Ngasekuqaleni kwaleli sonto, i-Google DeepMind ikhiphe i-Gemini Robotics-ER-1.6, umbono omusha nemodeli yolimi ukusiza amarobhothi enze umqondo wendawo ezungezile. Ukuze ibonise amakhono ayo, i-Boston Dynamics—enesivumelwano sokusebenzisa i-Gemini kumarobhothi ayo e-humanoid—ishicilele ividiyo yezinja zayo zamarobhothi zisebenzisa imodeli ukufunda i-thermometer ngesikhathi sokuhlolwa kwendawo yezimboni.
Naphezu kwamademo adonsa amehlo, imodeli entsha yamarobhothi ye-Google yathola izinzuzo ezikhulayo ngaphezu kwekhono layo lokuphakelayo lekhamera ngokuvumelana nemibandela yangaphambilini yokuphakelayo lapho iqedile ukusebenzisa amamodeli wokuphakelayo ngokwangaphambilini. Izilinganiso ze-Google. Kodwa lapho uthatha ama-feed amaningi amakhamera, imodeli ibonise ukuthuthuka. Lokhu kubalulekile, kusho i-Google, ngoba amarobhothi amaningi namuhla, njengalawo asemafekthri noma ezindlini zokugcina impahla, asebenzisa ukubukwa kwamakhamera amaningi njengekhamera engaphezulu kanye nekhamera ebekwe engalweni yerobhothi. Irobhothi kumelwe likwazi ukusebenzisa wonke lawo makhamera ukuze lenze ukuqonda okuhambisanayo kwalokho elikwenzayo futhi lazi lapho umsebenzi usuqediwe.
