The “Mixture of Experts” podcast host Tim Hong and experts Kate Soule, director of technical product management for Granite, Chris Haye, distinguished engineer and CTO of customer transformation, and Aaron Botman, IBM fellow and master inventor, discuss DeepSeek R1, a new open-source model, with the experts offering differing opinions on its significance. The conversation explores the accuracy of claims about the model’s training costs, the efficiency gains it demonstrates, and the role of reinforcement learning (RL) and chain of thought reasoning. The experts also discuss model distillation, its implications for competition in AI, and how it may change the incentive for companies to invest in large language models. Finally, the panel analyzes the responses of other tech companies like OpenAI and how DeepSeek may impact the competitive strategies of other AI companies.
Subscribe to the IBM YouTube channel at https://www.youtube.com/@IBMTechnology.
Subscribe for AI updates from IBM at https://ibm.biz/BdGKrT.