@AR0w posted this some years ago, but it's probably time to have a look at it again.
If you just try it with your own mind, you will lose.
You can try it with a coin and you might do better, or you can try it with dice (1-3 = 0, 4-6 = 1). The idea is that you make choices about whether to input a zero or a 1 and the computer tries to guess what you will say based on your previous inputs.
the traditional way to "fool" these is to use the binary encoding of a transcendental number, like Pi or Napier's Constant [aka "e"].
theoretically those aren't very high-entropy either, if you express them by the formula used to calculate their digits, although the complete lack of structure makes them evade the typical prediction algorithms.
that's cool. I hadn't thought of that!