I've created a genetic algorithm that learns to play snake. It scored 123 points

  • 3 Replies


  • Roomba
  • *
  • 2
Hi guys! A few days ago I've created a genetic algorithm that learns to play snake. It took 30 generations to reach 123 points. I used C++ and SFML. Each snake's brain is a neural network.

Here's the video:




  • Trusty Member
  • ************
  • Bishop
  • *
  • 1543
    • contrast-zone
Look at what the Nature managed to do only with a similar genetic production. Though She needed 5.4 billion years up to now. The key point in history was creation of self-duplicating unit. Genetics did the rest.
There exist some rules interwoven within this world. As much as it is a blessing, so much it is a curse.



  • Emerged from nothing
  • Trusty Member
  • *******************
  • Prometheus
  • *
  • 4511
  • First it wiggles, then it is rewarded.
    • Main Project Thread
So if we release duplicating nanobot cells...we just wait for everything to go fine right? Bad move oh no! Grey goo! Maybe intelligence has to do the rest!? ?O.o?

Those wireless artificial cells will need quicker smarter and more efficient guidance to be safe and have powerful thinking, manipulation, and sensing! They'll do so much at all levels!



  • Trusty Member
  • *********
  • Terminator
  • *
  • 828
  • It's Hip to be Square
Snake is already a solved problem for AI:
Feeling Chatty?