[smileysteve]

A post about mis-generalization seems to make a huge mis-generalization, seemingly from a lack of diversity.

The author recognizes that laundromats exist, but doesn't believe that washing machines are a luxury good (his yacht example of 60% owned by the top 1% likely applies somewhere around 60% to the top 30%). This indicates the author generalizes that high density living with shared appliances don't exist or apply to a large part of the population (that density doesn't shift statistics).

A few thoughts, instead of taking 340m people, you need to account for 2% homelessness, another 8% as housing insecure, 30%+ as high density possibly sharing at the apartment/condo building or laundromat level. A small upper 1% that outsource to services.

The infamous Mitt Romney, 47% of Americans don't pay income tax quote can be really shocking when you start thinking about the average American, and the wages that they earn compared to a higher income segment you might be in.

[bluGill]

Laundromats exist, but there are 100 million according to my web search which puts his estimate in the right area even if he is off by 30 million.

Most Americans would not call a washing machine a luxury good. Even in much poorer Mexico there are a lot of families living on $100/week that have a washing machine - they are an affordably luxury to many poor people.

[smileysteve]

If the statistical "right area" is within 30%, then 60m is similarly in the correct area.

But the context is most important, are AI's worst strategies worse than 30% off because of gross generalizations?

[bluGill]

Depends on the context. Sometimes I need to be correct to withing 0.0000001%. Sometimes I can be off by 500% and still be close enough.

[randallsquared]

> 2% homelessness, another 8% as housing insecure

Per a quick search, that's 0.2%, not 2%. Not sure about what "housing insecure" means, so that's harder to check. Also, that is just a one-night snapshot, and many of those won't be homeless if you check back later.