reliably know if it's done well for this specific image. So it can't be trusted. Which doesn't mean it's worthless. It's kinda like someone saying "I think I maybe have seen <something> when passing by but I didn't pay attention." It's not trustable as the human mind if fickle but it can be a good starting point.
  - Dah00n 5y ago
    Sure, if it is an option applied after the fact by the user editing the photo. Any camera that does this automatically is utterly broken. Otherwise where is the line? Some sharpening? Adding bigger boobs by default? Make people less black?
nitrogen 5y ago
AI enhancement of imagery injecting made-up data was a minor plot point in the novel Congo (the one about diamonds and apes), if I recall correctly, so this has been a known risk for a long time, and ML educators and practitioners aren't doing a good enough job of managing and talking about that risk.

gsliepen 5y ago

Obligatory Red Dwarf reference: https://www.youtube.com/watch?v=2aINa6tg3fo https://www.youtube.com/watch?v=2aINa6tg3fo

14 5y ago
Take my upvote. This has been posted to HN before but I love seeing it again. Red Dwarf was truly an amazing show way ahead of its time imo. For those who have never heard of it, it is a space based comedy I would recommend to all.
- chris_wot 5y ago
  Definitely the best spoof of Blade Runner I've seen so far.
sergiotapia 5y ago
https://www.youtube.com/watch?v=T5tIDad3xxs https://www.youtube.com/watch?v=T5tIDad3xxs I love these guys so much

dathinab 5y ago

"ML-powered" enhanced zoom can't be trusted. The moment you need ML magic to magnify you are basically telling the ML "guess what is there based on what we can see and the training database you where trained on". Also due to the nature of such algorithm they do not have any form of common sense or "realistic general purpose" world view, and tend to be black box so it's hard to check if they learned something "stupid"/"ridiculous". So it's totally possible that somehow it learned to place knifes into the hyper-magnified hands if they have a specific skin color and a turquoise sleeve on a picture with a color scheme as if it's a cloudy day. Or similar arbitrary things. Sure it's supper unlikely but not impossible and there is not good way to find such "bugs". Worse multiple ML systems trained on the same data might share some mistakes and there are just that many huge comprehensive datasets.

andbberger 5y ago
OK sure but the distribution of natural images is highly redundant ie pixels are not independent and there is structure to exploit. It's not magic obviously but done properly it's a pretty reasonable thing
- MauranKilom 5y ago
  > OK sure but the distribution of natural images is highly redundant ie pixels are not independent and there is structure to exploit. Sure, but that depends very strongly on the number of natural images you trained on and the specificity of the situation you present to the algorithm. Maybe one such structure in the training set just happens to be people holding knives given "a specific skin color and a turquoise sleeve on a picture with a color scheme as if it's a cloudy day".
  - andbberger 5y ago
    I said if done properly. My point is that in principle using a natural image prior to 'enhance' an image is not a totally crazy thing.
    - dathinab 5y ago
      It isn't (EDIT: isn't crazy) . But there is no good way to reliably know if it's done well for this specific image. So it can't be trusted. Which doesn't mean it's worthless. It's kinda like someone saying "I think I maybe have seen <something> when passing by but I didn't pay attention." It's not trustable as the human mind if fickle but it can be a good starting point.
- Dah00n 5y ago
  Sure, if it is an option applied after the fact by the user editing the photo. Any camera that does this automatically is utterly broken. Otherwise where is the line? Some sharpening? Adding bigger boobs by default? Make people less black?

nitrogen 5y ago

AI enhancement of imagery injecting made-up data was a minor plot point in the novel Congo (the one about diamonds and apes), if I recall correctly, so this has been a known risk for a long time, and ML educators and practitioners aren't doing a good enough job of managing and talking about that risk.