Monday Oct 14, 2024
Revolutionizing AI: New Models, UN Resolutions, and Viral Discoveries
In this episode, we explore groundbreaking advancements in artificial intelligence, including Google DeepMind's diversity-rewarded CFG distillation method that enhances generative models, and NVIDIA's innovative approach to upcycling large language models into sparse architectures. We also discuss the remarkable findings from the LucaProt model, which has uncovered over 251,000 new RNA viruses, reshaping our understanding of viral diversity. Additionally, we cover the US's historic UN resolution aimed at ensuring equitable access to AI technology globally. Join us as we delve into the implications of these developments for science, technology, and global equity.
Sources:
https://www.marktechpost.com/2024/10/14/google-deepmind-research-introduces-diversity-rewarded-cfg-distillation-a-novel-finetuning-approach-to-enhance-the-quality-diversity-trade-off-in-generative-ai-models/
https://www.marktechpost.com/2024/10/13/nvidia-ai-researchers-explore-upcycling-large-language-models-into-sparse-mixture-of-experts/
https://www.news-medical.net/news/20241013/AI-model-LucaProt-uncovers-251000-new-RNA-viruses-revealing-hidden-diversity-worldwide.aspx
https://apnews.com/article/un-artificial-intelligence-resolution-rules-governance-goals-b442f2701139780526b34ba0527e9425
Outline:
(00:00:00) Introduction
(00:00:44) NVIDIA AI Researchers Explore Upcycling Large Language Models into Sparse Mixture-of-Experts
(00:03:32) AI model LucaProt uncovers 251,000 new RNA viruses, revealing hidden diversity worldwide
(00:06:43) US spearheads first UN resolution on artificial intelligence