|
While there has definitely been some progress I feel like the field has been stagnant for years (outside of board games). It is usually pretty easy to break down enemy behavior into scripted moves in order to defeat them, and difficulty is scaled by damage/hp rather than enemy intelligence.
|
# ? Dec 11, 2021 12:57 |
|
|
# ? Mar 28, 2024 14:32 |
|
Most genres have been drifting towards "let the player win and give them a good experience" rather than outright challenging them. The challenging games are online PvP games. I've thought about 1v1 fighting games, which notoriously have real bad AI, and the CPU opponents often cheat. These are essentially expert AI's that don't even play with the same rules that you do, which is dumb in a game where having a fair match is the point. Nowdays with these games being online, it'd be possible to collect a massive amount of play data and use it to teach a neural net, and provide a much more reasonable challenge. Would fighting the resulting AI be fun? I don't know. I'd see neural nets being the next step for many games, but they won't fit all types of games. In a lot of games the AI opposition being predictable and letting the player figure it out and win is key to the experience. I'm sure we'll see a lot of new game design ideas based on neural nets in the coming years as the tech becomes more wide spread and accessible.
|
# ? Dec 11, 2021 14:11 |