I have a cool little challenge at the bottom of this post, doesn't require any knowledge on Brain-Flak just some maths.

thought it might be of interest.

@WheatWizard doesn't that just simplify to sum(i**2 for i in range(n))

using cs.cornell.edu/courses/cs211/2005sp/Sections/S2/Induction.pdf

or this

The goal is to show that yes.

@WheatWizard can't we just copy a proof from the question then add a few lines >_>

Not quite

there's a trick

huh?

Hmm

the first part sums to largest square, the rest sums to all the other squares?

Ah yes it does

I had a different way of doing it geometrically

@El'endiaStarman In case there is something I can help you with let me know!

@flawr I definitely will.

Holy crap. gizmodo.com/…

> Remember AlphaGo, the first artificial intelligence to defeat a grandmaster at Go? Well, the program just got a major upgrade, and it can now teach itself how to dominate the game without any human intervention. But get this: In a tournament that pitted AI against AI, this juiced-up version, called AlphaGo Zero, defeated the regular AlphaGo by a whopping 100 games to 0, signifying a major advance in the field.

wow

hmmm...actually, that's not as impressive as I first though

like, with computer training, its easy to overfit data

so, I could easily see new AIs stomping old AIs because that's who it is playing against all the time

so, I'm not convinced that that means 100 to 0 against Lee Sedol (although it definitely could)

similar to how a chess AI that looks 1 move ahead of another AI will always win

that doesn't mean that the +1 AI will always beat a human even if the +0 AI beat a human most of the time

How does randomness factor into AI vs AI games?

It depends on the AI. What likely happens is that it outputs a list of moves with weights, and it randomly picks one based on the weights.

Ah, but against an actual human it would just go with the highest weighted one

Other algorithms will pick a random move with N% chance on a given turn

Yeah. There's the tradeoff between learning and optimization.

If you pick the best move every time, you never find other moves that might be better

If you pick random moves each time, you never get sufficiently advanced to learn advanced moves

It's generally a parameter you pass into the AI. I can't remember the term

But thats just when its learning right?

when it comes to the competition you shift gears

That depends on what the people running the AI want. Sometimes they want it to learn more against a human

Ah, the term is exploration vs exploitation

en.m.wikipedia.org/wiki/Reinforcement_learning