
"Google DeepMind's Genie 3 marks a shift in artificial intelligence (AI) toward always-on systems that generate and sustain interactive environments without pausing. This world model creates photorealistic worlds from text prompts, running continuously at 24 frames per second for minutes, predicting actions and maintaining consistency. It points to persistent simulations that evolve independently, aiding agent training and approaching AGI. While the implications for this advancement span robotics and media, investors panicked on Friday over the threat it potentially represents to game engines."
"Genie 3 builds on prior models like Genie 2, which generated 3D worlds but lacked real-time interaction. This new version enables users to input text like "glacial lake with paths" and navigate the resulting 720p environment instantly via keyboard controls, with responsive elements like splashing water. It predicts frames on the fly, holding coherence for several minutes before drift occurs. DeepMind highlights its 11-billion-parameter scale, trained on vast video data for physics intuition."
Genie 3 generates photorealistic, interactive worlds from text prompts, running continuously at 24 frames per second for minutes while predicting actions and maintaining consistency. The model uses about 11 billion parameters and is trained on extensive video data to learn physics intuitions. Environments hold coherence for several minutes before drift occurs and can include responsive elements such as splashing water. Persistent simulations can evolve autonomously, support independent agents, and scale indefinitely to accelerate agent training and robotics development toward AGI. In gaming, prompt-based world creation can reduce manual design. Market reactions caused sharp stock declines for Unity and related companies, presenting a potential buying opportunity.
Read at 24/7 Wall St.
Unable to calculate read time
Collection
[
|
...
]