In Marcus Hutter's list of open problems relating to AIXI at hutter1.net/ai/aixiopen.pdf (this is not a link because markdown is behaving strangely), problems 4g and 5i ask what Solomonoff induction and AIXI would do when their environment contains random noise and whether they could still make correct predictions/decisions.

What is this asking that isn't already known? Why doesn't the theorem on the bottom of page 24 of this AIXI paper constitute a solution?