Comment What makes this (Score 2) 18
I've not had much luck in getting a better understanding on the algorithm used to generate these black hole images -- it sure sounds sounds like it's like similar to that trope in TV shows where a cop says "enhance!" at a blurry picture, and the original noisy image magically becomes much higher resolution.
Isn't this just feeding a handful of data points into a machine learning algorithm (can we please stop calling this stuff "AI" already) and generating an image from that, not entirely dissimilar from DALL-E or Midjourney or etc etc? If that's the case, how can we be certain the end result actually represents what is really there, and that the output image is actually scientifically useful? Or, how certain are we that the output from this algorithm actually reflects reality, and isn't just spitting back out what we think we want to see?
Honestly I trust that these researchers know what they are doing, but the aspect of what exactly the algorithm is doing and why we think it is trustworthy doesn't seem to get much mention.
Isn't this just feeding a handful of data points into a machine learning algorithm (can we please stop calling this stuff "AI" already) and generating an image from that, not entirely dissimilar from DALL-E or Midjourney or etc etc? If that's the case, how can we be certain the end result actually represents what is really there, and that the output image is actually scientifically useful? Or, how certain are we that the output from this algorithm actually reflects reality, and isn't just spitting back out what we think we want to see?
Honestly I trust that these researchers know what they are doing, but the aspect of what exactly the algorithm is doing and why we think it is trustworthy doesn't seem to get much mention.