Do Large Language Models Have a Duty to Tell the Truth?

Explore the ethical implications of large language models (LLMs) in this thought-provoking 25-minute talk by Professor Brent Mittelstadt, PhD. Delve into the concept of "careless speech" generated by LLMs and its potential long-term risks to science, education, and democratic societies. Examine the challenges of factual inaccuracies, misleading references, and biased information in LLM-generated content. Learn about proposed solutions, including a legal duty for LLM providers to ensure truthful outputs, grounded in EU human rights law. Gain insights from Mittelstadt's expertise as Director of Research at the Oxford Internet Institute on AI ethics, algorithmic fairness, and technology law.

- Introduction
- The Problem of Hallucinations
- Combatting Hallucinations
- Carless Speech
- Legal Duties to Tell the Truth
- Conclusion