Interview with Professor Adam Sobey

Loh Yee Wei: Good day everyone. In this episode of the SMI Horizon, we are pleased to have with us Professor Adam Sobey. Prof Adam is the Programme Director for Data Centric Engineering at the Alan Turing Institute, the UK’s national AI and data science institute, and Chair of Data Centric Engineering in the Maritime Engineering Group at the University of Southampton.

Prof Adam, can you share with the audience on what are the broad areas of research that you are undertaking?

Professor Adam Sobey: Yes. I am Programme Director for Data Centric Engineering at the Alan Turing Institute, and within that we are looking at how data science, machine learning and artificial intelligence can change engineering.

We have always used data within engineering dating back to the Egyptians, where we used empirical changes to model the structures of pyramids and ensure their integrity. But recently there’s probably been a change in the way in which we use that data. We have access to much larger quantities of data than ever before. We understand what’s happening in space and time across weather from satellites and have much greater regions than we have had previously. We have increasing sensors on our structures and we have access to compute at larger scales than we’ve ever had before. This is allowing us to generate new theories and approaches to engineering that rely more on this data and to generate new business models and new products that we haven’t seen before, as well as the necessary skill changes that we’re seeing.

So our program is trying to look at all of these elements from early stages research through to innovation and skills updates for our young engineers, as well as those with continuing professional development.

Loh Yee Wei: Great. The Turing Institute is a member of the ADViCE project under the UK’s government for AI for Decarbonisation Innovation Programme to support the UK’s transition to Net Zero. ADViCE focuses on sectors such as agriculture, manufacturing, energy and built environment. Can you talk on the key insights that could be relevant and applicable to the maritime sector, both in the UK and by extension globally?

Professor Adam Sobey: We have done 2 main pieces of work in this area. One is stakeholder engagement, where we’ve looked at all the different groups who could benefit from AI within the UK within these sectors. The second is around road-mapping the different technologies and how we see them being implemented within decarbonisation.

I can see maritime benefiting from both of these, both through sharing with the stakeholders we already have engaged to benchmark where we are as an industry in terms of our AI development and to understand the problems that other sectors are having in their implementation of AI. These are very similar to what we’re seeing in maritime and therefore we can share more broadly.

Also in terms of the technology roadmap, I think you’ll find that a lot of the work done in other sectors is very similar to that in the maritime sector. Therefore there’s the potential to share learnings, share research costs and share the product development by bringing everyone together on this decarbonisation journey.

Loh Yee Wei: Great to hear all this very exciting developments within ADViCE.

I understand that the UK Government recently launched an £8 million Smart Shipping Acceleration Fund to drive maritime innovation and harness AI to boost productivity in the sector. I believe this fund requires match funding from the private sector as well. The Turing Institute also has notable collaborations with Accenture and Lloyd’s Register in various projects, including joint research.

Can you share some insights as to how the Turing Institute, along with the agencies like Innovate UK, foster collaborations between academics, startups and the established industry players?

Professor Adam Sobey: I think it’s a difficult problem and if it had been solved, we probably wouldn’t still be talking about it. Each one of these journeys is a little bit different. They all start at different places and with different questions and so I think it’s around some flexibility in the models.

My preference, I think, and where we’ve had most success is around those longer term collaborations where you work together over a period of time. That doesn’t automatically mean a single funded project, although that can be the case, but it means that you are constantly talking to those industry partners. It means the academics really understand the challenges that the industry’s facing and that’s really important. It means that they can pull in research that’s funded elsewhere to support that company, and it means that the company starts to really understand the technology development.

We have such different timelines between these two different industries. When you look at the sort of industrial side of things that the maritime sector, they want answers in six months, whereas for a Research Institute or for a university, those are very short timelines. Much of the research that I am now implementing into industry was originally developed maybe 10 years ago. So you have very long sort of cycles of research with a burst of innovation perhaps at the end. So the longer you can continue those collaborations and find ways to do that, I think the more successful you will be.

We have been really lucky at the Turing Institute to have a partner like Lloyd’s Register Foundation who have given us continuous funding for nine years with a further funding for another 8. So we have a really good platform to provide a sort of center, a stability of core research that we’re then able to work with our industry partners and to continue to bring them along this journey. They don’t need to fund the entire partnership, but they can work with us continuously and start to understand how we’re evolving our research, while we get to understand their problems and continue to keep that in mind as we generate new research projects.

Loh Yee Wei: Thanks Prof Adam. UK is also positioning itself as a leader in safe AI systems and their deployment. In the field of IT and software engineering, we see AI being used for software testing and quality assurance. In the context of maritime, we are perhaps concerned less with the ethical issues currently, but with more of the assurance of AI technology services – how to enable reliable and robust AI applications to achieve operational goals such as safety and efficiency for ports and shipping. Could you perhaps share some of the current research in the UK that you are familiar with that probably helps the notion of ‘safe AI’ and ‘autonomy assurance’, and what are the key challenges?

Professor Adam Sobey: I think AI safety is such a large topic. A lot of what you are seeing in the news is very much focused around new developments in things like ChatGPT, around the issues we might see with democracies and the influence it might have on society more generally. I think you are right in terms of the engineering focus being slightly different to that.

I really like the term ‘assurable AI’ rather than ‘safe AI’ because I think it is really clear what that means that we have tests, that we can perform and that we can go through that process. I think it is very interesting that you see computer science for the first time not being able to validate their code with a sort of step by step process. We are now in a sort of soft computing world where that is more challenging. We don’t understand quite what’s under the hood, but that’s an area that engineers have always lived in. You never understood the product because it’s designed one way, it’s built differently to how it’s designed and then it’s operated for 20 or 30 years without you being able to see inside it. So I think we’re much more comfortable in that respect around how you would assure and use something which is a bit more black box.

So I think there’s a number of processes we can use for this. There’s human in the loop testing, which is very expensive, but we will need that in various areas, especially cybersecurity. Hardware in the loop testing, again where we can take these systems and start to understand them in isolation, perhaps in simulation approaches, or in different test facilities. Then software in the loop testing, which is something that we are trying to generate as much as possible where we can develop software to test and approve these different approaches. We have actually done the first assurance of a digital twin in the maritime industry two years ago using this sort of approach, and demonstrated that it’s actually quite a usable element. The difficulty comes in the sheer number of different test cases. You might see, for example, on autonomous system, if you think of 1 ship on one ship, it’s quite simple. There’s not too many variables, but as soon as you add a 3^rd, 4th or 5th ship in that area, it becomes really difficult. There is a huge number of tests you need to perform to ensure safety, and how we go around doing that in a way that we can afford with computational expense is a difficult problem.

Loh Yee Wei: Thank you, Prof Adam for the fantastic kind of like insights to the things of assurance AI. In 2014 when you were a Lloyd’s Register’s Educational Trust funded Research Fellow, where you were 50% of the time seconded to the Institute of High Performance Computing (IHPC) in Singapore, where you actually worked on novel algorithms in evolutionary computation. Could you describe a bit of your experience back then as a researcher in Singapore? SMI is eager to explore potential collaborations with the UK, particularly in AI. Any suggestions as to how we might develop such research relationships?

Professor Adam Sobey: So I was out here from the University of Southampton lab based in the Institute of High Performance Computing, and I really enjoyed the time out here. I thought the Singaporean ecosystem for research was just incredible. There are so many great researchers here and they attract so many great researchers from outside of Singapore that there’s this constantly changing ecosystem of top quality researchers discussing different ideas that you just don’t see outside. I think with a small number of universities, there is a great focus and a great community. Everyone knows each other and it’s very easy to go and talk to each other and have those discussions. So I think there’s a lot for us to benefit by working in Singapore from the UK.

From my experience, I think that if we have things that are mutually beneficial to Singapore and the UK, and we can find researchers who are attracted by those questions, then there’s a very great chance of success, and it would be a really great opportunity for us, certainly in the UK, to learn more about the great AI maritime research that is already occurring here in Singapore.

Loh Yee Wei: I think there are definitely a lot of areas where Singapore and UK are alike. I think there are, in terms of the stakeholders, is very similar in terms of direction. Definitely it’s a really great place where researchers from the UK, hopefully with Alan Turing where you are today, it’s a fantastic melting pot in that sense. So I’m kind of hopeful for the future for collaborations with like-minded people like yourself, the UK government and hopefully [the] wider maritime ecosystem where definitely there will be different ideas, where there are a lot synergies and there are a lot of sparks and then hopefully grow maritime knowledge and move this industry forward.

I hope you enjoy your trip in Singapore for this time around and hopefully to hear great things from you moving forward. Thank you very much.

Professor Adam Sobey: Pleasure. Thank you.