Intelligence is Free, Now What? Data Systems for, of, and by Agents – The Berkeley Artificial Intelligence Research Blog
https://bair.berkeley.edu/blog/2026/07/07/intelligence-is-free-now-what/
Source Domain: bair.berkeley.edu
Emerging Challenges and Opportunities with Low-cost AI
The reduction in the cost of artificial intelligence is monumental, transforming from $30 per million tokens early in 2023 to below $0.10 today. As per-token costs plummet, agents—software entities driving large datasets—are poised to dominate workloads in data systems. This article delves into three emerging challenges born from these shifts. Namely, data systems need redesigning to accommodate agents better, provide the reliable substrate agents require as they manage tasks and coordination, and synthesize bespoke data systems.
Redesigning data systems, termed “For Agents,” involves optimizing interaction protocols to harness agentic speculation—multiple agents exploring a high-volume, heterogeneous query space. New interfaces must support batch requests and high-level primitives instead of single-query interactions. “Data Systems Of Agents” entail building infrastructure to store pertinent memory and handling failures without livelock. Finally, “Data Systems By Agents” addresses agents’ ability to synthesize custom data systems quickly, though verifying their correctness remains critical. The future will see a co-evolutionary path where agents and data systems are symbiotically intertwined in tasks and infrastructure.
