Supporting Our AI Overlords: Redesigning Data Systems to Be Agent-First
6 hours ago
- #LLM Agents
- #Data Systems
- #Artificial Intelligence
- LLM agents are expected to dominate future data system workloads by acting on users' behalf.
- Agentic speculation involves high-throughput exploration and solution formulation, posing challenges for current data systems.
- Data systems need to evolve to natively support agentic workloads.
- Key characteristics of agentic speculation include scale, heterogeneity, redundancy, and steerability.
- Research opportunities include new query interfaces, processing techniques, and agentic memory stores.