NVIDIA (NASDAQ: NVDA) announced a collaboration with Ineffable Intelligence to develop infrastructure for large-scale reinforcement learning systems. The London-based AI lab emerged from stealth last week and was founded by David Silver, architect of AlphaGo.

The partnership focuses on building computational infrastructure to support reinforcement learning agents that learn through trial and error rather than from fixed datasets. According to the companies, this approach requires different technical requirements compared to traditional AI training methods.

"The next frontier of AI is superlearners — systems that learn continuously from experience," said Jensen Huang, NVIDIA's founder and CEO. "We are thrilled to partner with Ineffable Intelligence to codesign the infrastructure for large-scale reinforcement learning."

Silver stated that researchers need to develop systems that can discover new knowledge independently. "Researchers have largely solved the easier problem of AI: how to build systems that know all the things humans already know," he said. "But now we need to solve the harder problem of AI: how to build systems that discover new knowledge for themselves."

The technical work addresses challenges specific to reinforcement learning workloads, which generate data during operation rather than using predetermined datasets. The companies said this creates different demands on interconnect, memory bandwidth and serving capabilities.

Engineering teams from both companies are working together to create training pipelines that can operate at scale. The initial work uses NVIDIA's Grace Blackwell platform and will extend to the upcoming NVIDIA Vera Rubin platform.

The collaboration aims to develop hardware and software systems as AI development moves beyond training on human-generated data toward models that learn through simulation and experience.