Google DeepMind has shown an updated version of the AI model of the world called Genie 3. Such a system creates three-dimensional interactive worlds in real time, in which a person can move.
In the new version, the working session is longer, so the worlds exist not for tens of seconds, but for several minutes. They can also remember visual details, even if the user turns away.
World models are a special class of artificial intelligence that creates simulated environments for learning, playing games, or training neural networks and robots. The user gives such a model a text request, after which it generates a space similar to a game, but made not manually, but by artificial intelligence.
The virtual worlds in Genie 3 run at 720p resolution and 24 frames per second. They can store details in visual memory for about a minute and support dynamic changes. For example, the user can change the weather. At the moment, the model is available to a limited number of users, as Google wants to first analyze the risks and limitations before opening the model to the general public.
We previously wrote that artificial intelligence will destroy humanity with a probability of 95%.
To be continued…
Only registered users can leave comments