[Paper Review] The Dark Side of ChatGPT: Legal and Ethical Challenges from Stochastic Parrots and Hallucination

Zihao Li|arXiv (Cornell University)|Apr 21, 2023

Artificial Intelligence in Healthcare and Education36 citations

TL;DR

The paper warns that stochastic parrots and hallucinations in LLMs pose legal and ethical risks that may be underestimated by current EU AI regulations, and argues the European regulatory paradigm must evolve.

ABSTRACT

With the launch of ChatGPT, Large Language Models (LLMs) are shaking up our whole society, rapidly altering the way we think, create and live. For instance, the GPT integration in Bing has altered our approach to online searching. While nascent LLMs have many advantages, new legal and ethical risks are also emerging, stemming in particular from stochastic parrots and hallucination. The EU is the first and foremost jurisdiction that has focused on the regulation of AI models. However, the risks posed by the new LLMs are likely to be underestimated by the emerging EU regulatory paradigm. Therefore, this correspondence warns that the European AI regulatory paradigm must evolve further to mitigate such risks.

Motivation & Objective

• Highlight the legal and ethical risks tied to stochastic parrots and model hallucinations in large language models (LLMs).
• Critique the sufficiency of existing European AI regulation in mitigating these risks.
• Advocate for an evolution of EU regulatory frameworks to address emerging LLM challenges.

Proposed method

• Provide a focused correspondence discussing regulatory implications of LLM behavior (stochastic parrots and hallucination).
• Critically analyze the European AI regulatory paradigm and its transparency obligations.
• Reference related work to connect regulatory gaps with practical LLM risks.

Experimental results

Research questions

• RQ1 What legal and ethical challenges do stochastic parrots and hallucinations in LLMs introduce?
• RQ2 Are current EU AI regulations sufficient to mitigate these risks?
• RQ3 How should EU AI regulation evolve to better address LLM-specific risks?

Key findings

• Stochastic parrots and hallucinations generate novel risk profiles for accountability, transparency, and governance.
• Current EU AI regulatory approaches may underestimate or inadequately address these risks.
• There is a need for evolving transparency and accountability mechanisms within EU AI law to mitigate LLM-related harms.