When asked which is the most promising approach, it says:
"
It's difficult to say which approach is the most promising as it depends on the specific needs and goals of the game developers and the game itself. However, a combination of approaches is often used to improve the AI in strategy games like AOW4.
If I were to pick one approach that has shown great promise in recent years, it would be the use of deep learning algorithms. Deep learning has proven to be very effective in improving the performance of AI systems in a variety of domains, including game playing. By training the AI on a large dataset of human player moves and strategies, deep learning algorithms can learn how to play the game more effectively and make better decisions in future games.
Additionally, combining deep learning with other techniques such as reinforcement learning and Monte Carlo tree search can further improve the performance of the AI in AOW4. These techniques can help the AI make better decisions in different situations and increase its ability to adapt to different scenarios.
In conclusion, the most promising approach to improving the AI in AOW4 is likely to be a combination of advanced algorithms such as deep learning, reinforcement learning, and Monte Carlo tree search. However, the specific approach or combination of approaches used will depend on the specific needs and goals of the game developers and the game itself.
"
The next question would be: Why is video game AI not yet written with deep learning algorithms:
"
Video game AI can be written using deep learning algorithms, and in fact, there are many examples of games that already use deep learning for their AI systems. However, it is also true that not all games use deep learning for their AI, and there are several reasons for this.
One reason is that deep learning algorithms can be computationally expensive, especially during training, which can make them impractical for use in some games, particularly those with real-time constraints. In addition, game developers may choose not to use deep learning algorithms for their AI because they require large amounts of labeled data, which can be difficult and expensive to obtain.
Another reason is that game developers may opt for simpler, rule-based AI systems that are easier to implement and understand. These systems can be effective in many games, particularly those with well-defined rules and objectives.
Finally, game developers may also choose not to use deep learning algorithms because they may be more difficult to tune and balance compared to rule-based systems, which can affect the overall gameplay experience.
Overall, deep learning algorithms can be a powerful tool for video game AI, but their use depends on the specific needs and constraints of the game in question.
"
This is probably the answer here too. They chose rule based systems as deep learning algorithm tooling is not yet sufficiently mature to easily create complex game AI with. There may be examples of games, such as Starcraft 2, where this is done. But these are still research like examples that likely took a lot of effort on the developers part. If they did use it they may very well run into all sorts of weird problems that makes the game less fun, such as the AI running units around endlessly. Rule based AI is then a safer choice from a business perspective.
TLDR: Wait for AoW 5.