I think some napkin math with the number of celebrities and professional sports players, who presumably have far better access to testing than the US public (which has essentially zero access), would show that we are at least at 150k-200k infections, if not more. At least, immediately before Tom Hanks announced his infection, I saw some napkin math that showed that since no politicians, movie stars, or professional athletes were positive, the rate was less than 1/2000 in the US. Now that we have multiple infections in that group, we are in the > 1/2000 == 150k infection regime.

Does this control for the fact that celebrities and politicians do way more travel and meet way more people?

Nope! Napkin math will treat all cows as spherical. However, if they are traveling and meeting way more people, it's also likely they are spreading it far more in addition to having it at a higher rate than the more socially distant segment of the population. Also, would likely indicate vast amounts of community spread that is going completely undetected.

Canada has a population of 38 million, and only 150 confirmed cases.

But one of those cases is the Prime Minister's wife.

Do we really think there are only 150 cases?

It's obviously way, way, way higher than that.

Mme. Grégoire Trudeau seems to have contracted the novel coronavirus in the UK, so your datapoint doesn't really suggest anything about the situation in Canada.

Many people travel between UK and Canada.

>This is the interesting case where anecdote really does become data

I still disagree with this, although I somewhat agree with your sentiment. The anecdotes I have seen range from extremely mild to nightmarish. A large fraction of those made me false, deliberately or by accident. I find the statistics still tell the most concrete story.