What changed from existing technology?
This study shows how a single AI system can learn to control very different tasks by predicting what happens in the environment.
The study introduces a new approach for training AI systems to handle diverse control tasks using world models, which is different from previous methods that required task-specific training.
Current stage: Prototype
Classified from the research published in Nature and its publicly reported validation. Manufacturing and revenue evidence require separate confirmation.
What still needs to be solved?
Follow-up validation points are being checked.
Solving these issues would increase the chance of reaching products and industry use.
What could change if this succeeds?
Follow the path from technology to materials, components, equipment, and services.
This approach could be used to develop more versatile AI systems that can handle a wide range of tasks without needing separate training for each task.
Is a company actually connected?
Check public evidence such as research participation, patents, joint development, products, and facilities.
Company verified by public evidence
Alphabet
NASDAQ · GOOGL · United States
- Connection type
- Direct research participation
- Public evidence verified
- ✓
Keep following this technology
Revisit when new papers, patents, company participation, validation, or products appear.