What happens when AI stops playing games? [Paid Content for IBM]

What hap­pens when AI stops play­ing games? [Paid Con­tent for IBM]



IBM,AI,artificial intelligence,machine learning,deep learning,natural language,watson,gaming frameworks,epic studios,branded content,vox creative



The ques­tion, “can a machine be made to think like a per­son?” has always been tied to strat­e­gy games. Games, with their clear rules and obvi­ous win­ners and losers, were per­fect prov­ing grounds for ear­ly com­put­er sci­en­tists, who could break them down into clear­ly defined prob­lem sets. Audi­ences could mar­vel at the progress of man vs machine face offs, even if they didn’t ful­ly under­stand the under­ly­ing technology.

The ear­li­est AI game sys­tems relied on a brute force, top-down approach: pro­gram­mers down­loaded every pos­si­ble out­come into their AI sys­tems, which were built around nar­row­ly defined, rule-based cri­te­ria. In the 1950s, the ear­li­est ver­sions of neur­al net­works and machine learn­ing arrived. This rep­re­sent­ed a shift towards bot­tom-up pro­gram­ming, sys­tems designed to deter­mine the prob­a­bil­i­ty of var­i­ous out­comes based on train­ing data. One ear­ly sys­tem, a vir­tu­al rat solv­ing a maze, was made up of vac­u­um tubes, motors, and clutch­es. As the rat nav­i­gat­ed, the machine learned and shift­ed probabilities.

IBM debuted their first AI Grand Chal­lenge — a mul­ti-year effort meant to push the lim­its of arti­fi­cial intel­li­gence — in 1997. Deep Blue beat reign­ing chess world cham­pi­on Gary Kas­parov, and was seen as the final tri­umph for the top-down approach. In the last decade, advances in machine learn­ing, deep learn­ing, and nat­ur­al lan­guage pro­cess­ing paved the way for AIs like Wat­son on Jeop­ardy!, Alpha­Go, and Dota 2—increasingly com­plex sys­tems that still exist­ed with­in the frame­work of games.
Learn more here — 

[Paid Con­tent for IBM]



#stops #play­ing #games #Paid #Con­tent #IBM