Undoubtedly there are edge cases where that's true. But as an argument for UBI it works only if the cost savings from labourers being willing to undercut market price of robots in a UBI-era world across the whole economy exceeds the net cost of making welfare payments to a vast number of other people previously ineligible for them.

At present, the numbers of people claiming welfare because their skillset has been superseded by automation are vastly exceeded by the number of people on federal minimum wage jobs or choosing not to be employed...

And that's even before you look at the distributional impact of the subsidy (the benefits from cheaper labour tend to accrue to a subset of business owners; the cost of subsidising it is shared across all taxpayers, with the net burden probably falling on middle class wage earners)

If a government were to decide that preserving jobs in a particular automation-threatened industry/locale were worthwhile due to the difficulty of finding alternative occupations, a targeted subsidy per job would be far more efficient