Ben Telfer: 

Today’s topic, we are looking at “Preparing for catastrophic events using data, technology and machine intelligence to better mitigate and respond to risks”. We’re delighted to be joined by Jeffrey Bohn, who is Chief Research and Innovation Officer at the Swiss Re Institute. Jeff, delighted you can join us today, and I’ll hand over to you. 

Jeffrey Bohn: 

All right. Thank you very much. I’m happy to have the opportunity to talk to you about some interesting work that we’re doing at the Swiss Re Institute at the intersection of data technology and machine intelligence. I should say at the beginning that this does not represent Swiss Re’s policy or any of our actual strategy. This is very much our R&D efforts and we do work together with clients to figure out how to take the research and development efforts and turn them into commercializable efforts and projects. And we look forward to doing that on an ongoing basis. 

Today what I want to talk about is very much on everybody’s mind, and that is how do we think in terms of dealing with what I call the dynamically-changing normal? Often people will say there’s a new normal. I’ve seen new normal. And so, I’ve just given up and I now call it a dynamically-changing normal. I remember when I used to work as a risk, I would have traders and executives ask me, “Why are we seeing one and 10,000-year events every day?” And I think that there are two problems with that question, which I’m going to address as we go through this. One is that the underlying drivers of these risks are just not so stable for us to make meaningful predictions in that way, in a probabilistic way. And secondly, probabilities when we don’t have a lot of historical data really are more like metaphors, and so we really can’t use the same type of modelling capabilities that are developed for risks where we do have some notion of the underlying distribution. 

And 2020 has shown us this path of this dynamics where we have all different kinds of widespread and broad risk events, ranging from wildfires, I’m sure everybody’s tired of hearing about the COVID-19 pandemic, but that’s certainly at the top of everybody’s list. Cyclones, typhoons, and more recently, civil unrest. And one of the trends that started a few years ago in insurance, and this is really driving our research efforts at the Swiss Re Institute, is that insurance as an industry, which historically focused on understanding and covering, is now becoming more predict and prevent. When you think about it in that context, it means that while the insurance product design continues to be quite important to the industry, to the client, what people are asking more about these days are risk management and what we like to call resilience-as-a-service. 

So, this is where we use a broad range of tools to try to understand better what is going to happen in the future and how can we build resilience for that future effort. Now, before I move onto the next slide, I’m actually going to ask a polling question, just first of all to make sure there are actually people listening to me, but secondly, just to understand a little better people’s view. People often talk about black swans and how those come out of nowhere. I actually don’t think the pandemic is a black swan, however, we do have black swans, and these are events that we can’t really foresee and we do have to develop some notion of generic resilience. But just as a warm-up question, please answer, “Which of the following Black Swans” and you’ll just have to take my opinion on these actually being black swans. These were unforeseeable events that had widespread impact or consequences. So, go ahead and make you choice. 

And while that’s happening, I want to distinguish black swans from another animal metaphor, which is a grey rhino. Oh, quite interesting. Rise of the internet. Actually, that’s in line with my own view. This gives you some sense of the importance of technology in how we think about widespread extreme risk generating events. 

Now, getting back to the earlier point that I was making … And sorry. Actually, can you put the second question up? There’s a difference in the risk context between a black swan, which is really unknown and is based on unknown unknowns. But grey rhinos, which is a different metaphor popularized by an author, Michelle Wucker, who talked about high-impact obvious foreseeable events. And these, I think, are more relevant to thinking about resilience-as-a-service going forward and how we think about risk management. So, if you could choice which of the more obvious, obvious in the sense that these are risks that I think most people would agree are going to have widespread consequences. A few of these are in the past, but I think it’s relevant to the discussion today. If you could choose from that list, as well? 

And just while you’re choosing that, an important distinction between grey rhinos and black swans is accountability with respect to people making decisions. I think that you should hold the executives more accountable for the grey rhinos, because these are really things that you should have some notion of them coming along and given that they are foreseeable, you should repair. And interestingly, increasing wealth and equality seems to be the largest choice. And I think that I would agree with that as well. 

So, let’s return to the slides. So, once we have a reframing of this question of risk management, resilience as a service, and just to reiterate, the point I’m trying to emphasize here is that we should be doing all of these things in preparation. Often insurance is focused on after the fact, and this is more about saying, “What can we do with respect to this interconnected risks to develop a better capability for resilience?” And this is what we find to be quite exciting in the insurance R&D arena, because the advent of data, particularly unstructured data, the development of new machine intelligence, enhanced solutions. And let me just stop briefly there to specifically define machine intelligence. Here I mean the whole range of techniques that use machines. It could be conventional curb fitting, something as straightforward as logistic regression or it could be something much more sophisticated or less straightforward, such as deep learning, reinforcement learning, a whole range of some of these new neuromorphic computing techniques and what is often loosely defined as artificial intelligence. 

The point though is less about the complexity of the machine intelligence and it’s more about how that machine intelligence is deployed in the context of actual solutions. Just with respect to the technology, I’m going to ask one more question. So, if you can put up the question three? Just to get a sense of people’s view on which general purpose technology has been the most important, say, in the last 80 years? And while you’re thinking about that or answering the question, let me define specifically. General purpose technology are technologies that have very large impact. You can have technology such as, say, nanotechnology today, which has a very narrow impact. It may end up being a general purpose technology. But there’s many other types of technologies, I’ve listed five here. You get to get your view on which you think is the most important. 

This is also helpful for developing scenarios around the types of risks we should be managing. So, go ahead and make your choice there. Okay. Let’s see the results. Oh, internet. It’s actually quite interesting that the machine learning, artificial intelligence, although that again, as I mentioned before, that reflects a large range of technologies, that that was not number one. I have this argument with a number of people in my network about what is going to end up having more impact with respect to the economy. But I think the internet is definitely a good choice. 

So, general purpose technologies as a back drop, and if we think about the internet as an example there, that’s completely changed the way we interact with each other, it creates new risk drivers, it also creates new opportunities. There are a number of new technologies, which I’m sure most of you are familiar with. But I want to talk about how they can be coordinated together to really develop new resilience-as-a-service. And these technologies are the Internet of Things, the emerging fast networks, such as 5G, a whole range of machine intelligence, and distributed ledger technologies. So, these different new emerging, and maybe not all of these can be called general purpose technologies yet, but I think my forecast is eventually they will, they create opportunities and risks as we think about how to develop new solutions. 

One of the things that’s interesting in this context is that the insurance industry, specifically, has spent a fair amount of money over the last decade, and that’s a bit of an arbitrary time horizon, but quite a bit of money over the last decade investing in different kinds of machine intelligence. I mean, just a few examples are using random forests, which is a certain type of regression technique that’s a little more sophisticated than just linear regression, but using random forests to improve predictability in certain risk blinds. Developing computer vision capabilities to do claims processing. Those are some examples where we’ve seen some promise, but despite the 10 years of investment, sadly, it still has not transformed the industry in the way that a number of us, myself included predicted, say, 10 years ago. 

That said, I think that this convergence of these different technologies that I’ve outlined here do in fact suggest that in the next decade we will see quite a number of new opportunities for building better resilience, and that’s something that’s a big part of our R&D effort at the Swiss Re Institute. 

So, I want to give you just a picture of one approach. I’m not going to spend a lot of time talking about this. I’m happy to follow-up or we can have people within the Swiss Re Institute follow-up with people who have more interest in this type of modelling. But it’s about taking physics-based models and then hybridizing them with machine learning to develop much more robust resilience models. So, for example, you could look at a flood model for an urban area and begin with a hydrodynamic model based on the physics of flooding. And using data that is now becoming increasingly available around, say, the interconnection of the transportation, actually where people move, and just the positioning of buildings and even data around drainage, et cetera, you can overlay on top of your physics-based model a collection of machine learning models. Often this is called ensemble modelling. 

But this hybridized modelling is actually showing a lot of promise, at least in the research context, to provide very robust solutions to allow decision-makers, whether they be insurers working with a client or government leaders working in partnership with corporates, together with insurance companies. These are all different types of configurations that we thought about. But you can then think about where you want to focus investment and this becomes the basis for some very interesting resilience-as-a-solution approaches. You could think about what are the secondary impacts of certain infrastructure decisions, like where you place bridges or how you decide to lay out your water piping system? That could then be fed back into these models to understand how that impacts resilience. This is just an example of the type of exciting solutions that we’re working on today. 

Another related approach conceptually is to focus on this important problem of forward-looking assessment and portfolio optimization, and this is a problem that all insurers face. If you think about an insurance company or a mutual, this question of how you decide which liability segment’s to be focused on can be quite daunting even today. But we’re finding some new capabilities that, again, hybridize things like reinforcement learning, which is a type of machine learning together with interesting prediction models. And the one point here that I want to emphasize is this approach explicitly incorporates subject matter expertise into the modelling process and that’s something that I think is often forgotten in the technology arena. Maybe technologists erroneously believe that they can just entirely replace humans in these different workflows, and this is actually one of the reasons why I think we’ve collectively as an industry not been quite as successful in getting machine intelligence integrated into the workflow, and that is there’s been too many technology efforts, and I’m sure many of you have experienced this, where there wasn’t careful thinking about how to keep humans in the loop. 

And often, experts have tactic knowledge or extensive foundational understanding that is just difficult to train into the current generation of machine learning tools. But you can figure out how to integrate the two, and I really characterize this as the cutting-edge machine learning work, is taking the subject matter expertise when it’s really good and incorporating it into the machine learning. Just to give you a quick anecdote on that point, in the past, I’ve been involved in interesting efforts where we have looked at credit systems, so these are credit risk assessment systems. This is before I worked at Swiss Re. But I think it highlights the point. And what we found is when we benchmarked the machine, the new credit risk scoring machine that was using a variety of data models, it was outperforming the human credit risk analyst approximately 85% of the time. 

What was interesting is we then analysed 15% of the time where the people defeated the machines and that gave us a lot of insight, one, in how to improve the algorithms, but more importantly, identifying areas where it didn’t seem like anytime soon the algorithms or the machine learning approaches were going to be essentially 100% better all the time. And now you have these new approaches that allow you to incorporate this human insight in a systematic way, which I think creates just much better hybridized models. The other point to emphasize is the incredible importance of curation of data, and this include unstructured data, which is another great opportunity that has developed in the last few years. So, this is the ability to process text and process even audio and video in different contexts, and your creation capability can be an important differentiator for how well the systems work. 

And we’re finding that in developing resilience-as-a-service, resilience-as-a-solution, this is an incredibly important part of the process, making sure that we curate all available data and that we properly integrate the human expertise together with the machine learning capabilities. 

Now, this has given us the foundation for some development efforts that we’re doing at the Swiss Re Institute, one of which we call the Risk Intelligence Factory. And this embeds a lot of the capabilities that I’ve been talking about. It’s the ability to properly harvest data, curate that data, develop a replicable and scalable framework. And that’s another important point, just to pause and emphasize, and that is that the failures to make enterprise machine intelligence work in the past, I think, is arisen from the fact that often those technological frameworks were not replicable nor scalable. And I think that has changed, and so that is another reason why I’m much more optimistic going forward that we can make some of these work. 

And then the final point is once you have the foundational data in place, once you have the replicable and scalable framework working that has properly curated data, that can then be enriched with the accumulated risk insight, which again will be a mix of model-based outcomes and also subject matter expert outcomes. I should also just point out on the left-hand side of this slide that the evolving industry ecosystems that are incorporating some of these technologies I talked about before, the distributed ledger technologies, interactions with IoT, et cetera, such as the Food Trust and TradeLens and we.trade, et cetera. These constitute new sources of very interesting data that I think will be quite important for developing global resilience solutions because this will give us more insight with respect to the supply chain, and it also provides examples of digital platforms that we will continue to see evolve as we go into the future, and those will create interesting data sources that can be combined with the large amount of data we already have. And so, I think that that creates, again, a great foundation for new resilience-as-a-service. 

Now, I want to shift just to make this a little more practical to a case study in the supply chain, because this is the area that we find further along in incorporating some of these new ideas. I mean, this is certainly not the only area, but it’s one that I think is quite interesting and the interaction of the pandemic with supply chains, I think is particularly atypical today, just given where we are. 

Now before COVID-19, the supply chain was already in the midst of a pretty substantial transformation, and this is something that Swiss Re and the Swiss Re Institute have been an integral part of as this transformation is going along. And this relates to these new technologies that I highlighted before. So, we have these IoT devices that are becoming more ubiquitous, the digital ecosystems that are evolving, the creation of digital twins that allow risk managers and executives to do quite interesting analysis at points of need, the interaction of distributed ledgers. And I should mention here that not every application needs to have a distributed ledger approach, however, the immutability of data that is a characteristic of distributed ledgers or blockchain can be quite important for facilitating trust among different counterparties. And there is a parallel effort that is now a big part of our research agenda called confidential computing. This is something that not enough people are talking about in the blockchain or distributed ledger community, but I think we will see more of these. 

These are techniques, whether they be hardware-based or software-based, that allow us to essentially keep data encrypted over most of the data value chain, and that is going to become increasingly important as these platforms scale and you have more questions of trust among a variety of counterparties. There’s a branch of confidential computing that we call privacy-preserving analytics that’s quite exciting, and that’s the ability to use new techniques to essentially manipulate data in the context of an analytical model without ever decrypting it. 

Now, we are still very much at the research stage for these approaches, just given the fact that they are very expensive computationally. They require quite a bit of compute power and that means that they’re still not ready for enterprise deployment. It would just be too slow. But I do think that going forward, that will be an incredibly important part of these platforms in order to ensure that the data’s being processed not just in line with regulation like GDPR in Europe, but also in a way that maintains the trust across all the participants. Which as an aside, this has been a limiting factor in developing these platforms today, but I think that that will improve over time. 

And then finally, the development of parametric insurance and the adjacent resilience and risk-management-as-a-service that comes off of this data, which I had been talking about before. And we’re even aware of some interesting efforts, such as being undertaken by ICMIF member Ibisa that’s developing some parametric solutions for improving agricultural risk product and service design in a number of countries. So, I think we’re already seeing some of these technologies deployed in very interesting ways. 

Now, the pandemic has highlighted the new nature of some of the systemic risks. So, one of the things that we’ve been able to do as we’ve watched this play out is look at how the pandemic impacts the non-systemic risks such as supply chain unpredictability, the operational risks, the network risks. I think this pandemic is causing many firms and different market players to take a step back and ask the question, “To what degree do they need to have redundancy in the system?” And that just creates a new set of conversations about risk management. I think it’s also true that it’s accelerating interest in process and controls to improve the ability to do things like predictive maintenance in a much more automated and systematic way. So, these are things that are being accelerated by the COVID-19 pandemic. I mean, they had been in the research arena, but now we’re finding much more interest in actual commercial deployment opportunities to understand them. 

And then if we look at the systemic risks, which are hard to diversify, actually sometimes impossible to diversify, but they do provide some insight into how things like pandemics are interacting with risks that were already there, such as financial risks. What is the interaction of the locking down these different economies and the economic impact, and how does that affect the interest rate environment, and things like even expectations around pricing for things like insurance. We have to think more carefully about the impact of pandemic. I mean, if you go back to my polling questions, certainly for Swiss Re, pandemic was not a black swan. You know how you read that sometimes in the press. But for many well researched risk groups, it was clear that pandemic was a risk that people were thinking about and preparing for, but the cascade of economic and social consequences, I think some of those have been surprises and that financial interaction, that economic interaction is something that can be analyzed in more depth now that we do have this experience. 

And then the other grey rhino that’s out there is this mix of cyber risks that continue to be a growing challenge. It’s an opportunity for insurance because this a fast-growing risk line, however, it also … or the pandemic’s highlighting some of the vulnerabilities that exist in the world with respect to cyber. 

I’d also like to emphasize the notion of algorithmic risks separate from just cyber risk. So, if you think about the fact that we are increasingly reliant on systems and networks, and I think this is something that the COVID-19 pandemic has also highlighted as well as accelerated, there are two broad types of problems that arise from that vast increase of network scale and system integration. The first is system vulnerability and that plays out in terms of cyber risk. So, you have vulnerability points potentially and then that ends up being potentially a problem is somebody tries to hack you for a variety of reasons. But there’s also system fragility and that could arise from just the complexity of the system. Often, systems are not well architected. That could lead to operational failure unrelated to, say, getting hacked by some kind of outside party. Or you could have what we call algorithmic malpractice. So, you could have components in the system that were not properly coded. Maybe they generated biases that end up having downstream impact or they might be propagating errors that eventually cause the system to fail. 

These are developing risks that I think have been highlighted by the fact that so many people had to move online in terms of their work as this pandemic has developed throughout the world. So, if we think then in terms of the post COVID-19 supply chains, we are seeing a material increase in demand for resilience and risk transfer solutions already and this ranges from looking at tier dependencies. The developing datasets that are coming off of these networks and our ability to join them with other data creates a very interesting opportunity to analyse those dependencies. I think that there’s increasing questions of how one does geographical diversification. In the past, the question on supply chain was much more about how to optimize movement of the different material goods and inventories across that chain without thinking too much about the geographical question. That illusion is gone. Now we have to really think about, we think carefully about geographical diversification and what it means, what are the dependencies that develop there. 

Again, to do that analysis in a robust, replicable and scalable way, you do need to have these new tools that I’ve been talking about. We now are able to deliver insurance at the point of need. There’s still lots of questions about how this will work in practice, but we are confident that this will continue to be a growing area in the parametric space and we think it’s quite interesting. And then we are finding more demand from the corporate and also from governments around supplier diversification. Again, the ability to rely on a mix of data and models I think is very important to do that effectively. 

And I want to just finish now with a discussion of the types of solutions that we’re thinking about. We’re strong members of ICMIF and the mission that is being put forward in the context of this organization. And the ability for us to deliver solutions that are enabled now by data and technology is just better than ever. And that ranges from performance type solutions that allow for better portfolio steering, exposure management. I think this has been probably one of the areas where the technology has moved a little faster and more successfully than others, despite the overall slow pace in terms of seeing enterprises getting transformed. But reducing the volatility in the underwriting portfolio, the ability to extract key risk drivers. We think that that is becoming much more possible with this mix of capabilities. And then the ability to comply with regulation. 

Now, one of the more exciting solution sets that we’re developing is in the sustainability space, and this is where we essentially are able to use the UN SDGs as guidance, and then develop benchmarking and tools to understanding better how one’s strategy and developing a portfolio and impact are related specifically to those SDGs. And we think this is a very exciting area of development and look forward to rolling this out with a number of interested clients in the near future. 

One of the important aspects of this work is the ability to score performance against SDGs and do that specifically to a particular business of focus. This would allow you to benchmark yourself against ICMIF members and the insurance industry. You could also understand more systematically how your exposures are impacting your progress toward compliance with any given SDG. So, this particular insurance SDG (iSDG) calculator is one of the more exciting solutions that is being enabled by this new mix of data and technology. 

We also have another piece of technology that will allow us to expand insurability across a number of risk areas, and this is based on a technology capability that’s called Sphere. Right now this sits within the Swiss Re Institute. It continues to be an R&D developed technology. But what’s important about it is that it provides a comprehensive capability to automate claim processing, facilitate more risk prevention based on data, improved quoting, and then potentially add a variety of value-added services. Hopefully, we will be able to get this video to run.

So, what the Sphere technology allows is the ability for a client or a customer service representative from any organization that’s looking to offer insurance to go through and essentially video different items or insurable items, say in a home. I mean, this is, I think this is the example with this video. So, the user would go and follow a set of simple instructions with their smart phone. This would allow the insurer to then have an upload of all the different items in the home. I mean, I guess this particular example would be for some kind of home insurance. What’s interesting about this is that many of us in the technology space have talked about this capability for, at least for me, since 2012. But it’s been very difficult to see technology deploy that actually can handle the scale of a real business and has computer vision recognition at a level that doesn’t require extensive error correcting after the fact. This is really where we’ve seen the breakthrough. 

So, this Sphere technology boosts the user engagement. Some people might even find it more interesting than your more typical insurance experience. And on top of that, we’ve developed with a partner a proprietary scoring model that allows us to compare a user’s house topology with the more typical topology specific for a particular market segment. So, this takes us one step closer to doing much more personalized covers than we’ve seen before. So, that’s something that we’re quite excited about. 

So, just in summary, or actually before I get to the end, I have one final question that I’d like to put up. This is more about thinking into the future. I’ve shown you some examples of technology that’s working now. There’s some interesting technologies going into the future and I guess I do have to apologize. I’ve been a little too generic on the machine learning and artificial intelligence because underneath, there’s very specific capabilities that are developing. So for example, reinforcement learning is a machine learning approach that hasn’t really moved beyond robotics as much as I think it should, but that’s an example of what may potentially be quite an influential technology in the future. 

But before I conclude, I would like to get people’s view on which one do you think will be the most influential technology in the medium-term? Say, the next decade? And I will finish. These are all technologies that we see in development today. We think they will have an impact. It’s hard to tell what that impact will be, but it’ll be interesting to see how that plays out. Oh, interesting. So, I don’t know if my reinforcement learning pitch persuaded you, but I do agree. I think that in the next 10 years, it’ll be quite important. I should mention on the quantum computing side, there is a parallel effort that’s often called quantum-inspired algorithms, which I think is a type of machine learning that could be quite influential as well. So, that’s something if you haven’t heard about it, you might want to look into it as well. Quantum computers are still probably quite a number of years away, but quantum-inspired coding or quantum-inspired algorithms may be quite soon in terms of our ability to use them. 

And I will just conclude. So, in terms of Swiss Re Institute’s new data driven offerings, and again, let me repeat, these are still very much research and development efforts. But we do like to do these efforts together with clients and are happy to explore those types of initiatives. But they break into three big areas, risk products and services, risk analytics and risk studies. And as I mentioned, we have some very interesting developments in the sustainability solutions space. The development of scores that would allow you to benchmark against others as you work to comply with SDGs or even have clients comply with SDGs. The development of end-to-end risk platforms. I think one important point to emphasize here is that despite the very interesting developments in the machine learning and artificial space as it applies to, say, consumer retail or social media or even just in the academic community that they’ve made a lot of their progress in innovating algorithms, we still don’t see the types of transformative deployments in enterprises that you would expect. But as I mentioned, these technology trends are creating the platform for that to change, and we expect to be part of that process. 

In the risk and analytics face, our ability to do better predicted risk analytics by hybridizing subject matter expertise from humans together with machines, we think that’s a very promising area. And as I mentioned in our Risk Intelligence Factory, data curation becomes an incredibly important part of that end-to-end process. And then the creation of these new tools, such as Sphere that we think will revolutionize home insurance, that is quite exciting as well because we’re finally seeing the actual realization of promise with respect to things like computer vision and the processing of information at scale that can be used to do things like more personalized underwriting. And this all feeds into a wide range of risk studies, which is Swiss Re Institute has a long history of generating. You’re probably familiar with our Sigma publications and we also publish a variety of risk expertise papers in these contexts. 

But we’re very excited about the continued co-creation of these new resilience-as-a-solution tools and we look forward to working with interested parties to develop it further. With that, I think I’ll close and leave a little bit of time for Q&A. 

Ben Telfer: 

Thank you very much, Jeff. That was a really fascinating presentation. You touched on a number of really key themes that both ICMIF and ICMIF members are heavily interested in. The idea of leveraging those emerging technologies for social impact, risk prevention, resilience and sustainability and the Sustainable Development Goals. We do have a number of questions that have come in already. Again, that’s a reminder, please do send in any additional questions. And if you would like to ask your question to Jeff, please raise your hand by clicking the little hand icon on the platform, and we’ll try and enable your mic so you can ask Jeff directly. 

Jeff, first question. You may have already touched on it. The question actually came in before your last voting question and it was that 50% said that the internet was the general purpose technology that had the biggest impact in the last 80 years. What do you think will be the technology that will have the biggest impact in the next 80 years? 

Jeffrey Bohn: 

Well, I think it will be or my prediction is it will be something out of the machine learning, artificial intelligence face. There’s a lot of debate about the risks of AI, for example, and I think we should continue to have that debate. But the ability to process data to improve decision-making, I think will just continue to have and outsize impact relative to these other technologies. There’s been things, despite the fact that the internet has massively changed the world, it hasn’t always changed the world for the better, and I think that some of these new technologies together with human intuition and guidance, and dare I say, ethics and a moral outlook, that could have a much better impact than this. So, that’s the area where I see the next 80 years going. But it’s by no means deterministic. I mean, there’s a lot of things that could happen that could take us off different paths, but right now I think we’re on that path. 

Ben Telfer: 

It’s interesting you mentioned ethics because that’s referenced in the next question. So, it just says “the use of data is going to be a game changer for insurers and to insure resilience. How will this be impacted by data regulation, increase in data regulation that is likely to come in and also the ethical considerations of data, considering some of the high profile breaches and the exploitation that is being seen by large corporates already?” 

Jeffrey Bohn: 

That’s a very good question and there’s two components of my answer. One is going back to that confidential computing. Right now, data anonymization techniques that are typical are actually just not strong enough in light of some of these new machine learning capabilities. It’s actually quite easy … Well, easy maybe’s not the right word, but straightforward to break many of these anonymization approaches as they’re implemented. And I think that in the context of the data privacy regulations, which is quite important, if we don’t strengthen things like privacy-preserving analytics, that will limit what we can do analytically. So, I think those two work hand-in-hand. I’m a strong believer in data regulation. I don’t buy into this narrative that it’s going to somehow hinder innovation. I mean, it might create short-term challenges, but if you look at, for example, the privacy-preserving analytics research, that is a reaction to the data regulation, and I’m confident we will end up with very robust confidential computing, which will improve our ability to both comply with regulation and deliver better analytics. So, that’s one aspect of my answer. 

The second is that if you think about artificial intelligence or machine learning or any type of machine intelligence, unless we deliberately decide to develop it in an ethical way, in a way that is free as much as possible from bias, and we think carefully about the impact on society, I think that we will overall be better off. But again, that puts cost, that puts some short-term hurdles, but I think we need to be having those conversations. So, I think that more of these conversations around there’s different slogans, AI for good, ethical AI, I think that that’s something that we all need to be talking about. Because there’s nothing inherent about these technologies that make them ethical. We decide to make them ethical. 

Ben Telfer: 

That’s a very good point, Jeff. And something that as mutual and cooperative insurers, I know there’s been a lot of questions about the use of technology, the use of big data, artificial intelligence, and how you can do that in an ethical way, and how that can be a differentiator to how potentially stock companies will be using data in these technologies. We’ve got time for a couple more questions. This is really interesting. Well, I thought this was a really interesting question. Imagine that we’re 10 years down the line in terms of how 10 years of advancements in machine learning and other technologies. How would the response to the COVID-19 pandemic have been different if we were 10 years ahead in terms of implementing these technologies? Nice, simple question for you, Jeff.  

Jeffrey Bohn: 

No, actually, that’s a great question. Here you’re going to get my scenario development hat. But I think that if we had better network systems in place, again, subject to all of the controls I was talking about, you got to regulate the data, et cetera. And assume that you had low cost, fairly fast tests, and then that could be integrated with contact tracing. I think that you could manage through pandemics without having to lock down your whole society. And so, that’s what I would hope or I would expect that to happen 10 years from now. What I suspect is that type of capability will not be rolled out equally across the world. I mean, even today, we’ve seen some countries that have better testing and contact tracing in place so that they can be more surgical in terms of dealing with the super spreaders, in particular. 

I’m actually quite surprised that my own country, US, at how behind we are in terms of having tests and contact tracing, et cetera. And so then you have to go for these very blunt instruments where you just shut down the whole economy. So, I’m hopeful that we’ll be much better at managing that 10 years from now. 

Ben Telfer: 

Thank you, Jeff. I think one more question we’ve got time for. The question is, prevention is a growing theme in the insurance industry and especially in the mutual and cooperative sector. The person who asked the question says that it was a big theme at the ICMIF Conference last year in 2019. The question asked, “is prevention seen in other markets around the world, especially in the shareholder-owned companies? And do you see it having more of an impact at that large macro level with large reinsurers helping to help create resilient cities or do you think it’s more apparent in the micro level through microinsurance?” 

Jeffrey Bohn: 

Wow, that’s a very insightful question. Let me take the last part of that first. Whether it’s micro or macro, I actually think that to get prevention to work well, it probably has to be more micro-driven. So, I think that that is important. I mean, we’re firm believers in public-private partnerships, so we have some efforts underway with some selected governments that we’re trying to work together to put the tools in place around the data and the analytics to do better prevention and then make that available to the different parties in those areas. So, I think there is a macro component. However, I think we’re still at the stage where we need more experimentation. And my own experience is that that only happens at the micro level. So, people taking advantage of, say, smaller targeted groups and figuring out stuff that works and then they can bring that to the larger industry. 

In terms of the industry generally and reinsurance, yes, this prevention question comes up everywhere, so I don’t think it’s specific to this group. I think that the big issue I still find is that collectively as an industry, and this would be not just cooperatives and mutuals, but all insurance entities, whether they be reinsurers or share-based insurance companies, we’re still collectively not that great at deploying technologies at an enterprise level, and we’ve collectively under-invested in data curation. And I think those two components create obstacles to really doing some of these more robust prevention solutions. And so we had to, it’s not … we’re finally at the point where I feel like the technology together with the data availability is creating opportunities, but there’s still this foundational work that has to happen, and that is investing in proper data ingestion and curation. 

And then secondly, also thinking more carefully how this gets deployed in real enterprises, whether it be a cooperative or large reinsurers. But large reinsurers, we’re a bit prisoners of our old organizational paradigms, and those have to shift as well. So, it’s not just about technology. It’s also about how people make decisions and organize. 

Ben Telfer: 

Thank you, Jeff. I’m afraid we’re going to have to end the webinar there, but I mean, I’m sure we could have carried on speaking for many hours with all these questions. Thank you again, Jeff. Really excellent presentation, discussion, and I’m sure we look forward to more about how Swiss Re and other ICMIF members are working together to- 

Jeffrey Bohn: 

And please reach out to me or anybody in the Swiss Re Institute staff is you have more questions. We’re happy to come and talk to you and look forward to further conversations in this space. 

Ben Telfer: 

Well, thanks for that, Jeff. You almost teed up my what I was going to say next about how we have one of your colleagues, Astrid Frey speaking in this afternoon’s webinar. Well, this afternoon, 2pm BST, Astrid will be talking about the impact of COVID-19 on the insurance industry and the economy, and again, following on from some of Jeff’s things that he’s presented there.  

So again, a final thank you, Jeff. Great to have you join us today and I look forward to hearing, speaking and participating with everybody at later this afternoon at 2pm BST. Again, thank you and enjoy the rest of your day. 

The above text has been produced by machine transcription from the webinar recording. ICMIF has made every effort to ensure that transcriptions are as accurate as possible, however, in some cases some text may be incomplete or inaccurate due to inaudible passages or transcription errors. Listening to or watching the webinar recording will allow you to hear the full text as delivered during the webinar but this is available in English only. Our transcriptions are provided to enable members to select the language of their choosing using the dropdown menu above.