Exactly ! this makes the problem potentially millions of times easier, since you know with certainty that every ray fired is going to contribute to the image, whereas firing rays from the light source would guarantee you never see most of them, the processing power is wasted and your image never converges