Shortage of high-quality data threatens the AI ‘boom’

A number of fundamental issues, including a shortage of high-quality data with which to ‘train’ the technology is threatening the AI ‘boom’, according to a new white paper from the Open Data Institute. The paper Building a better future with data and AI is based on research carried out by the Institute in the first half of 2024. It claims to identify significant weaknesses in the UK’s technological infrastructure that threaten the predicted potential gains – for people, society, and the economy – from the surge of interest in artificial intelligence and its applications. It also outlines the ODI’s recommendations for creating diverse, fair data-centric AI.

Based on its research, the ODI is calling for the new government to take five actions that will allow the UK to benefit from the opportunities presented by artificial intelligence while mitigating potential harms:

1. Ensure broad access to high-quality, well-governed public and private sector data to foster a diverse, competitive market;
2. Enforce data protection and labour rights in the data supply chain;
3. Empower people to have more of a say in the sharing and use of data;
4. Update our intellectual property regime to ensure models are trained in ways that prioritise trust and empowerment of stakeholders;
5. Increase transparency around the data used to train high-risk models.

The ODI’s white paper argues that the potential for emerging AI technologies to transform industries such as diagnostics and personalised education shows great promise. Yet significant challenges and risks are attached to widescale adoption, including – in the case of generative AI – reliance on a handful of machine learning datasets that ODI research has shown lack robust governance frameworks.

The paper argues that this poses significant risks to both adoption and deployment, as inadequate data governance can lead to biases and unethical practices, undermining the trust and reliability of AI applications in critical areas such as healthcare, finance, and public services. It claims that these risks are exacerbated by a lack of transparency that is hampering efforts to address biases, remove harmful content, and ensure compliance with legal standards.

To provide what it says will be a clearer picture of how data transparency varies across different types of system providers, the ODI is developing a new ‘AI data transparency index.’

Sir Nigel Shadbolt, Executive Chair & Co-founder of the ODI, said, “If the UK is to benefit from the extraordinary opportunities presented by AI, the government must look beyond the hype and attend to the fundamentals of a robust data ecosystem built on sound governance and ethical foundations. We must build a trustworthy data infrastructure for AI because the feedstock of high-quality AI is high-quality data. The UK has the opportunity to build better data governance systems for AI that ensure we are best placed to take advantage of technological innovations and create economic and social value whilst guarding against potential risks.”

Before the General Election, Labour’s Manifesto outlined plans for a National Data Library to bring together existing research programmes and help deliver data-enabled public services. However, the ODI says that first, we need to ensure the data is AI-ready.

The ODI believes that, as well as being accessible and trustworthy, data must meet agreed standards, which require a data assurance and quality assessment infrastructure. The ODI’s recent research suggests that currently – with a few exceptions –  AI training datasets typically lack robust governance measures throughout the AI life cycle, posing safety, security, trust, and ethical challenges related to data protection and fair labour practices.

Other claims from the ODI’s research include:

• The public needs safeguarding against the risk of personal data being used illegally to train AI models. Steps must be taken to address the ongoing risks of generative AI models inadvertently leaking personal data through clever prompting by users. Solid and other privacy-enhancing technologies have great potential to help protect people’s rights and privacy as AIs become more prevalent
• Key transparency information about data sources, copyright, and inclusion of personal information and more is rarely included by systems flagged within the Partnership for AI’s AI Incidents Database.
• Intellectual property law must be urgently updated to protect the UK’s creative industries from unethical AI model training practices.
• Legislation safeguarding labour rights will be vital to the UK’s AI Safety agenda.
• The rising price of high-quality AI training data excludes potential innovators like small businesses and academia.