Trenchant critique of reinforcement learning.

Mahadevan, a fellow of the AAAI, sets out his evolved views on the limits of reinforcement learning. It is not that RL cannot perform really useful functions. But, if your goal is to develop artificial general intelligence, then RL is by no means enough. In parallel with the spread of RL applications, especially in the commercial world, there is growing criticism of RL’s limitations. Machine intelligence is increasingly crying out for alternative techniques, whether to master common sense or hypothesising or other abilities that human beings take for granted.

Link to Answer:

You may also like to browse other AI posts: