Anyone who sees the agent learn to play for the first time will be amazed by it--it seems so human-like. The initial moves are always clumsy; slowly, the agent learns which way to go, though it is slow and often misses the ball. However, as the learning continues, the agent becomes an expert.

Yet it is very unlike us; once we learn to play a game, we can easily use that knowledge in any other similar situation. The RL agent will not be able to do so--even a simple thing such as changing the size of environment space will bring it back to zero. Transfer learning, a technique being investigated by researchers, can help the agent use the knowledge it learned in one environment space in another environment space, and perhaps pave the way for true artificial general intelligence one day.