|LLM|AI AGENTS|ETHICAL IMPLICATION|ETHICAL RISKS|

Power Corrupts: Hierarchies, Persuasion, and Anti-Social Behavior in LLMs

Unraveling Power Dynamics and Ethical Implications in LLM Agents

image created by the author using AI

The only thing necessary for the triumph of evil is for good men to do nothing. — Edmund Burke

The greater the power, the more dangerous the abuse. — Edmund Burke

Large language models (LLMs) showed amazing cognitive, reasoning, and general abilities. For these reasons, these models are not used for specific tasks but can be adapted for various applications. Today, these models are increasingly integrated into different tools, in dynamic and collaborative roles.

A Requiem for the Transformer?

Especially with the latest models, these LLMs are used in collaborative settings where they simulate human behaviors, can almost be considered peers, and could enter decision-making products.