New research proposes "subjective functions" for AI goal generation

A new paper introduces the concept of "subjective functions" as a framework for understanding how agents, including humans, can generate their own goals and objectives. The research explores how these higher-order functions, defined relative to an agent's internal state rather than external tasks, can lead to the synthesis of new objectives. Expected prediction error is presented as a specific example of such a subjective function, drawing connections between machine learning, psychology, and neuroscience. AI

IMPACT Introduces a theoretical framework for AI goal generation, potentially influencing future research in agent design and reinforcement learning.

RANK_REASON The cluster contains a new academic paper on a theoretical concept in AI. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• Samuel J Gershman
• Subjective functions

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →