AI Confidential Building transparent, open-source AI with Sriram Raghavan
Apr 22, 2025
Sriram Raghavan, Vice President of IBM Research AI, leads a global team focused on advancing AI with a commitment to open-source principles. He discusses IBM's contributions to the Linux Foundation, emphasizing smaller, fit-for-purpose models instead of large language models. The conversation touches on the importance of transparency in AI, the challenges of regulatory compliance, and the exciting potential of innovations like the Model Context Protocol. Sriram highlights how responsible AI practices can enhance enterprise efficiency while fostering community collaboration.
AI Snips
Chapters
Transcript
Episode notes
IBM's Open AI Platform Vision
- IBM commits to open source for AI platform technology to support diverse needs and environments.
- They contributed Data Prep Kit, Docling, and BeeAI to Linux Foundation for community innovation.
IBM's Open Source Model Principles
- Open source models should be transparent with data bills of materials and permissive licenses.
- IBM's Granite models follow this with Apache 2.0 license and various sizes for developer choice.
Three Axes of AI Innovation
- Innovation in AI will span model architectures, multi-agent composition, and new data modalities.
- Domain experts and open collaboration are key for advancing diverse application areas.