Xnor.ai, spun off in 2017 from the non­prof­it Allen Insti­tute for AI (AI2), has been acquired by Apple for about $200 mil­lion. A source close to the com­pa­ny cor­rob­o­rat­ed a report this morn­ing from Geek­Wire to that effect.

Apple con­firmed the reports with its stan­dard state­ment for this sort of qui­et acqui­si­tion: “Apple buys small­er tech­nol­o­gy com­pa­nies from time to time and we gen­er­al­ly do not dis­cuss our pur­pose or plans.” (I’ve asked for clar­i­fi­ca­tion just in case.)

Xnor.ai began as a process for mak­ing machine learn­ing algo­rithms high­ly effi­cient — so effi­cient that they could run on even the low­est tier of hard­ware out there, things like embed­ded elec­tron­ics in secu­ri­ty cam­eras that use only a mod­icum of pow­er. Yet using Xnor’s algo­rithms they could accom­plish tasks like object recog­ni­tion, which in oth­er cir­cum­stances might require a pow­er­ful proces­sor or con­nec­tion to the cloud.

Xnor’s saltine-sized, solar-pow­ered AI hard­ware rede­fines the edge

CEO Ali Farha­di and his found­ing team put the com­pa­ny togeth­er at AI2 and spun it out just before the orga­ni­za­tion for­mal­ly launched its incu­ba­tor pro­gram. It raised $2.7M in ear­ly 2017 and $12M in 2018, both rounds led by Seattle’s Madrona Ven­ture Group, and has steadi­ly grown its local oper­a­tions and areas of busi­ness.

The $200M acqui­si­tion price is only approx­i­mate, the source indi­cat­ed, but even if the final num­ber were less by half that would be a big return for Madrona and oth­er investors.

The com­pa­ny will like­ly move to Apple’s Seat­tle offices; Geek­Wire, vis­it­ing the Xnor.ai offices (in inclement weath­er, no less), report­ed that a move was clear­ly under­way. AI2 con­firmed that Farha­di is no longer work­ing there, but he will retain his fac­ul­ty posi­tion at the Uni­ver­si­ty of Wash­ing­ton.

An acqui­si­tion by Apple makes per­fect sense when one thinks of how that com­pa­ny has been direct­ing its efforts towards edge com­put­ing. With a chip ded­i­cat­ed to exe­cut­ing machine learn­ing work­flows in a vari­ety of sit­u­a­tions, Apple clear­ly intends for its devices to oper­ate inde­pen­dent of the cloud for such tasks as facial recog­ni­tion, nat­ur­al lan­guage pro­cess­ing, and aug­ment­ed real­i­ty. It’s as much for per­for­mance as pri­va­cy pur­pos­es.

Its cam­era soft­ware espe­cial­ly makes exten­sive use of machine learn­ing algo­rithms for both cap­tur­ing and pro­cess­ing images, a com­pute-heavy task that could poten­tial­ly be made much lighter with the inclu­sion of Xnor’s econ­o­miz­ing tech­niques. The future of pho­tog­ra­phy is code, after all — so the more of it you can exe­cute, and the less time and pow­er it takes to do so, the bet­ter.

The future of pho­tog­ra­phy is code

It could also indi­cate new for­ays in the smart home, toward which with Home­Pod Apple has made some ten­ta­tive steps. But Xnor’s tech­nol­o­gy is high­ly adapt­able and as such rather dif­fi­cult to pre­dict as far as what it enables for such a vast com­pa­ny as Apple.

Source link