Simular AI has developed a new agent named S2, aimed at enhancing automation for computer tasks. Unlike typical large language models, S2 combines various specialized models to improve effectiveness in using apps and interacting with operating systems. CEO Ang Li emphasizes the distinction between computer-using agents and traditional models, highlighting that S2 learns from experience through a memory module that records user feedback. Its performance exceeds that of existing agents, excelling in benchmarks like OSWorld and AndroidWorld, showcasing its potential in the evolving landscape of AI-driven automation.
"S2 is designed to learn from experience with an external memory module that records actions and user feedback and uses those recordings to improve future actions."
"Computer-using agents are different from large language models and different from coding," says Ang Li, cofounder and CEO of Simular. "It's a different type of problem."
"On particularly complex tasks, S2 performs better than any other model on OSWorld, a benchmark that measures an agent's ability to use a computer operating system."
"Li, who was a researcher at Google DeepMind before founding Simular in 2023, explains that large language models excel at planning but aren't as good at recognizing the elements of a graphical user interface."
Collection
[
|
...
]