However, it turns out this algorithm might not choose left — because if it always chooses right, it might not have well calibrated probabilities about what would happen if it chose left instead.
The ambiguity between choosing right (as in well*) and choosing right (the direction) is slightly confusing here.
*Which would be the direction left.
It also seems like a proving program would be able to get around needing to explore if it had access to the source code for a scenario.
The ambiguity between choosing right (as in well*) and choosing right (the direction) is slightly confusing here.
*Which would be the direction left.
It also seems like a proving program would be able to get around needing to explore if it had access to the source code for a scenario.