Instead of saying:

“Write an email.”

Imagine assigning an AI a permanent responsibility.

Examples:

• Meeting Coordinator
• Research Analyst
• Documentation Owner
• Personal Chief of Staff

The AI continuously owns that responsibility without waiting for instructions.

Would role-based AI be more effective than prompt-based AI?

Would a worker who can autonomously think of other things to do at the gym be better than one who dispenses a towel when you press the towel button?

The entirety of recognizable user-facing AI (i.e. not applications like facial recognition of passing cars’ drivers and license plates, or a robot not falling over) is generative AI, though.

Generative LLM AI means: supply an input context sequence or embedding, such as a message. Receive a sequence that the model has been trained to produce by a reward algorithm. History and memory are limited by the input length that can be supplied and run.

That somewhat limits independent free-thinking.