Google DeepMind has launched a analysis preview of SIMA 2, the up to date model of its synthetic basic intelligence (AGI) agent.
This new mannequin combines Gemini’s language and reasoning talents with the talents wanted to function inside digital environments.
The primary SIMA system relied on lengthy recordings of individuals taking part in a wide array of 3D video games. By this coaching, the mannequin realized how you can handle numerous duties throughout many various recreation worlds, together with some it had by no means practiced with.
Do you know?
Subscribe – We publish new crypto explainer movies each week!
What’s Ethereum & What’s it Used For? (Animated Rationalization)
SIMA 2 makes use of the Gemini 2.5 flash-lite mannequin. DeepMind makes use of the time period AGI to explain programs that may deal with many forms of considering duties, be taught new talents, and apply data throughout completely different conditions.
DeepMind’s researchers give attention to “embodied brokers”, which implies the system has a digital physique. It could observe its surroundings and take actions in it, a lot as an individual or robotic would.
Throughout a press briefing, DeepMind senior analysis scientist Joe Marino shared how a lot the mannequin has improved. Joe Marino, senior analysis scientist at DeepMind, mentioned:
SIMA 2 is a step change and enchancment in capabilities over SIMA 1.
One other DeepMind researcher, Jane Wang, famous that SIMA 2 is just not solely about taking part in video games. She defined that the agent should perceive what is going on round it, make sense of the person’s request, and reply with fundamental reasoning.
Google not too long ago introduced its Gemini AI assistant to Maps in India, together with new instruments centered on street security and route updates. What did Miriam Daniel, Head of Google Maps, say? Learn the complete story.
