type shit

I've been alive for ~14 million minutes. In none of those minutes has a banana spontaneously appeared in front of my face and started break dancing. By Laplaca's law of induction I can justifiably be 99.99999% sure this will not occur in the next minute. Except there's an easy disproof of this method of reasoning: You can make your credence arbitrarily high by choosing different partitions. I've been alive for 10^51 Planck times, but clearly 1 - 1/10^51 would be wildly overconfident. Where does the term for "model error" start becoming relevant? What's the chance that I've forgotten a break dancing banana, or that the laws of physics may just decide to change at any moment? Not sure. But the "argument from practicality" shows that it has to be somewhere lower than around 1 in ten million, since we make decisions based on probabilities in that range all the time. (e.g. wearing a seatbelt when you drive.) Same goes for the lottery examples some people are bringing up in the replies. If I program my computer to output a random number from 1 to 10^500, what chance should I ascribe to it being any given number? I think it's much greater than 1/10^500, because my brain can't even store that much information. The moment I consider a *specific* number and try to assign a credence to it, I have to take into account the fact that there's some reason I'm thinking of that number instead of any of the others, and that reason could be related to the program's output.