We’ve detected that you are using an outdated browser. This will prevent you from accessing certain features. Update browser

Podcast interview

The Safety Pioneers Episode 4: Data Engineering for Safety

Improving the way our computer systems and machines make decisions. We hear from Professor Sir Nigel Shadbolt the Chairman of the Open Data Institute and a Professor of Computer Science at the University of Oxford.

Episode transcript

Narrator:

In this episode we focus on the risk and safety challenges surrounding data engineering - the way that information is gathered, analysed, stored and accessed by computer systems and machines. 

This is the foundation for the process by which digital systems make decisions. If we don’t get the data engineering right then the potential for poor, unsafe or biased decision-making will affect all areas of our lives from transport safety, recruitment and credit scores to healthcare and criminal justice.  

Professor Sir Nigel Shadbolt:

The pace and rate of technological change in this field, in my field of data analytics, artificial intelligence is extraordinarily fast, the cycles of development extraordinarily short.

Narrator:

Professor Sir Nigel Shadbolt is Chairman of the Open Data Institute which he co-founded with Tim Berners-Lee, the inventor of the World Wide Web. Professor Shadbolt is also a Professor of Computer Science at the University of Oxford. We caught up with him at a networking event.

Professor Sir Nigel Shadbolt:

The reason that I’m so interested in safety and engineering is because fundamentally my trajectory through engineering has been right from the beginning in AI. Forty years ago I began my research in AI in building decision support systems, since then in all sorts of application contexts from aero to brewing to chemical engineering to the stock exchange. But what is always crucial in these AI apps is the data, and it’s about engineering the data that I’ve been particularly preoccupied, and the work we’re doing with the Open Data Institute and the Lloyds Register Foundation is in that area.

Narrator:

Artificial Intelligence - creating computer systems and machines that work, react and learn like human beings - is of course leading to amazing developments and possibilities for the future, from self-driving cars to improved prediction of flooding. But the more progress we make in this area the more we have to assess and deal with the potential ethical, social, financial and legal risks. 

Professor Sir Nigel Shadbolt:

You could think about the data-engineering aspects of a sector like power generation, maritime, transport safety - all of these areas are increasingly dependent on digital data - that it’s reliable and engineered to a high quality. If that data is not engineered or representative in the way it needs to be is that an ethical use of it? So I do regard the challenge of data for the public good, the challenge of data as infrastructure as a fundamental platform on which you build a safer world in the digital era.

Narrator:

So what could the longer-term prospects be for improving the process of data engineering? 

Professor Sir Nigel Shadbolt:

In a future world I would hope that we’ve taken a view that that is just too important to be left to particular researchers to decide just how they might like to describe their data, in fact I imagine a world in which self-describing data, data itself becomes more intelligent, more capable of describing the properties that it has, the terms of use under which it might be used, and this might sound ambitious, but I think it’s one of the fundamentals we’ve seen often is that to really improve the world of safety we actually inject more information into a domain not less, we inject more information about how the data itself is to be understood and used, the limits, the provenance, where it came from, how recent is it. So that whole world, we call it metadata – the information about the information - is I think an area of engineering that we’ll see really grow in years to come.

Narrator: 

Professor Shadbolt also highlights that the safety of digital systems is being compromised by the obstacles that are in the way of sharing data, due to differences in systems and standards of data analysis.

Professor Sir Nigel Shadbolt:

Take the Internet of Things, take the extraordinary profusion of devices that pervade our everyday working space even our domestic spaces at home. The internet of things of home - it’s a babel, a babel of different languages, different standards and until we get much broader agreement on the standards we can use to adopt and share information, it’ll be much harder to effect regulation for example, to be sure that the information that we want to release is being released in a way that is safe and proportionate. So I think there’s a huge amount to go out in this space and it’s almost as if we’re seeing the enthusiasm for the rapid use of an application outbidding the fact that the underpinning data has to be thought about first of all and that manifests itself in all sorts of challenges we see around people worrying about – is the algorithm fair? It might not be fair if the data it’s being fed was not to the quality we might wish in the first place.     

Narrator:

So inevitably this raises the issue of putting more resources into the way our digital systems are set up and work across different platforms.

Professor Sir Nigel Shadbolt:

I think the area of certification and safety that Lloyds has always pioneered provides some of the best use cases for why this is worth investing in. When you’re designing nuclear reactors, power generation, when you’re thinking about insuring the safety of huge amounts of our shipping and transport infrastructure, you do that on the basis of knowing that you need good data and that you need to be engineering to reliable and reproduce-able standards.

Narrator:

Another issue that needs addressing is how to encourage the sharing of data when so many organisations want to stay one step ahead of their competitors. 

Professor Sir Nigel Shadbolt:

The challenge is to overcome what is sometimes incumbent preferences, so that people have got their own standards and don’t feel they would want to shift, in fact their standards sometimes give then advantage by crowding out other players in this area. So the whole idea of having open, shared, core data assets can feel like particular agencies are giving up their advantage in a field. But I think there’s a completely different way to see it and what we’ve seen in health and safety, in the kind of work that Lloyds does, is that the whole industry, a whole sector can benefit from shared concerns and shared rules of the road, standards, operational laws and formats that they can share together.

Narrator:

The way we gather, use and share data in our increasingly digital world affects every single one of us. When assessing the challenges and addressing the risks, Professor Shadbolt stresses that we still have a long way to go if we are going make the best possible use of digital systems in the future. 

Professor Sir Nigel Shadbolt:

I got into the data engineering concerns that have preoccupied me for a little while now, off the back of my deep enthusiasm for artificial intelligence and advanced computer decision support. The work that people like Tim Berners-Lee and myself did on helping open up more governmental data, showed the pure innovation you can unlock when you make that data to a high quality available. But we also realised that there was still a huge amount of friction in the system, a huge amount of inefficiency to overcome, opportunities lost.

I think if we’re really going to realise the full value of a digitally enabled society, if we’re going to use AI to help humans flourish and not oppress them, we need to get much more on top of the basic standards to engineering data, the equity and access and the ethics that surround its use and regulation, and how we steward it in the first place. So the work that we’ve been doing at the ODI with the Lloyds Register Foundation, that all speaks to these fundamental concerns. 

Hit enter or the arrow to search Hit enter to search

Search icon

Are you looking for?