Why data spaces are key to unlocking AI’s potential

AI and data share an inseparable bond, raising important questions about their relationship and data quality's role. What drives AI's need for massive amounts of high-quality data? Why does Europe champion collaborative models like data spaces to feed AI's growing data appetite? Let's explore these questions.

by Gianfranco Cecconi

What we commonly call AI is a group of different algorithms designed to simulate or replicate what we consider “intelligence”. While these algorithms often seem magical to the uninitiated, they primarily process massive amounts of data to identify patterns or properties, e.g. large language models (LLMs) mimic how humans use language.

In simple terms, AI is only as good as the data it processes. Feed it poor or biased data, and you’ll get poor and biased results. For instance, if a generative AI system were trained solely on the speeches I’ve given in my less-than-perfect English, it wouldn’t produce eloquent content like that of Oprah Winfrey or Winston Churchill – it would just replicate me.

Collaborative models matter

Generative AI’s rise, led by products like OpenAI’s ChatGPT and Microsoft’s Copilot, exacerbates these systems’ enormous data needs. Current AI engines train on datasets that can be purchased or found on the public Internet, but these sources have limits. What happens when they are exhausted?

Traditional methods of sourcing data are nearing their limits. Research suggests that accessible data sources may be exhausted by next year. Moreover, the practice of indiscriminate data collection has led to a backlash, with previously open datasets now being restricted.

Many other valuable datasets remain untapped because they aren’t strictly for sale. These are datasets that individuals or organizations may be willing to share under specific conditions:

• Data sovereignty: They want control over their data rights and usage consent
• Trust: They need confidence in the data requester’s identity
• Value exchange: They may look for benefits that go beyond money, like partnership opportunities

This is exactly what data spaces can enable – thanks to the Dataspace Protocol. It’s a standardized set of rules for secure and interoperable data sharing. To participate in a data space, organizations use a data space connector: a software component that applies these rules.

Think of it like email: the email protocols define how messages should be formatted and transmitted, while email servers and clients are the software that actually implements these protocols to send and receive messages. Similarly, the Dataspace Protocol defines the rules for sovereign data sharing, while data space connectors are the software that puts these rules into practice.

How the Dataspace Protocol works

So, how does the Dataspace Protocol help us with AI? It addresses those three key requirements I’ve touched upon earlier:

• Data sovereignty: The Dataspace Protocol puts data providers in control by allowing them to set clear rules about how their data can be used. Through detailed usage policies and agreements, organizations can confidently share their data while maintaining oversight of its use.
• Trust: The Dataspace Protocol creates trust through standardized processes and integrating with other technologies dedicated to managing securely digital identities and claims. Just as you trust a bank to handle your money because of established banking protocols and practices to verify the identity of people and businesses they’re doing business with, the Dataspace Protocol ensures all participants follow consistent rules for identity verification and agreement negotiation. This means organizations can be confident about who they’re sharing data with.
• Value Exchange: Beyond simple buy-sell transactions, the Dataspace Protocol enables more sophisticated forms of collaboration. Organizations can create agreements that deliver mutual benefits – whether that’s joint research projects, operational efficiency across different organizations, enabling shared AI models, or other forms of innovation.

The International Data Spaces Association leads efforts to establish the Dataspace Protocol as an international standard. This standardization will create new opportunities across industries through a common data-sharing framework.

Data spaces provide a sustainable alternative, unlocking new data sources while ensuring data sovereignty, trust, and fair value exchange. This approach doesn’t just address the immediate data needs of AI but also aligns with ethical and collaborative principles.