Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

For practical reasons mostly yes.

For scientifically correct scaling (think forensics) you will of course need to interpolate the source data instead of generating synthetic data which is done by AI upscaling.



Not entirely sure the AI scaling is more or less "correct" than the "interpolate the source data" method.

Both use priors, the former a large and complex one, the latter a simplistic one, but both "invent data based on assumptions".


You can completely understand and intuit what interpolation does, and it doesn't drag in extra detailing from anywhere.


That's correct but probably also practically irrelevant.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: