Entropy, Power, and Models

Power is derived from harnessing entropy.

Power is derived from harnessing entropy.

The most powerful technologies (letters, numbers, printing press, computers, internet, science) harness the power of replication.

While global entropy increases, all technologies are simply local entropy sinks. These are places where you can implement a new framework to wash out unpredictability. The creation of local entropy sinks is the key to value, to power, to knowledge, to control. The reduction of randomness is critical to our progress as humans and allows us to make reliable changes on an abstracted layer above an unusable layer.

“All models are wrong some are useful”

New technology does not have to map to perfectly to reality. New technology simply implements a slightly more accurate framework, or a framework with a more favorable distribution of outcomes. Unfortunately, many times we mistake the most accurate map, or the most reliable map, for the final map. It would be wise for us to remember how far away from the final map we are in nearly every domain.

Applied Problem: Survivorship Bias

“Napoleon is reputed to have had a profound insight into the human soul; Shakespeare also. And their vision has nothing in common.” -Shestov

Most successful entrepreneurs swear by fairly different heuristics and give different advice. Less successful people say that you should ignore all their advice due to an idea they made up called “Survivorship bias”. While the term is real, it has been consistently misapplied. In business, you can now read that term as “I don’t know why they succeeded, but rest assured, I’m certain their advice is unhelpful”. 

Survivorship bias can safely be ignored most of the time. All the successful people ignore it.

In actuality, we’re so far removed from the final map that even tiny reductions in entropy seem to produce outsized results. Those who refuse to use models never succeed because the underlying layer is far too complex to be usable. You need to operate on an abstracted layer. Is inverting every problem the key to it’s solution? Unlikely, but it reduces the entropy enough to focus the mind onto the highest impact elements. The consistency of this process allows the odds to increase significantly enough to achieve desired outcomes reliably. 

Collect Models, Harness Entropy

This is why a wide collection of mental models is so valuable. It consistently minimizes the variables that could go wrong and has been stress tested by many others. Your multi-model testing is stress testing the next abstract layer. While the situations may change, the models may change, you can choose chaos or simplicity. Only one produces reliably.