数据：英国人工智能初创公司的生命线

Author Vinous AliDeputy Executive DirectorStartup Coalition AboutStartup Coalition The Startup Coalition is an independent advocacy group that serves as the policy voice for the UK’stechnology-led startups and scale-ups. We were founded in 2010 by Mike Butcher, Editor-at-Large ofTechCrunch, and Jeff Lynn, Chairman and Co-Founder of online investment platform Seedrs. We fightfor a policy environment that enables early-stage British tech companies to grow, scale, and competeglobally. Our network includes over 4000 startups, scale-ups, and investors. We have been instrumental inbuilding proactive coalitions of businesses and investors on issues integral to the health of the UK’sstartup ecosystem. Our work has seen many successes, from establishing the Future Fund to creatingthe Scale-up visa. Acknowledgements We are grateful to all the startup founders who contributed to this report and supported us by providingtheir time, insight and expertise. Introduction ThePrime Minister himself described AI as “the defining opportunity of our generation”1 with theGovernment’s own sources citing that AI can boost the UK’s economy by 1.5 percentage points inproductivity and be worth up to an average of £47 billion to the UK each year over a decade.2However,the window to capitalise on this opportunity is closing fast, and the UK must act swiftly and decisively tostay ahead. It is generally accepted that the progress and capability of AI stands on three legs: the ability to accessand use data, compute and talent. While all three are indispensable, data plays a foundational and highlycritical role in the development of AI. Without the ability to access and use data, AI models cannot learn,identify patterns, make predictions or improve their performance. Startups, especially, rely on publicly available datasets to innovate. They often aren’t able to leverageexistingproprietary data nor have the resources to engage in data partnerships.Therefore,democratisingaccess to usable data levels the playing field helping startups to compete withincumbents. But where is this data coming from? Much has been written about models trained on data that is publicly available online, but what is clear isthat today, much of the world’s data, particularly high-quality datasets, are unstructured, inaccessible orunusable to researchers and developers alike. The National Data Library (NDL) represents a uniqueopportunity for the UK to unlock siloed public sector datasets - and those that sit in arms length bodies(ALBs) - to give UK startups and researchers a competitive edge in developing the next generation of AIsystems. At Startup Coalition, we believe the National Data Library could supercharge discovery andsupportmission-led Government by finding new pathways to solving some of the biggestchallenges facing society today. But with so much potential data to draw upon, sensitivities at every turn and IP and ownership issues,success is not guaranteed. Described as a ‘ten-year’ or ‘two-term’ project, it is clear that the Governmentis in it for the long haul. However, quick wins could put the NDL on the map and encourage greaterengagement from even the most sceptical. What follows is a roadmap for the National Data Library outlining three actionable workstreams toestablish the NDL as a valuable interlocutor in the journey to AI leadership. TheCritical Role of Data Artificial Intelligence fundamentally depends on data. Classic machine learning techniques refine largedatasets, and modern large language models learn by processing vast amounts of information to identifypatterns, correlations and concepts across datasets. Scale as well as variety is key, directly impacting amodel's performance, accuracy, and ability to generalise to new, unseen data. Neglecting either can leadto suboptimal or even failed AI deployments. While frontier models, predominantly trained outside of the UK, have trended upwards in needing moretraining data over time. Therefore, the UK has an opportunity to seize on the opportunity higher up thestack, leveraging the pre-trained models and fine-tuning them. The UK is well positioned to accrue value at this layer in the stack given the following strengths: 1.A robust R&D ecosystem:The UK boasts world-class universities such as Oxford, Cambridge,andImperial College London,which are consistently at the forefront of AI research. Thisacademic excellence creates a strong pipeline of highly skilled AI talent. Institutions like the AlanTuring Institute further bolster national AI capabilities through research and collaboration.2.Government commitment and investment:One of the first acts of this new Labour governmentwas to commission what became the AI Opportunities Action Plan. This has been backed by£2bnin cash in the Comprehensive Spending Review with money going to build out AIinfrastructure (including increasing compute capacity at least 20x by 2030) and investing in AI skills. These strategic investments create a fe