THE EVOLUTION OF VIDEO GAME AI

Over time, game AI has become more refined.

- 

ANYONE'S WHO ever played Space Invaders or one of its many clones remembers how the aliens relentlessly bear down on the player until they're either destroyed or collide with your ship. This is one of the earliest examples of game Al (Artificial Intelligence), where developers made clever use of scripted responses to enemy NPCs capable of challenging players.

These days, Al can be used to do much more than have the Pac-Man ghosts run around a maze. Innovations like procedural content generation can spontaneously generate in-game items and even entire worlds. LLMs can even be used to inform enemy and other NPC behaviors, creating more realistic interactions.

Still, the course of in-game Al hasn't always run smoothly. Here, we'll chart its progress, from chess programs that were laughably easy to defeat, to how bleeding edge Al is used in modern games to enhance play and graphics. -NATE DRAKE

From lead to gold

IN OCTOBER 1960, philosophy professor Hubert Dreyfus published a RAND Corporation memo entitled 'Alchemy and Artificial Intelligence'.

In the paper, which formed the basis of his later book, What Computers Can't Do, Dreyfus launched a scathing attack on over-optimistic predictions about the field of AI, comparing it to the mythical process of alchemy.

He took particular issue with the claim by AI pioneer Herbert A Simon in 1957 that in ten years' time, a digital computer would be the world's chess champion. Dreyfus recognized the ability of Simon's GPS (General Problem Solver) programs to solve relatively simple tasks like the 'Cannibals and Missionaries' logic problem. However, he pointed out that the NSS chess program, which Simon had co-created in the late 1950s, had been beaten by a ten-year-old novice.