How to Win Snake: The UNKILLABLE Snake AI

Why does he remind me of Hiccup

ah yes, THE OF GRID.

You can use flood fill and a time-weightmap to make it be able to predict how it can dodge its tail.

Use flood fill to make a weightmap of how many steps it would take to reach a given location, and if the step count at the index of that weightmap during the Hamiltonian search is greater than the snake segment's distance to its tail at that location, that location will be available to the Hamiltonian search as a clear spot.

Idk if that made any sense, but if it did, I think it'd be a better step optimization than "Applefear Mode".

It's a great achievement, thanks for sharing! How about making a Mine Sweeper AI?

autism + adderall =

couldn't you use a similar code to what a mouse race robot uses the snake goes the nearest route to the apple if the snakes in the way do the next shortest way a bit harder cause you have to think like 4 apples Infront

Can the algorithm be improved by relaxing the necessity of a Hamiltonian cycle, and only insisting that if part of the board gets cut off, the snake has enough time to escape from an area.

Another thing I noticed, when going straight for an apple that is not horizontally or vertically aligned, it would seem that a possible optimisation could be made by minimising the area on one side.

:D

The of grid

Hamilton is the best. Oh, uh, I mean, Hamilton-ion or whatever.

The most impressive thing about this video is that you're using Matlab and not going insane

Thanks for showing this. Now I can implement it in my professional snake gaming career.

Didnt undertsood a thing , yet loved it

Unreal how much thought and time went into this - and for such a frivolous matter

snake was what people did to kill time in 2001 before phones could surf the web

I wonder if it can be improved by giving it a planning horizon whereby it can make any move it wants, so long as a hamiltonian path is possible within the next N steps after making that move.

I would have loved to see the recording of the best game it played

Could make a nice video game boss, where player tries to get out of the snake's path.

Has anyone ever told you you sound like Mark Rober? Great video by the way!

"it could just go "zzzt zzzt" or even "z z z z z z z" but instead it has to go you know BLREHHER"

You sound like mark rover or is it just me

You sound exactly like Mark Rober

this guy sounds like mark rover

P=NP

The snake is not unintelligent, it always has a plan. You just want it to speedrun Snake

I played that game till I wore out that’s all us oldies was content with !

THOABRGG

Why does he sound like mark rober

What about a Hilbert curve for a 2^n×2^n

I loved learning code on MathLab, I coded a sound pitch detection algorithm. It is really not efficient. I think your renders could have taken 10 minutes in c++, even python instead of 6 hours on ML ;)....
Very enjoyable video, thanks

now, do that on hex grid... :)

it whot proboly bi fster to först run a algoritom and hafe waje trov so swetses it the algoritom automaticaly when neded

slow

I don't know if prophecy/coincidences are real but I strongly felt that someone close to me is a "snake" an hour or so ago only to find this video in the recommended

Well done on taking something so arcane seeming and making it accessible and interesting. You made me appreciate all the thinking, hard work and collaboration between people that goes into these ai programs

you sound similar to mark rober

Do you realy need AI. just snake back and forth leaving the side column open to allow you to return

He sounds like mark rober

GREAT

Do ppl really code in matlab?

my snake won in 39k moves

Bro sounds like Mark Rober… 😐

i would create an ai that finds the move that gets it closest to the apple while creating the least amount of failure points and always making sure it can get to every square that is safe and not filled in

I've won a game of snake this without even knowing it was an actual thing

Just use supercomputers to make it go exponentially faster