00:00¿Cómo estás hoy, Apolo?
00:01Estoy un poco nervioso. ¿Eso es un micrófono?
00:16Para el robot, lidiar con los desafíos del mundo físico es imposible.
00:23Pero para nosotros, los humanos, se viene tan natural y tan intuitivamente
00:27que casi no nos piensan como inteligencia.
00:32Usted probablemente ha visto todos estos increíbles videos
00:35de especializados humanos, robots,
00:37running, jumping, backflipping.
00:39La diferencia principal en Gemini Robotics 2
00:42es que intentamos construir un modelo generalista de la tecnología
00:45que va a dar mucho más valor
00:47si un robot puede hacer muchas tareas de tareas.
00:50A.I. models offer a way
00:53to navigate the messy complexity
00:56of our human environment.
00:58The way to think about it is the brain, Gemini Robotics,
01:01is what controls
01:02the whole body of the humanoid,
01:05the delicate movements of the sharper hand,
01:07and the grippers of the duo.
01:09We focus on three main things
01:11for this release.
01:12The first one is whole body control.
01:14We are very used to move our body
01:16in a very corny manner,
01:18but it's not true for robots.
01:20Even though this field is a very simple task for us,
01:22but for robots,
01:23I need to make many, many decisions
01:25across the entire body
01:26to achieve this.
01:28The second thing is about dexterous hands
01:30and particular manipulations of objects,
01:33way beyond just pick and place.
01:36As a field, we're humbled
01:37by things like screwing a light bulb,
01:40where you're interacting with smaller objects,
01:43and those are really hard problems
01:45that we are trying to solve.
01:48When we first came up
01:50with the trash bag task in particular,
01:53several people did think it was impossible.
01:55You don't think about
01:56how to drive 22 separate joints
01:57when you operate your hand,
01:59but that's what we're asking
02:01these AI models to do.
02:03And third thing is in Gemini Robotics 2,
02:05we add a feature
02:06that enables multiple robots
02:09to collaborate to accomplish
02:10the same task simultaneously.
02:12Hey, duo.
02:13Kit all tools in the bin,
02:15close the kit,
02:16and put the kit back into the bin.
02:18On it.
02:19Each of these two robots
02:20is running that same stack,
02:21and so rather than having
02:22one neural network
02:23that controls both robots,
02:24each have their own copy,
02:26and they're each doing
02:26their own individual thinking,
02:27and they're actually orchestrating
02:29through reasoning.
02:30Awesome precision, duo.
02:31Keep going.
02:33We're really after intelligence
02:35and generality.
02:36Being able to react
02:37to the scene changing,
02:38that is where
02:39the intelligence comes in.
02:42I'm super excited about robotics.
02:44I think it's moving really fast,
02:46and AI is really
02:48the missing piece
02:49of the entire robotics puzzle.
02:54It is quite exciting
02:55to be in the spotlight.
02:56Thank you.
Comentarios