Fundamental Limitation
Fundamental limitations in artificial intelligence research currently focus on identifying and addressing bottlenecks in model capabilities and performance. Active research areas include exploring the limitations of large language models (LLMs) in reasoning, particularly compositional abilities and handling complex tasks; analyzing the inherent quadratic time complexity of transformer architectures and the challenges of developing subquadratic alternatives; and investigating the impact of data quality and size on model performance and safety. Understanding these limitations is crucial for improving the reliability, safety, and efficiency of AI systems and for developing more robust and generalizable models across various applications.
Papers
Jailbreak Defense in a Narrow Domain: Limitations of Existing Methods and a New Transcript-Classifier Approach
Tony T. Wang, John Hughes, Henry Sleight, Rylan Schaeffer, Rajashree Agrawal, Fazl Barez, Mrinank Sharma, Jesse Mu, Nir Shavit, Ethan Perez
A Primer on Large Language Models and their Limitations
Sandra Johnson, David Hyland-Wood