10172343A Novel Dynamic Wave Reservoir Architecture for High-Performance Low-Power Generative AIClosedFast Start ResponseInnovate UKArtificial intelligence is changing our world. Today, users with no prior experience of programming or knowledge of AI can hold fluent conversations in ordinary language with computers using systems such as ChatGPT (Open AI), Gemini (Google), and Claude (Anthropic). These and other 'generative AI' systems are used by millions of people worldwide to perform many personal and business tasks such as generating, summarising or translating documents, making plans, checking regulations and laws, generating sounds, images, and videos, and many others. They are amplifying the capabilities of people in a wide range of occupations to perform their work in much less time. They are able to respond creatively to a wide variety of personal questions and requests for ideas. This new technology is game-changing but has its drawbacks. These include the need for powerful and expensive computational hardware restricting the market to tech giants outside the UK such as OpenAI, Anthropic, Google, Meta and Tesla, creating inflexibility for UK developers. Also the prevailing architecture today consumes unsustainable amounts of electricity and water. Nonetheless, the use of the technology is expected to increase significantly, _e.g._ the UK Government expects this technology to be applied widely across the public and private sectors with a significant impact on economic growth. Vision Scientific Ltd has created a new computational architecture that can revolutionise future machine learning and generative AI by providing new kinds of chips that work differently to conventional AI chips. This technology can deliver more focussed computational power, using much less electricity and water, with a more flexible architecture creating new opportunities for UK businesses and entrepreneurs. Our new _Wave Reservoir Computing Architecture_, or _WaveRCA,_ challenges the conventions of rigid neural network architectures by a fluid, flexible, and symbolically rich model, where signals do not merely activate through static layers, but instead propagate, interact, and resonate within a dynamic reservoir inspired by physical principles of self-organisation. This new _WaveRCA_ architecture can be implemented in compact low-cost chips embedded in a wide variety of applications, enabling new generations of products and systems. This Project will develop a _WaveRCA_ demonstrator and benchmark data to enable potential partners, funders and UK semiconductor companies, to evaluate the technology. This is a major step towards the investment needed to take the new technology to market. It will make the UK a major player as a producer of AI hardware and software over the next decade.