LLMs are not like you and me – and never will be

11 days ago

Copy Link

LLMs are fundamentally different from humans and do not think like humans despite some superficial similarities.
LLMs lack proper world models, leading to errors in tasks requiring temporal reasoning, common sense, and factual accuracy.
Examples of LLM errors include incorrect historical facts, inability to play chess reliably, and failure to account for inflation or time.
LLMs operate based on autocomplete mechanisms rather than genuine understanding or reasoning.
The field of AI has not adequately addressed fundamental reasoning frameworks like time, space, and causality.
Using LLMs as agents for complex tasks is unreliable due to their lack of proper world models and reasoning capabilities.
Critics argue that claims of LLMs thinking like humans are based on ignorance or denial of how both LLMs and human brains work.
Despite improvements, LLMs remain limited and cannot be trusted for tasks requiring deep reasoning or factual accuracy.

Hasty Briefsbeta