As long as the model itself understands that when it says "foo bar" in its own reasoning trace it means "pursue theory foo; no, this didn't lead anywhere, let's backtrack and pursue theory bar", where's the loss in clarity?
As long as the model itself understands that when it says "foo bar" in its own reasoning trace it means "pursue theory foo; no, this didn't lead anywhere, let's backtrack and pursue theory bar", where's the loss in clarity?