AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Is There a Transfer Learning Dimension?
We do train the original m network that was, let's say, used in alpha zero or in the open source lilo version to train the initial maia models. Then we then can take that network and find tuneit with games played by an individual player to learn a personalized model. And you know what is enough really has varied, and we've been en looking at this data problem quite deeply as well too. But five thousand games are great. Ten thousand even. Or if you've played a decent number of games, we can then kind of fine tune these base models to the individuals. So i'm excited about it. Think there's a lot of potential. I think we have