counter example: we won't know how to fight a robot apocalypse unless we go ahead and develop some super-intelligent kill bots ourselves, so we can train agains

counter example: we won't know how to fight a robot apocalypse unless we go ahead and develop some super-intelligent kill bots ourselves, so we can train against it and develop strategies. Given enough time and random variation, someone or some government is bound to create them, so we should not stick our heads in the sand or let fear-mongering stand in the way of research. In reality COVID-19 required several key mutations to be as harmful and to bind as effectively to human cells as it did. The likelihood of all of these mutations occurring naturally, in animals only, without intermediate variants to observe and develop immunity against, is very small. If only 64 amino acids must mutate exactly correctly to create the deadly variant then it would require 4^64 or 2^128 mutations, which is a large number. Assuming it mutates 100B times per year, that's still more years 3e27 years. For example, this experiment on more transmissible HIV requires hundreds of acids (600-1000, and a 32 amino prefix). https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2836558/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2836558/ The point being that creating extremely unlikely events to "see what would happen" is only justified if the risk of the observation is less than the natural risk of the event. If the risk of a world-ending pandemic is 1e-128 per year, and the risk due to human-made viruses is 0.1%, then "not sticking your head in the sand" has raised your risk by an astronomical factor.