How playing video games improves AI

The glob­al video game mar­ket is a colossal enter­tain­ment sec­tor, cur­rently estim­ated at over €100bn. Cer­tain tech­no­lo­gic­al devel­op­ments have been made only because of the new oppor­tun­it­ies offered by video games. The most strik­ing example is com­puter graph­ics cards that evolve rap­idly today thanks to the video games sec­tor push­ing real-time 3D to its lim­its. The same goes for real-time ren­der­ing algorithms. Invest­ments and the sub­sequent pro­gress that goes with them would not hap­pen at the same pace without the video game eco­nomy. Not to men­tion aug­men­ted real­ity (AR) or vir­tu­al real­ity (VR), for which the same it hap­pen­ing. As such, video games have an impact on research, par­tic­u­larly arti­fi­cial intel­li­gence (AI).

What’s more, AI imple­men­ted in video games is a spe­cial­ised sec­tor. Indeed, in a video game, con­trary to what one might think, the goal is not neces­sar­ily to beat the play­er at all costs. Rather, it’s more about offer­ing a chal­lenge. The AI must win some of the time, but also lose from time to time oth­er­wise play­ers stop play­ing. But AI in video games is not just about anim­at­ing the oppos­i­tion. There are oth­er object­ives, too.

Object­ive 1: under­stand players

First, AI is use­ful in terms of the player’s per­cep­tion of their envir­on­ment; mean­ing how the cam­era pos­i­tions itself in a coher­ent way in rela­tion to the avatar, the way light dif­fuses on the envir­on­ment, the music, etc. When I am mov­ing in a game, my envir­on­ment must adjust based what I am doing, and anti­cip­ate what I want it to do to assist me.

But what people first per­ceive of AI in video games are the NPCs, the “non-play­er char­ac­ters”. Their pur­pose is to cre­ate a brief illu­sion. In a war game, most play­ers don’t play around watch­ing guards for half an hour. Instead, the play­er will see the char­ac­ter for a few seconds before shoot­ing it or being shot them­selves. An AI that works well is above all an AI that you don’t see is there, they blend into the game. The situ­ation to avoid is an NPC that stumbles end­lessly on the edge of a table being blocked from return­ing to ‘nor­mal’ behaviour.

The AI of oppon­ents is thus much easi­er to code than that of com­puter-con­trolled allies.. This is because, with the oppon­ent, the rela­tion­ship is brief. He doesn’t need to really under­stand what I’m doing, where­as an ally does. It must know if I need help catch­ing enemies in secret or cov­er­ing me in an attack. That’s why we often use orders, it greatly sim­pli­fies the pro­gram­ming of ‘ally’ AIs.

Object­ive 2: Sur­prise players

If the first level of AI in a game is to do everything to pro­duce a coher­ent and cred­ible envir­on­ment, the second level is to sur­prise the play­er; a more advanced task. Some suc­ceed, oth­ers don’t. The his­tory of gam­ing is littered with many fam­ous examples of AIs. From Pac-Man’s ghosts, through HalfLife’s mar­ines (able to coordin­ate for tac­tic­al actions), Cap­tain Blood’s ali­ens (with very real­ist­ic dia­logue cap­ab­il­it­ies), Creatures (real pets with a cer­tain intel­li­gence) or Black & White’s semi-autonom­ous avatars (able to learn your play style to adapt their beha­viour), or Event0’s space sta­tion (emo­tion­al AI that you have to con­vince to help you), many games are based more par­tic­u­larly on more com­plex AIs.

Object­ive 3: adapt to players 

The third func­tion is to ana­lyse the play­er, his reac­tions to the chal­lenges, where we find the trend of adapt­ive AI or stat­ist­ic­al learn­ing. The game sends data to the serv­ers of the pub­lish­er or stu­dio, and this data will be ana­lysed to find out what was appre­ci­ated or not in the game. And this data ana­lys­is will be traced back to the game design­er. Today, in a mar­ket that has become largely online, this pro­cess has become sim­pler. Very early on, alpha or beta ver­sions of a game are launched to have as many people as pos­sible play them and col­lect as much data as pos­sible so as to modi­fy the game accord­ingly for the time of the offi­cial release; it has a lot of pos­it­ive effects.

For example, it cre­ates a com­munity and some game stu­di­os, very few of them, are doing research on real-time con­tent modi­fic­a­tion. It’s a real break­through in the concept of real-time ana­lys­is. A new step would be to have a rela­tion­ship with the play­er, not neces­sar­ily via dia­logue, but a rela­tion­ship that is cred­ible enough for the play­er to feel that they are deal­ing with real human intel­li­gences. This would be in line with the ini­tial uto­pia. But we are not there yet. For video games, and even if it plays an import­ant role, AI is still a sec­ond­ary arte­fact, second only to the visu­als, which are still the main selling point. It’s easi­er to quickly con­vey the key points of a game (its uni­verse, its story, its char­ac­ters) through visu­als than through AI, which takes longer to exper­i­ment with and is there­fore less of a mar­ket­ing point.