First, are we assuming that the only measure of success is how much money we get? Some people need $1,000 more than others. If someone has taken my child hostage and I know that the child will be released if I can find $1,000, Choice 1 becomes the "no-brainer." Have the results in experiments been broken down along socioeconomic lines?

If money is the only object, and the odds of the AGI being right are greater than one in one thousand, my mathematical expectancy is greater for Choice 2 than choice 1. As the alternative to zero in Box B drops, the AGI's accuracy must rise. What if there were either 0 or $1,001 in Box B? Has anyone graphed the outocome with different max pay-offs?

We don't know how good a guesser the AGI is. Does our assessment of the AGI's success depend on our life's experiences with people smarter than we are? On the language used to describe the AGI? That assessment seems key to the probability approach.

That, and how badly we need a thousand bucks.