It's true: I don't know how to predict job creation. Some new kinds of jobs are obviously enabled or required by new technology. But many (most?) changes are either socially or politically driven and so very hard to predict.

So it might be that enough jobs for everyone will be created for non-technological processes. But I still think it likely that job automation eventually won't leave enough room for this "jobs for everyone" scenario.

Jobs can resist automation in many several ways:

• Require domain-specific intelligence. E.g., programming.
• Require human interaction skills where the hard part is understanding the customer's problem or explaining to them what they must do. E.g., front desks, clerks.
• Require physical dexterity and mobility. E.g., a waiter serving tables.
• Legally required to employ humans, for reasons of liability (e.g. doctors), tradition (e.g. judges), or public sector make-work schemes.
• Jobs where human customers will pay a premium for interacting with other humans.
• A long tail of highly specialized jobs with few workers each, where the upfront cost of automation isn't worth the resulting savings. This can also be cast in terms of automated systems having insufficient flexibility.
• Probably many other reasons unique to various sectors.

Clearly this leaves a lot of room for a future society where jobs cannot be automated away.