Personally as an amateur 5 dan Go player, I played the game of Go since I was 8. I stopped playing after going to high school, but I've always been paying attention to the world of Go. In 2016, Alpha Go defeated Lee Sedol 4 - 1, shocked the world as well as myself. Since then, as a computer engineering student, I've wanted to create my own Go AI, and as a metric, to defeat me.
Going into my third year, I learned more about machine learning and artificial intelligence, and skilled with frameworks like PyTorch. I decided to follow the Alpha Go published articles and build my own Go AI. And with some knowledge in web development, I also created the frontend with React to be able to play against my own AI.
Install all necessary dependencies listed in requirements.txt.
pip install -r requirements.txtRun ai_play.py to run the websocket and start a simple server to run the models.
python ai_play.pyStart up the frontend with npm, so make sure npm is installed in your device.
# Assume in the root directory
cd frontend
npm run devYou should be able to see the web page, click on the Play button on the top right corner, select Human Vs. AI to play against the models.
Recommand to play against difficulties 1 or 2 since 3 uses MCTS algorithm and the response time could be significantly longer than 1 and 2 as they only use
the model to inference.