Release2 sources
DeepMind reveals Gemini Robotics 2.5 and open safety benchmark "ASIMOV-jailbreak"
Ars Technica · By Ryan Whitwam ·
The Brief · written from 2 sources
- Google DeepMind revealed Gemini Robotics 2.5, a new generation of its AI-powered robotics platform, which it describes as its most capable vision-language-action model.
- DeepMind says the model enhances robot dexterity, reliability and safety, and Google says some Gemini robots completed actual work tasks.
- DeepMind also released an open safety benchmark called ASIMOV-jailbreak.
Why it matters
The release shows general-purpose robot models moving closer to completing real-world tasks, and the open ASIMOV-jailbreak benchmark gives researchers a shared way to test robot safety against jailbreak-style attacks.
Who’s affected
- robotics researchers
- AI safety researchers
- Gemini Robotics 2.5
- #robotics
- #embodied ai
- #vision-language-action
- #ai safety
- #benchmark
- #deepmind
Google DeepMind has revealed a new generation of the AI-powered Gemini Robotics platform. In the 2.5 release, Google says its robotic AI is getting tantalizingly close to completing real-world tasks as well as humans do. In fact, some Gemini robots were able to get actual work…
Timeline
Google DeepMind announces Gemini Robotics 2.5
Google DeepMind News
Gemini Robotics 2.5 accelerates real-world general dexterity
AI - Ars Technica
DeepMind reveals Gemini Robotics 2.5 and open safety benchmark "ASIMOV-jailbreak"
Full coverage · 2 sources
- 2h ago ↗
AI - Ars Technica
DeepMind reveals Gemini Robotics 2.5 and open safety benchmark "ASIMOV-jailbreak"
- 22h ago ↗
Google DeepMind News
Gemini Robotics 2.5 accelerates real-world general dexterity