DeepMind unveils Gemini Robotics 2.5 and ASIMOV-jailbreak safety benchmark
Audio narration · Coming soon
DeepMind has introduced Gemini Robotics 2.5, a new model that demonstrates up to twice the success rate on certain tasks compared to previous versions. Alongside this, the company has released ASIMOV-jailbreak, an open safety benchmark aimed at evaluating and improving AI safety measures. This announcement marks a significant step in advancing robotics capabilities and safety standards.
Why this matters
PRISM scored this story 65/100 for interest.
Originally published by arstechnica