I think his argument is that without type inference, programmers will avoid more complex types, and therefor model their domain with less precise types.
This makes no sense.
The basic idea is that type inference would shift the sweet spot for the right amount of typing upwards.