I decided to split this out as a separate post...
Regarding ISO (supposed speed):
Any optical sensor system works by taking a sample of the sensor output. That occurs in an instant. The sensitivity of the system is what determines the "highest ISO" that can be achieved.
Lower ISO settings are done by summing the results of those instantaneous samples, with each "stop" being twice or half as many samples as its neighbors.
Noise (as being discussed here) is generally random. It won't appear at exactly the same location on multiple samples.
Thinking in terms of a single pixel on the sensor, when you add samples, the majority of them will not have noise at that pixel. If you sum a large number of samples, only one of which exhibits noise at that location, that noise averages out to nothing.
This is the ONLY reason that low-ISO exposures have less noise than high-ISO images.