We're talking about the global spam problem, not your localised end user spam problem. Even if gmails spam filter was perfect and stopped all spam (impossible), they'd still have a problem because they'd be using vast amounts of human and computing resources to do it.

Maybe at first, but I think the GP's idea is more generally to make sending spam not a viable way of making money. If end users never see spam and spam profits drop from 0.2% of email addresses or whatever they're at now to 0.000000000002% then other methods to make money will be used.

You simply can not get that sort of accuracy without vastly increasing the number of false positives.

If it was perfect, spammers would no longer have an incentive.

I already specified that perfection is impossible. Even if every email is routed via a human filter instead of a computer filter, there would still be mistakes.

If a spammer can't get a message through, they will keep tweaking it until they succeed.

If even 999 out of 1000 spams are blocked, they'll still keep firing.

The perfect is the enemy of the practical.