It might be tacky, but it would work well-enough to just rephrase it as "How many non-programmers read HN?" and have a single voting option. We can assume that HN's uniqs/month, minus that number, are the number of programmers.
No, it would drastically underrepresent non-programmers. That only works if every single person that visits HN this month will see this poll and will take the trouble of voting (possibly having to create an account to do it) if they are a non-programmer. It also requires the uniques/month number to be accurate.
I'd guess that you would get an answer 2 orders of magnitude too small doing it that way, and that might be a generous estimate.
