Generative Code Strategies
Subbarao discusses the brute force nature of generating diverse code possibilities through LLMs, emphasizing the importance of external knowledge to guide the generation process. He introduces the concept of prompt diversification as a means to enhance creativity and improve verification outcomes. Tim shares insights from his conversation with Ryan, who believes in the potential of future LLMs to autonomously verify their solutions, highlighting the evolving landscape of machine learning capabilities.In this clip
From this podcast
Machine Learning Street Talk (MLST)
Prof. Subbarao Kambhampati - LLMs don't reason, they memorize (ICML2024 2/13)
Related Questions
Have you seen a way to unit test large language models (LLMs) that are super helpful, as discussed in the episode How to Systematically Test and Evaluate Your LLMs Apps // Gideon Mendels // #269?
Have you seen a way to unit test large language models (LLMs) that are super helpful, as discussed in the episode How to Systematically Test and Evaluate Your LLMs Apps // Gideon Mendels // #269?
What's your opinion on using large language models (LLMs) for scientific research, especially for generating new ideas for hypotheses, as discussed in the episode 888: Marc Andreessen | Exploring the Power, Peril, and Potential of AI and the clip The Power of Computer Creativity?