AI for Chemistry: AIchemy

点击翻译按钮获取中文摘要

Chemistry impacts most areas of our lives, including healthcare, energy production, and the environment. It is also the UK's second largest manufacturing industry, employing 140,00 people. This hub will bring the transformative power of artificial intelligence (AI) to the area of chemistry, and by doing so have a major societal impact. Both AI and chemistry are fast-moving and historically separated research disciplines, and there is huge untapped potential to collaborate at the interface of these two fields. Today, relatively few UK experimental chemists are exploiting AI (e.g., for reaction optimization), and few have corresponding automation facilities to do this, which is a missed opportunity. The use of machine learning methods is more common in computational chemistry, but here also we are often data poor, and data is sparse. In some AI fields, such as natural language processing, there is also rapidly evolving, leading-edge industrial research, necessitating a cross-sector approach if we are to exploit the cutting edge of this technology.This hub (AIchemy) will bring together leading researchers in AI and trailblazers at the interface of AI for chemistry, spanning both university and industry. We will exploit unique established facilities and institutes in the four core partner institutions (Universities of Liverpool, Imperial, Cambridge, and Southampton) where cross-discipline working has already been achieved: this includes the Materials Innovation Factory (MIF), the Institute for Digital Molecular Design and Fabrication (DigiFAB), and the I-X Centre for AI in Science. In addition to the 6 lead investigators, we have aligned 25 other investigators across nine institutions, spanning the areas of AI, robotics, and a diverse range of experimental and computational chemistry sub-disciplines, and career stages. The team also includes unique expertise in robotics and automation (Liverpool & Imperial), natural language processing for chemistry problems (Cambridge) and data curation in the Physical Sciences Data Infrastructure (PSDI, Southampton). This diverse team and associated facilities give us the breadth of expertise and critical mass to become the core of a UK hub for this activity. AIchemy will carry out world-leading research at the AI/chemistry interface, building on distinctive UK strengths in this area and developed initially via 6 Forerunner Projects. The Hub will also build an approach for sharing chemistry research data and code in a common format to unite the currently fragmented UK research landscape. We also aim to dramatically broaden the number of AI researchers tackling chemistry problems, and vice versa, through a mixture of pump-priming funding in the hub, bespoke training, access to datasets, and events (e.g., AI challenges using hub-generated data). To ensure the long-term health of this discipline, we will also focus resource on projects that are led by early career academics.The hub will build a UK-wide consortium involving university and industry stakeholders outside of the core partners, including a broad set of 15 day-one industry partners across the sectors of AI and chemistry, to be further expanded in the full proposal. The team has an excellent collective track record in industry engagement and knowledge transfer; e.g. MIF collocates 100 industry researchers in a common facility with academics; Chemistry is co-located with IX at Imperial's £2 Bn White City campus, and there are shared spaces to enable 800 scientists and industry partners to work together on common challenges, with tailor-made labs and offices for early stage companies. Mirroring the enormous benefits that have been achieved in other science areas, such as structural biology, this hub will transform the UK landscape for the discipline of chemistry, transforming engagement with AI from a relatively niche activity to a core, platform methodology.