Revolutionizing AI: Intel’s LoNAS and OpenAI’s Free Model Release

Simply AI

Explore Intel Labs' innovative LoNAS framework for efficient LLM compression and OpenAI's strategic release of the o3-mini model for free. Delve into the implications of MLCommons and Hugging Face's massive speech dataset and DeepSeekAI's RAGEN framework for agentic AI models. This episode highlights transformative advancements and ethical challenges in AI, offering insights into the future of technology and its societal impacts.
Sources:
https://www.marktechpost.com/2025/01/31/intel-labs-introduces-lonas-a-hybrid-approach-combining-low-rank-adapters-and-neural-architecture-search-for-efficient-llm-compression/
https://techcrunch.com/2025/01/31/mlcommons-and-hugging-face-team-up-to-release-massive-speech-data-set-for-ai-research/
https://www.theguardian.com/business/2025/jan/31/openai-to-release-new-artificial-intelligence-model-for-free
https://www.marktechpost.com/2025/01/31/meet-ragen-framework-the-first-open-source-reproduction-of-deepseek-r1-for-training-agentic-models-via-reinforcement-learning/
Outline:
(00:00:00) Introduction
(00:00:43) Intel Labs Explores Low-Rank Adapters and Neural Architecture Search for LLM Compression
(00:03:55) MLCommons and Hugging Face team up to release massive speech dataset for AI research
(00:06:57) OpenAI to release new artificial intelligence model for free
(00:10:31) Meet RAGEN Framework: The First Open-Source Reproduction of DeepSeek-R1 for Training Agentic Models via Reinforcement Learning