Minecraft beat Minecraft after 2B years | Hacker News Reader

Minecraft beat Minecraft after 2B years

pcgamer.com

26 pointsmorder9 comments

Cool idea. The way the video explains how he used probabilistic models to simulate when an action is expected to happen doesn't feel the same as it actually happening (whatever tf "happen" means in a simulated game to begin with).

It feels like saying some specific sequence of head and tails is 99.9% likely to occur after 3 million tosses, therefore let's pretend that it did occur. Though I'm not sure where the game footage came from so maybe there was some legitimate game engine running or I'm misunderstanding.

Clearly taking some liberties. Pumpkins can be grown, but wild pumpkins don't replenish. Same with snow blocks, basically. So they will get exhausted, there's no infinite supply of snow golems to eventually supply all the luck.

This is YouTube Minecraft clickbait of the sort which usually earns a "never recommend this channel" from me.

If we're nitpicking we also need a player logged in, stationed in the end and not dying for those billions of years, since MC does not update the whole world on its own :)

One player since spawn does get updated continuously, if spawn has enough pumpkins and snow...

They were specifically interested in the question of whether there were enough of them in the world for this to realistically occur, so their model did take this into account even if it didn't simulate every tick of the world directly. (and the answer is yes: on the seed they found there are enough pumpkins and snow blocks for this to eventually happen).

Including the random walk from where the snow golem happened to get put and the one open portal in the world? That doesn't seem likely at all to me.

That's pretty much what they describe in the video. The random walk itself is not simulated tick-by-tick, but turned into an faster-to-compute model of the longer term statistical behavior of the walk. It's not quite a case of 'it would happen in the real engine at exactly these points with this seed', but more a statistical argument that it would be expected to happen on that kind of timescale.

But he's assuming that sometimes snow golems causes a creeper to target them, so these are random walks on a minefield, very finite.

Sure, but it doesn't need to be infinite, just large enough.

It is a little unclear, but I think they're basically running an accelerated simulation using a probabilistic model for mob movement and enderman block pickup/putdown, but otherwise running the simulation on a block grid that's a direct mapping to the actual world, so they can load up a snapshot of the simulation into the actual game to show the state of it. I do suspect they need a bit of movie magic to show the actual events they are describing happening in-engine though.

Reply on news.ycombinator.com