Today we're joined by Peter Hase, a fifth-year PhD student at the University of North Carolina NLP lab. We discuss "scalable oversight", and the importance of developing a deeper understanding of how large neural networks make decisions. We learn how matrices are probed by interpretability researchers, and explore the two schools of thought regarding how LLMs store knowledge. Finally, we discuss the importance of deleting sensitive information from model weights, and how "easy-to-hard generalization" could increase the risk of releasing open-source foundation models.
The complete show notes for this episode can be found at twimlai.com/go/679.
Controlling Fusion Reactor Instability with Deep Reinforcement Learning with Aza Jalalvand - #682
GraphRAG: Knowledge Graphs for AI Applications with Kirk Marple - #681
Teaching Large Language Models to Reason with Reinforcement Learning with Alex Havrilla - #680
Coercing LLMs to Do and Reveal (Almost) Anything with Jonas Geiping - #678
V-JEPA, AI Reasoning from a Non-Generative Architecture with Mido Assran - #677
Video as a Universal Interface for AI Reasoning with Sherry Yang - #676
Assessing the Risks of Open AI Models with Sayash Kapoor - #675
OLMo: Everything You Need to Train an Open Source LLM with Akshita Bhagia - #674
Training Data Locality and Chain-of-Thought Reasoning in LLMs with Ben Prystawski - #673
Reasoning Over Complex Documents with DocLLM with Armineh Nourbakhsh - #672
Are Emergent Behaviors in LLMs an Illusion? with Sanmi Koyejo - #671
AI Trends 2024: Reinforcement Learning in the Age of LLMs with Kamyar Azizzadenesheli - #670
Building and Deploying Real-World RAG Applications with Ram Sriharsha - #669
Nightshade: Data Poisoning to Fight Generative AI with Ben Zhao - #668
Learning Transformer Programs with Dan Friedman - #667
AI Trends 2024: Machine Learning & Deep Learning with Thomas Dietterich - #666
AI Trends 2024: Computer Vision with Naila Murray - #665
Are Vector DBs the Future Data Platform for AI? with Ed Anuff - #664
Quantizing Transformers by Helping Attention Heads Do Nothing with Markus Nagel - #663
Create your
podcast in
minutes
It is Free
20/20
The Dropout
FiveThirtyEight Politics
Ten Percent Happier with Dan Harris
World News Tonight with David Muir