Comment by AutumnCurtain | Hacker News Reader

Comment by AutumnCurtain | Hacker News Reader

>AlphaZero’s neural network evaluation function doesn’t have the same level of structure as Stockfish’s evaluation function: the Stockfish function breaks down a position into a range of concepts (for example king safety, mobility, and material) and combines these concepts to reach an overall evaluation of the position. AlphaZero, on the other hand, outputs a value function ranging from -1 (defeat is certain) to +1 (victory is guaranteed) with no explicitly-stated intermediate steps. Although the neural network evaluation function is computing something, it’s not clear what.

I find it fascinating that we can create something that can "learn" in some sense in a way we don't fully understand. Or am I misunderstanding the meaning of this passage?

The way in which I think about it is that "learning" (in humans) often occurs is in the form of heuristics - certain rules or shortcuts which are not guaranteed to be optimal, but do a good job at simplifying (the search-state of) the problem. Instead of trying many simple comparisons between legal moves, we instead use a heuristic to score them.

Non-NN chess engines are really all about doing a good job at balancing these heuristics (like it says above). There's fundamentally an upper bound in how good the overall approach can be, simply because the underlying heuristics are defined and selected by humans: there could be many more heuristics that haven't been considered. This version of "learning" is really just removing human forced constraints which were introduced to help learning, by starting from a position without bias.

Thanks so much for the response! It's really interesting stuff.

We do understand how they learn, but the structure of their knowledge is currently hard to discern in the same way that you can't get thoughts or memories out of somebody's head with a MRI scan.

It's interesting to think that if we create a human-level artificial intelligence using ML, we'll probably have no idea how it works.

It's okay, it won't either

To be fair, we have no idea how we think either!

Reply on news.ycombinator.com