sawwit

There is a huge memory leak in Preview since El Capitan and the display of PDFs sometimes fails when you zoom in. I have no idea how many users are affected by this, but it seems strange that Apple hasn't fixed it yet. Perhaps it is just a tricky bug.

Wait, so Steve Jobs wasn't a visionary. He was just a salesman exploring unexploited markets as any salesman does? Albert Einstein wasn't a visionary. He was just a physicist that explored unconsidered theories as any physicist does?

It seems to me that in philosophy there is just as much "groupthink" as in almost any human endeavor—there is a maybe little less in the hard sciences where the systems give you feedback about whether an idea is correct or not.

Skim.app is very suitable for Latex PDF preview (it even supports syncing given you have an Editor that supports it too, e.g. Emacs AUCTex or TexShop).

Care to list those reasons. Genuinely interested, and not sure which reasons you are referring to.

> The difference isn't easy to describe, but one such difference would be that a single extra stone can change a Go position value much more than a single pixel changes an image classification.

A CNN can still distinguish extremely subtle differences of various animal breeds, exceeding human performance in such tasks. Why was that advance not a warning sign? The rotational-translational invariance prior of the convolutional neural network probably helps because, by default, local changes of the patterns can massively change the output value without the need to train that subtle change for all translations. Also, AlphaGo does a tree search all the way to the games end, which can probably easily detect such dramatic changes of single extra stones. Reality is likely much too unconstrained to to able to efficiently simulate such things.

Yes, the latter number was perhaps too conservative; no doubt about deeper predictions being easily possible, but I doubt even expert players consider many alternative paths in the search tree. They might recognize overall strategies which reach many moves into the future, but extensive consideration of what will happen in the upcoming moves is probably constrained to a only few steps; at least relative to the number and depths of paths that AlphaGo considers.

It's because they have a much larger stack size than a human brain (which does not have a stack at all, but just various kinds of short term memories). An expert Go player can realistically maybe consider 2-3 moves into the future and can have a rough idea about what will happen in the coming 10 moves, while this method does tree search all the way to the end of the game on multiple alternative paths for each move.

1. The value network is trained with gradient descent to minimize the difference between predicted outcome of a certain board position and the final outcome of the game. Actually they use the refined policy network for this training; but the original policy turns out to perform better during simulation (they conjecture it is because it contains more creative moves which are kind of averaged out in the refined one). I'm wondering why the value network can be better trained with the refined policy network.

2. They just run a certain number of simulations, i.e. they compute n different branches all the way to the end of the game with various heuristics.

Great achievement.

To summarize, I believe what they do is roughly this: First, they take a large collection of Go moves from expert players and learn a mapping from position to moves (a policy) using a convolutional neural network that simply takes the 19 x 19 board as input. Then they refine a copy of this mapping using reinforcement learning by letting the program play against other instances of the same program: For that they additionally train a mapping from the position to a probability of how how likely it will result in winning the game (the value of that state). With these two networks they navigate through state-space: First they produce a couple of learned expert moves given the current state of the board with the first neural network. Then they check the values of these moves and branch out over the best ones (among other heuristics). When some termination criterion is met, they pick the first move of the best branch and then it's the other player's turn.