Episode SummaryIn this episode, Maribel Lopez interviews Kate Soule, Director of Technical Product Management for IBM's Granite products. They discuss IBM's third-generation AI models, their focus on efficiency and enterprise readiness, and the latest advancements including vision capabilities and reasoning features.GuestKate Soule - Director of Technical Product Management for IBM's Granite productsKey Topics & Timestamps00:04 - IntroductionMaribel introduces the show and Kate SouleBrief overview of IBM Granite as fit-for-purpose, open-source enterprise AI models00:48 - What is IBM Granite?Designed as core building blocks for enterprises building with generative AIFocus on efficiency with smaller model sizesMonthly innovation updates to keep pace with rapidly evolving field02:19 - Understanding AI ReasoningExplanation of reasoning capabilities in AI modelsHow allowing models to generate more text at inference time can improve performanceCost/benefit tradeoffs of reasoning features03:13 - Enterprise AI Model Selection CriteriaMoving beyond "one model to rule them all" thinkingImportance of fit-for-purpose modelsWhy smaller models can be customized more easilyTrust and transparency considerations05:38 - AI Governance and SafetyHow to evaluate models for governance requirementsSafety evaluations and benchmarks as table stakesSystems-based approach to safety with guardrailsIBM's Granite Guardian and protection mechanisms08:55 - Benefits of Smaller ModelsWhy size matters: cost, latency, and customization advantagesSmaller models are easier to customize and require less computing powerIBM's transparent approach to training data10:13 - Future of AI EvaluationPerformance per cost becoming the key evaluation metricThe growing importance of flexibility in model selectionHow the "efficient frontier" between cost and performance will differentiate providers12:41 - IBM's Vision ModelsIBM's pragmatic enterprise focus for multimodal capabilitiesVision understanding (image in, text out) for practical business use casesSpecialization for documents, charts, and dashboardsDelivering powerful capabilities in only 2 billion parameters15:25 - Understanding Model Size ContextEvolution from millions to billions of parametersPractical considerations of deploying different-sized modelsFinding the right cost-benefit trade-off for specific use cases

Ver todos los episodios