Deep reinforcement learning from human preferences
This reduces the cost of human oversight for RL systems, making it practical for state-of-the-art applications, though it is incremental in improving human-in-the-loop methods.
The paper tackles the problem of communicating complex goals to reinforcement learning systems by using non-expert human preferences between trajectory segments, enabling effective solution of tasks like Atari games and robot locomotion with feedback on less than 1% of interactions and training novel behaviors in about an hour of human time.
For sophisticated reinforcement learning (RL) systems to interact usefully with real-world environments, we need to communicate complex goals to these systems. In this work, we explore goals defined in terms of (non-expert) human preferences between pairs of trajectory segments. We show that this approach can effectively solve complex RL tasks without access to the reward function, including Atari games and simulated robot locomotion, while providing feedback on less than one percent of our agent's interactions with the environment. This reduces the cost of human oversight far enough that it can be practically applied to state-of-the-art RL systems. To demonstrate the flexibility of our approach, we show that we can successfully train complex novel behaviors with about an hour of human time. These behaviors and environments are considerably more complex than any that have been previously learned from human feedback.