But if there was a bot that was well behaved (for whatever that means for a given service), and somehow legitimately viewed ads, and could legitimately decide to buy something with real money from an ad, there's a lot of sites that would no longer want to ban that particular bot. Now they need to distinguish between that bot and the other "bad" bots. In general, if a bot can "behave well" we don't necessarily want to kick them out just for being bots.
In the end, the more bots become like humans and humans through things like the Mechanical Turk become more like bots, service providers will be required to very carefully think through what it is they actually want to demand of their users. Splitting people into "human" and "bot" is already only an approximation on the grounds that there's plenty of humans services don't want around, and there's already some bots that services are fine with (e.g., some helpful reddit bots), so that split isn't going to be good enough. The process of thinking through what is really desirable at a much higher level of detail will be fun to watch, and there will be a lot of different answers for different services.
The mistake he made is adding blockchain and crypto to that physical verification.