LLM Risks, Limitations, and Mitigation Strategies - Lecture

Explore the risks, limitations, and mitigation strategies associated with Large Language Models (LLMs) in this comprehensive 44-minute presentation by Kaushik Roy from the AI Institute at UofSC. Delve into various types of risks, including those posed by AI giants and individual actors. Examine the concept of stochastic parrots and the potential existential threat to the creator economy. Investigate the impact of AI on the creative landscape and the implications of automation through new technologies. Consider the environmental concerns related to the carbon footprint of LLMs. Learn about advanced techniques such as chain of thought prompting, mind mapping, and multimodal AI. Gain insights into methods for controlling text generation and develop a deeper understanding of the challenges and opportunities presented by LLMs in today's rapidly evolving AI landscape.

Intro
Types of Risks
AI giants
Individual points
Stochastic parrots
Existential risk to Creator economy
Impact of AI on the creative economy
Automation with new technology
Carbon footprint
Chain of thought prompting
Mind map
Multimodal AI
Control Text Generation