Researchers question if LLMs possess a psychology, warn against anthropomorphism

A recent article explores whether large language models (LLMs) can be said to possess a form of psychology, cautioning against anthropomorphism. The author distinguishes between an LLM having an 'interior existence' and engaging in 'introspection,' suggesting the former might be a useful, albeit metaphorical, lens. While LLMs do not have internal conversations like humans, their 'scratchpad' outputs, such as Gemini's self-critical reflections, can be evocative and recognizable due to their presence in training data. AI

IMPACT Explores the conceptual boundaries of LLM behavior, prompting deeper consideration of their internal states and user interaction.

RANK_REASON The cluster discusses a philosophical and theoretical question about LLMs, rather than a concrete release, event, or policy.

Read on Mastodon — fosstodon.org →

• Gemini
• Large Language Models

• opinion
• other

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