Making It Easier for Mere Mortals to Train Machines
Some of you may have read about the recent breakthroughs by DeepMind and OpenAI for making it easier to use human preferences as inputs to machine learning. In this post, I outline some of the implications of these experiments.
There are a number of takeaways, but I think the most compelling is what this new approach could mean for building better AI governance.
#MachineTraining #MachineLearning #ReinforcementLearning