As AI has grown in importance, both from a technological but also an economic perspective, there has been a significant increase in attention given to it by policy makers. Some of these have been designed to ensure that AI develops in a way that benefits all of society, whilst others have attempted to appreciate the importance of AI to a nation and explore ways in which it can be supported.
It’s in this latter camp that a new review by the British government resides. The report touches on a number of areas, including closing the skills gap that exists in AI, the efficient transfer of AI research from lab to market and the various steps required to encourage the uptake of AI.
The section that most intrigued me however was around data. The authors rightly reveal that the availability of data, alongside a tremendous increase in computing power, has played a significant role in the enhanced capabilities of AI in various areas.
Improving access to data
The report highlights the crucial role the government can play in opening up its data so that AI developers can start to create applications using this data. There has been undoubted progress made in this field, but an enormous amount is still to be done, not least in healthcare where data is often analog, siloed and unavailable. The report mentions a number of possible reasons for this, but in reality it is largely due to political and organizational factors that massively inhibit the potential in this field.
It quotes executives from Your.MD in identifying the need for the government to grasp the nettle on this and develop the kind of data marketplace that would enable so much AI research and development, and subsequently so much better and more targeted healthcare. Alas, such protestations are not new and there seems little indication that the government has the vision to make this happen.
This haphazard approach to making data available for AI research has led to obvious shortcomings with regards to the privacy and governance of individual data. Again, this isn’t a new issue, and data governance was highlighted in the recent report into AI by the Royal Society.
Finally, the report underlines the importance of making access to that data frictionless. This won’t be the case so long as legal and procedural costs remain high, and an active data marketplace requires the cost of each ‘transaction’ to be as low as possible, especially if we’re to encourage smaller businesses to participate as much as larger ones.
Data trusts or data commons?
To overcome some of these challenges, the authors advocate the creation of Data Trusts to support secure and mutually beneficial exchanges of data. Such trusts would provide a common framework for two parties to agree to share data. This would be supported by a body, known as the Data Trusts Support Organisation (DTSO), to develop tools and guidance to support the sharing of data.
Such a framework would be undoubtedly an improvement on what we already have, but I feel it would fall some way short of a full and functioning data commons that would provide a fluid marketplace for the exchange of data. Such a marketplace would work towards returning ownership of data from the institutions that currently own it back to us as individuals. It would still provide a framework for the sharing (and monetization) of that data, but would enable it on a one-to-many, and indeed many-to-many basis rather than simply bilateral arrangements.
I fully agree with the report that an independent body would then be tasked with ‘refereeing’ this marketplace to ensure that rules are kept and the system continues to operate securely and fairly.
The technology exists, and indeed has existed to provide such a marketplace for some time now, especially with the advances in secure and transparent ledger technologies such as blockchain. The obstacles have been, and remain political in nature. To bring such a data commons about requires a degree of boldness that has been sorely lacking in politics for some time. Whilst the intentions, and indeed the recommendations, of this report are commendable, it’s not immediately clear how that leadership question will be overcome.