## Comment: Re:How on earth (Score 1) 128

I can imagine a couple of applications of these transistors though...

Many numerical simulations require repeated random sampling of some process, and then combine the results in the end. If you're averaging *some billion simulations*, the result should be quite robust to fluctuations in the results of each simulation. Thus it might well be worth it to use 10 billion unreliable transistors instead of 1 billion reliable transistors, if they cost the same.

Another application could be to generate random numbers. Let's say that you have a pseudorandom number generator with periodicity N, and your unreliable transistors makes the algorithm do a random jump after an average of N/100 numbers. Wouldn't that be "random enough" for more applications than just the pseudorandom number generator itself?