<< Back to Programming Forum   Search

Posts 1 - 1 of 1   
Article in Wired re: AI causality, not correlation: 10/15/2019 14:24:23

Level 57
TL;DR: Good article that should be of interest to anyone interested in Deep Learning AI, with a note about pioneering gaming developer Will Wright.



I am a programmer, in the sense that the fry cook at your local greasy spoon restaurant is a chef. I'm terrible at real math, but am respectable enough at Warzone, I believe. In other words, don't ask me to program anything, but I might have some worthwhile ideas.

Warzone's AI isn't really an AI, just a set of complex algorithms, I suppose. There's almost certainly formal levels of AI, but for the sake of my laziness, I will call this Level 1. I think WZ's AI is reasonably good for level 1, and I have a number of suggestions for making it better, involving memory of past turns, efficient wayfinding, etc.

Level 2 is the AI learning about its opponent, and developing counter-tactics, feints, and ultimately counter-strategies to defeat that opponent. I believe the earliest attempt at true gaming AI was Will Wright's classic game for the Commodore 64 in 1984; Raid on Bungeling Bay (ROBB). ROBB actually learned the game player's techniques as the scenario developed, and used different sets of tactics against the player. WZ's AI could implement these features, but it would need an understanding of tactics, operations, and strategies, even grand strategy. Wright's work with AI led into his development with the Sims, and with Spore. IMHO, Spore missed its potential to be an MMORPG giant.

Level 3 is the incorporation of big data to learn from individual players versus the AI and individual players versus other players. The massive amount of data from past WZ games incorporated into deep learning AI could bring about a major upgrade in gameplay. All it takes is time and money, heheheh!
Posts 1 - 1 of 1