Bus 377 week 6 discussion
The AI would first create a large pool of potential moves and then select those that it thinks will produce success. To do this, it might compare each move against historical data from past games, in addition to evaluating any potential benefits or drawbacks associated with those moves such as making a favorable trade-off between attacking strength and defensive security. As it plays the game more and more times, the AI will continue to refine its decision-making process through trial and error until eventually settling on a ‘best’ strategy for winning the game.
In addition to learning through experience, reinforcement learning algorithms also allow us to set certain objectives that we want our AI agent to strive for when selecting moves. For instance, we may want our agent to prioritize defense over attack or try moving certain pieces first before others if they offer greater long-term benefits. By setting these goals within the context of our desired outcome (i.e., winning), we can help ensure that our AI is able make better decisions while still taking into account all relevant factors related to each move being considered.