The limitations of current assistive AI are they are ‘horizontal,’ and lack depth. They can tell you the weather or order pizza, but they can't talk to you about your goals, support a complex task, or manage your daily life. They have no emotional IQ and they function as highly skilled ‘question-answer machines.”

They can perform ‘single turn’ requests like “turn on the music,” analyze a lot of data and come up with a single solution to a specific question, but they cannot perform complex tasks nor reason like a human. They cannot form and maintain a deep relationship, or understand our goals.

Large Language Models have significant limitations. They focus on inputs and outputs and need brute-force learning. They have no true understanding of what they are doing, are expensive to run, and only address shallow use-cases (question-answer, or single turn prompt->output.)

GEAR focuses instead on true meaning, event trajectories with multiple steps, achieving state-based goals; and understanding an individual human and the environment they are in.