As applications of artificial intelligence spread, more computation has to occur — and more efficiently with lower energy consumption — on local devices instead of geographically distant data centers in order to overcome frustrating delays in respons