>>1382832 (previous)
Chess computers just took a giant step in the opposite direction.
Old style (Deep Blue, Rybka, etc.) There's a move tree, you can see what the computer picked and exactly why, what it evaluated, what was almost as good, etc.
New style (Alpha Zero): you have no idea what's going on inside that neural network, it just pops out one tremendously strong move. You have no clue what if any plan or prediction is there, if any other choice almost (in some ill-defined sense) was picked … but it's somehow a better move than the old computers picked.