In a study conducted by MIT last year, they reported that AI that can truly replace humans is still decades away from becoming a reality. Human efficiency in carrying out tasks using vision alone prevents AI from putting millions of people out of work. The Information today reported that OpenAI is developing two "agents" that can take control of a device before performing the tasks assigned to it.
Agents are AI software developed to perform specific tasks. Two agents are reportedly being developed by OpenAI. The first takes over the owner's device (perhaps a PC) and performs complex tasks that have traditionally required humans.
The latter can be tasked with doing web-based work such as buying plane tickets and collecting user data. In other words, both agents perform the process of automating human tasks in the office. OpenAI has not yet issued a comment on this report.
To be sure, the concept of agents here is not something new. Humane AI Pin and Rabbit R1, two AI devices that have been announced are marketed with similar capabilities. Rabbit R1 developers showed a demo of Rabbit OS being taught to perform tasks using the camera. Once trained, it then performs the same task again without supervision.