London, UK – Takara.ai, a leading innovator in Artificial Intelligence, has announced the launch of its latest breakthrough, SwarmFormer. This state-of-the-art AI architecture is inspired by natural swarm intelligence and boasts an impressive reduction in computational resource demands of up to 94%. This positions the UK as a leader in sustainable AI innovation, in line with the government’s AI Opportunities Action Plan.
SwarmFormer: A Nature-Inspired Solution
Drawing inspiration from the collective behaviour of swarming insects, SwarmFormer achieves unparalleled efficiency. By combining local token interactions with cluster-based global attention, it matches the performance of industry-standard models with just 6.7M parameters, significantly less than the 108M parameters in traditional systems. This not only reduces infrastructure costs by 70%, but also allows for seamless operation on consumer-grade hardware, making advanced AI applications more accessible.
Jordan Legg, Chief AI Officer at Takara.ai, said, “SwarmFormer is a game-changer. It showcases the UK’s ability to lead in AI not just through investment, but through innovation. By democratizing access to powerful AI tools, we are enabling organisations of all sizes to harness the transformative potential of AI.”
Pioneering Sustainable AI
SwarmFormer addresses key priorities in the government’s AI strategy, including:
– Sustainable Infrastructure: By reducing energy consumption and computational overhead.
– Democratised Development: By lowering barriers to entry for smaller organisations.
– Homegrown Innovation: By establishing the UK as a global leader in AI research and application.
Technical Excellence with Real-World Impact
SwarmFormer utilises a hierarchical local-global attention mechanism, enabling decentralised multi-hop propagation of information and efficient global context representation. Its cluster-based architecture significantly reduces memory and computational requirements while maintaining exceptional accuracy in text classification tasks. Experimental results have shown SwarmFormer to achieve up to 90% fewer parameters than baseline models like BERT, while surpassing them on key benchmarks.
For more technical details, visit the SwarmFormer insights page on the Takara.ai website.
Derick is an experienced reporter having held multiple senior roles for large publishers across Europe. Specialist subjects include small business and financial emerging markets.