After all, we could use robustness as our optimization criterion.
> After all, we could use robustness as our optimization criterion.
It's almost impossible to optimize for "robustness". You are always optimizing some measure. You would have to deoptimize for all measures, and that's just silly from an engineering perspective.
The only thing you can do is to simply not optimize to the last 0.1%.
The whole toilet paper fiasco is a good example. Toilet paper is such a low-margin business that everybody optimized it to hell and back. Lines are already running 24/7 even in normal times. The consumer and industrial production is split so they can optimize them independently to squeeze those last couple fractional percentage points out by making commercial toilet paper even shittier.
Consequently, there is zero ability to absorb the demand shock that occurred when lots of commercial usage transferred to consumer usage.
(Eg, it might be possible to make to make space access more robust by spending the next 100 years of the world's combined effort to build a space elevator. But that's not going to happen because there's no perceived need to have that extra robustness given how much human focus is required.)
Forward error correction systems are designed to improve signal robustness in the face of noise - based on a given noise model. It can be made more robust by assuming a much higher noise than what actually exists.
This would probably require more transmission time. But the robustness level, by any usual quality measure, would increase.
"The whole toilet paper fiasco is a good example" ... for my clarified version, yes. That system was was optimized for profit extraction, probably also with the expectation that the government would throw money at them if things went really bad. ("Socialized risk.")
It was not at all optimized for an efficient transition to a widespread shelter-at-home world.
It could have been at least somewhat optimized for that, which is why I think your original statement was too broad. But it wasn't, because it was maximizing for profit, not social stability or happiness or other more nebulous criteria.
