NVIDIA Backs Ineffable to Build Next-Gen Intelligent Systems

According to reports on May 13, semiconductor and AI industry leader NVIDIA has announced a strategic engineering collaboration with a newly established artificial intelligence startup focused on superintelligence research, Ineffable Intelligence.

Founded in late 2025 by University College London professor and former head of the reinforcement learning team at Google DeepMind David Silver, Ineffable Intelligence is positioning itself at the frontier of next-generation AI systems. Despite being only a few months old, the company has rapidly attracted significant industry attention due to its ambitious focus on “super-learning” systems capable of autonomous improvement through trial-and-error experience.

The company confirmed that it will engage in deep technical collaboration with NVIDIA at the engineering level, aiming to build advanced reinforcement learning infrastructure that enables AI systems to learn continuously from experience rather than relying primarily on pre-existing human-generated datasets.

In April, Ineffable Intelligence completed a record-breaking $1.1 billion seed funding round. The round was co-led by Sequoia Capital and Lightspeed Venture Partners, with participation from NVIDIA, DST Global, Index Ventures, Google, and the UK sovereign AI investment fund. The scale and investor profile of the funding highlight growing institutional interest in reinforcement learning–driven AI architectures.

NVIDIA CEO Jensen Huang emphasized that the next phase of artificial intelligence development will center on “super learning systems”—AI models capable of persistent self-improvement through accumulated experience. He noted that the collaboration aims to establish foundational infrastructure for large-scale reinforcement learning, enabling a new generation of intelligent systems beyond static training paradigms.

Unlike most current large language models, which are primarily trained on extensive human-curated datasets, Ineffable Intelligence focuses on reinforcement learning approaches that allow models to evolve through interaction with environments and feedback loops. The company stated that its systems will rely on diverse experiential datasets fundamentally different from natural language corpora, requiring novel model architectures and training methodologies.

The partnership will also involve joint engineering teams working on scalable reinforcement learning pipelines. The project is expected to leverage NVIDIA’s latest computing platforms, including the Grace-Blackwell architecture and the Vera Rubin AI computing stack, designed for high-performance AI training and inference workloads.

David Silver highlighted the shift in AI research priorities, stating that much of the “simpler problem” of AI—building systems that capture existing human knowledge—has largely been addressed. The remaining challenge, he argued, is developing systems capable of independent discovery and continuous learning from real-world experience, marking a transition toward truly autonomous intelligent agents.