Have you tried this? Does it work?

I don't understand much of this, and I want to, so let me start by asking basic questions in a much simpler setting.

We are playing Conway's game of life with some given initial state. An disciple AI is given a 5 by 5 region of the board and allowed to manipulate its entries arbitrarily - information leaves that region according to the usual rules for the game.

The master AI decides on some algorithm for the disciple AI to execute. Then it runs the simulation with and without the disciple AI. The results can be compared directly - by, for example, counting the number of squares where the two futures differ. This can be a measure of the "impact" of the AI.

What complexities am I missing? Is it mainly that Conway's game of life is deterministic and we are designing an AI for a stochastic world?

The first requirement:

Even as chicks, geese cannot be handled by a human, or encounter other geese who have been.

suggests that a FAI would not tell us that it exists. In other words, the singularity may already have happened.

You can pay someone $8/hr to do menial tasks 20 hrs/ week, for a total of about $8000 / year.

With payroll taxes and insurance, I would expect this to cost at least $12000 a year.

Sure, but you could have a limit on how many rounds back they remember, or you could fill in the history with some rule.

Or just prohibit the bots from knowing which round they are playing.

It is not at all clear that the people resistant to addictive drugs are reproducing at a higher rate than those who aren't.

I think the sub-proposal is too complex and involves too many trivial inconveniences. I up-voted the original proposal.

almost every researcher in CS flaunts copyright, posting their papers on their own websites

Many journals explicitly allow you to distribute a "preprint" of your journal articles on your personal website. For example, the Elsevier policy states that authors retain:

the right to post a pre-print version of the journal article on Internet websites including electronic pre-print servers, and to retain indefinitely such version on such servers or sites for scholarly purposes

There are open access journals. I recommend supporting them.