I have an entirely different critique of the idea, based on the breadth of the experiment. I think there is a good case that the more people that have the basic income, the better chance there will be of network effects arising.

For example, Let's say there are 10 or so people who are given the basic income. You are one of them and you are looking for a cofounder for a drone startup, or some other project. You need to gel with one of those other 9 people. Because everyone else is busy earning their monthly nut. So if you want to test the idea and really allow for the chance for those kind of network effects to arise, you need a massive sample size of recipients. That means for practical reasons, you need to do it in a place where people live on low incomes. That's exactly what GiveDirectly is doing, and it makes a lot of sense. In Oakland, it could cost ~$50M annually for every 1000 people in the program, not including overhead costs. That's not a lot of data to extract meaningful data from.

Realistically no Basic Income could be assumed to be for life. After all, governments and policies can change. So rational actors should always assume it is just for a limited time, and have a contingency plan.