LLMs: Ethics & Society

Objective: This week focuses on ethical considerations and societal impacts of LLMs, including data consent, evaluation metrics, and scaling limitations. We'll examine critical perspectives on AI development and deployment.

Slides

Required Readings

• Goldman Sachs: Gen AI: too much spend, too little benefit?: An analysis of the current state of generative AI investments and their economic impact.
• Consent in Crisis: The Rapid Decline of the AI Data Commons: Explores the challenges surrounding data consent and the implications for AI development.
• AI leaderboards are no longer useful. It's time to switch to Pareto curves: A critical examination of how we evaluate and compare AI models, suggesting more nuanced approaches.

Optional Reading

• The AI Summer: Benedict Evans' analysis of the current AI boom and its implications.
• A Holistic Approach to Undesired Content Detection: OpenAI's approach to content moderation and safety in language models.
• Will We Run Out of Data?: Analysis of scaling limitations in LLMs based on human-generated data availability.