Is it possible to input a correct image in to a neural networks layer and compare it with distorted image and make a model ? This is artificial intelligence . Does it work for large deep sky images - like andromeda - comes from distorted mirror ?
And why spin casting is unsuccessful ? I have an new idea. If I am not wrong , liquid mirrors are sensitive to dirt and wind and operators coat mercury with epoxy. I mean they pour epoxy on liquid mirror . If this epoxy cure without self distortion and without distorting the liquid mirror under it , may be this method of mirror making is efficient ? Am I right ?
Do you know anybody who made spin casting successfully ?
And if spin casting is not working , why people talking about it and Scientific American publish about it ?
And which epoxy is good ? slow curing or fast curing ? Is there very low shrinking polymers out there ?

• Correction - bobmay - July 24, 2004 - 21:02 UTC