Google's Gemini Can Now Stomp Around as a Humanoid Robot (wired.com) 35
Google DeepMind's Gemini Robotics 2 combines vision, language, and action models to control multiple types of robots, including humanoids performing tasks such as organizing shelves, tying bags, and replacing lightbulbs. "It's another milestone in our path towards really getting towards what we call like physical AGI, which means we get a robot to do anything that a human can," Carolina Parada, head of robotics at Google DeepMind, tells WIRED. From the report: Gemini Robotics 2 combines several different AI models into a single system. Taken together, they allow a robot to make sense of its surroundings and how to act in it. A vision language model (VLM), which understands images and video, can communicate with humans and reason how to perform different tasks. Two vision language action (VLA) models, trained to understand how to move in physical space, control the robot's full-body movement as well as the movements of grippers or hands.
In video demonstrations shared ahead of the release, the company showed several different robots performing complex tasks autonomously using the amalgamated model. In one demo, Apptronik's Apollo 2 robot used hands from a company called Sharpa to tidy shelves. Google DeepMind trained the model to perform these tasks using a mix of human teleoperation, video examples, and simulations -- it's not yet possible for AI models to perform a wide range of complex tasks without specific training.
[...] Parada says Google takes a multi-layered approach to safety, with guardrails applied on each model layer. It's also introducing ASIMOV-Agentic, a new benchmark for measuring the safety of various AI systems collaborating to control a robot. The benchmark detects whether a command will result in harmful or uncertain outcome.
In video demonstrations shared ahead of the release, the company showed several different robots performing complex tasks autonomously using the amalgamated model. In one demo, Apptronik's Apollo 2 robot used hands from a company called Sharpa to tidy shelves. Google DeepMind trained the model to perform these tasks using a mix of human teleoperation, video examples, and simulations -- it's not yet possible for AI models to perform a wide range of complex tasks without specific training.
[...] Parada says Google takes a multi-layered approach to safety, with guardrails applied on each model layer. It's also introducing ASIMOV-Agentic, a new benchmark for measuring the safety of various AI systems collaborating to control a robot. The benchmark detects whether a command will result in harmful or uncertain outcome.
Scary (Score:4, Insightful)
This will be scarrier when Grok does it.
Re: (Score:1, Funny)
It'll be really good at sieg-heiling. A fluid, natural movement for sure.
Re: (Score:1, Offtopic)
I find that distinction amusing: Israel itself is a Nazi country now, if you define Nazi as "someone who did what the Nazi's did to a group of oppressed people" and not just "people who hate Jews".
If you look at how the Palestinian people are losing their country, and having all their land and possessions taken from them, being pushed into what can only be described as "ghettos" ... tell me that isn't exactly what the Nazis did to the Jews, gays, gypsies, etc. (just before they took it a step further with t
Re: (Score:2)
It's actually a pretty common definition, and in fact people often generalize "Nazi" far more than I did. Ever hear of Senifeld's "Soup Nazi"? He had nothing to do with Jews or stealing people's land and possessions .. but America clearly understood his title.
Re: (Score:2)
Re: (Score:2)
Re: Scary (Score:2)
Re: (Score:2)
What could possibly go wrong? (Score:2)
Re: What could possibly go wrong? (Score:2)
ASIMOV? (Score:2)
Re: (Score:2)
You really need the 0-th law for that to be a relatively safe position. Even that is going to hinge on your definition of "human".
Re: (Score:2)
Even that is going to hinge on your definition of "human"
Human = "politically connected multi-billionaire citizen and land owner"
Obviously.
Re: (Score:2)
Re: (Score:2)
I mean, that's certainly one take on his stories.
Another would be that for thousands and thousands of years (in his universe) the laws of robotics not only worked great, they were the entire foundation for a pivotal technology (robots). That technology was in turn essential for letting an entire civilization colonize countless planets.
Without the laws of robotics, humanity would still be stuck on a single planet.
Every Sci-Fi Movie Ever (Score:5, Interesting)
The machines won't kills us all because they're smart/evil. It will be people who are so stupid they think the Machine are Oracles, giving more control to the random word generators, trained on the best and worst parts of humanity, and allowing them to destroy us (really us allowing them to destroy ourselves) by pure randomness and not any actual malicious intent.
But yea, keep making those robots with massive security holes and continue to give them guns.
Re: (Score:2)
They'll kill us because far too many people think they're smart and will trust the perfect magic box to do smart magic box things when really they're just following a reddit circle jerk about the worst recipe possible.
Re: (Score:2)
The decision makers, the owner class, aren't ignoring the sci-fi stories. They just see them in a different way from us ordinary folks. The view them and sit there all juiced up thinking it's a beautiful and wonderful prediction of the power they will control. Because, of course, the believe themselves to be the gods in control of the new digital gods. They truly are some of the most arrogant people to have ever lived, and the warnings our authors and creative teams have provided us are viewed by these indi
Was the evil sociopath problem solved? (Score:4, Insightful)
Re: (Score:2)
I'm anticipating the first time I see the word roboticide in the news.
Re: (Score:2)
I'm anticipating the first time I see the word roboticide
I bought a can of this about a year ago, it works pretty well. I sprayed it all around the outside perimeter of the house, and a few choice inside spots. You have to re-apply it every 6 months or so, but worth it.
Re: (Score:2)
Re: (Score:2)
Shouldn’t we solve the evil sociopath problem first, before we let them out of the digital sandbox and into the real world?
The evil sociopath problem? You mean the fact that all of our elected officials, all of our lawmakers, all of our big decision makers, and all of our business leaders are evil sociopaths? Not sure how we solve that problem when they hold all the power. I mean, we sometimes shuffle them around with elections, but there's a whole lot of monetary indoctrination for newcomers, and they either get on board with the sociopathy, or they get a manufactured outrage engine pointed at them to take them down.
Oh, you me
Re: (Score:2)
Shouldn’t we solve the evil sociopath problem first, before we let them out of the digital sandbox and into the real world?
The evil sociopath problem? You mean the fact that all of our elected officials, all of our lawmakers, all of our big decision makers, and all of our business leaders are evil sociopaths?
AIs are sociopaths that far exceed nearly all of those.
Not sure how we solve that problem when they hold all the power.
Sociopaths don’t want competition from other sociopaths. Which is why the human kind seem so willing to consider restricting the AI kind, especially in the sense of maintaining their control over that AI. So a sufficient number of human sociopath decision makers, that would outnumber the cartel of human sociopaths imagining they can control a particular AI sociopath, should work.
I for one welcome our new rob... (Score:1)
on second thought, run for the hills!
Thermodynamics ? (Score:2)
ORLY (Score:1)
Well, considering his past assertions (Score:2)
Calling him a Nazi may be technically incorrect, but Elon seems to be favorably impressed by Hitler's actions. Enough so time cause emulations like MechaHitler.
Great (Score:4)
Glad Google is working on that instead of shoving more of this garbage into my phone. Ever since I got gemini I can no longer control my alarm or maps with my voice, it lies to me and says it adds stuff to my shopping list when it doesn't, and it forgot how to say "ninety" (it says ninnety which rhymes with inifinity).