Large Language Models are Biased Because They Are Large Language Models

TL;DR

This paper argues that harmful biases are an inherent consequence of the fundamental design of large language models, suggesting that addressing bias requires rethinking their foundational principles.

Contribution

It challenges the assumption that bias can be mitigated within current LLM frameworks, advocating for a fundamental redesign of AI based on LLMs.

Findings

Bias is an inevitable outcome of current LLM design

Addressing bias requires rethinking foundational assumptions

Current LLMs inherently encode harmful biases

Abstract

This position paper's primary goal is to provoke thoughtful discussion about the relationship between bias and fundamental properties of large language models. I do this by seeking to convince the reader that harmful biases are an inevitable consequence arising from the design of any large language model as LLMs are currently formulated. To the extent that this is true, it suggests that the problem of harmful bias cannot be properly addressed without a serious reconsideration of AI driven by LLMs, going back to the foundational assumptions underlying their design.