Akira Urushibata <a...@wta.att.ne.jp> writes: > Mr. Kaz Kylheku claims that artifacts that never existed in reality > may show up when images are magnified. If anybody believes that this > is true, please show me examples. I believe they would be of interest > to follow list members.
Recent demonstrations of AI "upscaling" old low-resolution photos or analog video to "4K quality" are an excellent example of this. They fill in details where none exist, based on copying them from other similar data where it does. > In a previous post I provided links to two newspaper articles, The Rittenhouse trial was televised. Did you watch it yourself, or are you relying on third-party, likely biased, sources? (Note: I have not watched it myself, and base my comments below on this analysis video: https://www.youtube.com/watch?v=NxoYNpBMaCg ) > Both considered the judge's decision declaring the image inadmissible > bizarre. The judge did not declare the image inadmissable at all. He declared that Apple's software could not be used to zoom the image unless the prosecution provided expert testimony that the admissibility of the image would not be changed by such zooming. An expert was not available in the time allotted, so an alternate software that was previously vetted was used instead. > We already have software that add color to old black-and-white > photographs. This is an excellent example of adding details that aren't in the original image - if the image was colorized, and the color was used as evidence in the trial, I would hope the judge would declare it inadmissible. > Judges must be careful when deciding whether something can be admitted > as evidence or not. Where judges are elected, voters should consider > whether a candidate understands technology, at least the kind which > is now commonplace. I think in this case, the judge *did* understand that our *current* technology MIGHT be able to use "AI" to add details to zoomed images, and thus the prosecution had to do due diligence to prove that such was not the case. It was not the image which was a concern, it was software which might modify the image in such a way which might taint the evidence. Whether you believe that Apple software does this or not is irrelevent; it still must be proven in court. > As programmers we must keep in mind how much difference an accurate > record can make. We also must be willing to be informative to > consumers who need advice on the matter. Agreed 100%