Ekuqaleni kwale veki, i-Google DeepMind ikhuphe i-Gemini Robotics-ER-1.6, umbono omtsha kunye nemodeli yolwimi ukunceda iirobhothi ukuba zenze ingqiqo yendawo ezingqongileyo. Ukubonisa amandla ayo, iBoston Dynamics-enesivumelwano sokusebenzisa i-Gemini kwiirobhothi zayo ze-humanoid-ipapashe ividiyo yezinja zayo ze-robot isebenzisa imodeli ukufunda i-thermometer ngexesha lokuhlolwa kwendawo yoshishino.
Nangona i-demos ekhangayo, imodeli entsha ye-robotics ye-Google yafumana inzuzo eyongeziweyo ngaphezu kwekhono layo lokutya eligqityiweyo kwi-camera eyodwa, xa igqibile ukukwazi ukuxelela imodeli yangaphambili. Iibenchmarks zikaGoogle. Kodwa xa uthatha iifidi zekhamera ezininzi, imodeli yabonisa ukuphucuka. Kubalulekile, utshilo uGoogle, kuba iirobhothi ezininzi ziseta namhlanje, ezinjengezo zisefektri okanye iindawo zokugcina iimpahla, zisebenzisa iimbono ezininzi zekhamera njengekhamera engaphezulu kunye nekhamera ebekwe kwingalo yerobhothi. Irobhothi kufuneka ikwazi ukusebenzisa zonke ezo khamera ukwenza ukuqonda okuhambelanayo kwento eyenzayo kwaye yazi xa umsebenzi ugqityiwe.
