This week, hosts Conor Bronsdon and Atindriyo Sanyal discuss the fallout from DeepSeek's groundbreaking R1 model, its impact on the open-source AI landscape, and how its release will impact model development moving forward. They also discuss what effect (if any) export controls have had on AI innovation and whether we’re witnessing the rise of “Agents as a Service”. To tackle the increasing complexity of agentic systems, Conor and Atin highlight the need for robust evaluation frameworks, discussing the challenges of measuring agent performance, and how the recent launch of Galileo's agentic evaluations are empowering developers to build safer and more effective AI agents.Chapters:00:00 Introduction02:09 DeepSeek's Impact and Innovations03:43 Open Source AI and Industry Implications13:44 Export Controls and Global AI Competition18:55 Software as a Service19:29 Agentic Evaluations 25:14 Metrics for Success31:34 Conclusion and FarewellFollow the hostsFollow ⁠⁠Atin⁠⁠Follow ⁠⁠Conor⁠Follow VikramFollow ⁠⁠Yash⁠⁠Check out GalileoTry GalileoShow NotesOn DeepSeek and Export ControlsIntroducing Agentic Evaluations

Ask this episode

“What did the council decide on item 42?” — answered with the speaker and the moment.

Transcript

Related