[00:00:00] Hello, I’m Veronica Combs, and this is the Quantum Spin by HKA. For season four, we decided to do something a little different. In March, we attended the APS Global Physics Summit in Anaheim, California. We took advantage of this amazing event to talk to the leaders in academia, industry, as well as the creative folks who helped make the event such a compelling experience.

[00:00:23] I hope you enjoy these conversations that really reflect what’s happening in the industry right now.

[00:00:29] Veronica: Today I am talking with Dr. Frederik Flöther, who is the Chief Quantum Officer at QuantumBasel, along with several other titles and job responsibilities. Thank you for joining us today.

[00:00:39] Frederik: Thank you for having me, Veronica.

[00:00:40] Veronica: So you are the Chief Quantum Officer at QuantumBasel, Deputy CEO at QuantumBasel, and also a member of the University of Basel’s Center for Quantum Computing and Quantum Coherence.

[00:00:50] So it sounds like you’re very busy on a regular basis.

[00:00:53] Frederik: Yes. I think that’s a fair statement. I do have different roles, so some are more on the managerial side, others more on the research side at QuantumBasel. So I should say, first of all, QuantumBasel is the center of competence for quantum computing and AI as part of the uptown Basel Innovation Campus. In Switzerland and at QuantumBasel, Chief Quantum Officer is a role. That’s right. The title has not yet caught on, but I’m confident that it will. There I lead our team of quantum algorithm researchers and at QuantumBasel we also have data scientists and other functions.

[00:01:25] And as deputy CEO I’m responsible for helping steer the work such that QuantumBasel can contribute to the long-term success of the uptownBasel campus, an ecosystem. The QC two Center for Quantum Computing and Quantum Coherence. That’s a joint effort that we have with the University of Basel in terms of scaling a Quantum computing institute and closely linking the quantum hardware research with the application research to progress the path towards commercial use cases.

[00:01:54] Veronica: Yes, they really have to advance together. Absolutely. Or we won’t get very far with you, hand in hand. QuantumBasel is part of the work that you do to support entrepreneurs, right? To provide them with resources and connect them with experts. Is that right?

[00:02:07] Frederik: Yeah. So we work with a range of different organizations, startups, companies, universities, research institutes, and in fact we also have a sister company they’re called QAI Venture and they are a startup accelerator investor and ecosystem builder that is active in that space.

[00:02:25] Veronica: You specialize in healthcare and quantum technology and I’m curious about what your work is like in that area.

[00:02:32] Frederik: Yes, so I think for the last decade or so much of my work has been at this intersection of quantum computing, machine learning, and medicine, healthcare, pharma.

[00:02:42] I first, after my PhD, joined IBM and was working in data science there, looking specifically at data science healthcare applications. And following that, I joined a new team. I was the first employee of the team at IBM looking at how we can use some of those quantum algorithms in real world settings.

[00:02:59] And in particular there I was also looking at healthcare and life sciences. And yes, indeed that’s exactly the sweet spot where much of my research has happened. So for example, I’ve worked with a range of different organizations companies like Amgen, Pfizer, and Rush. One of the projects I was driving there was with Amgen where we looked at classical machine learning as well as quantum machine learning, applied to the problem of predicting persistence, the length of time until a patient discontinues a therapy.

[00:03:31] So predicting the persistence of rheumatoid arthritis patients. And so we investigated. How do some of those quantum algorithms compare with some of the classical approaches?

[00:03:40] Veronica: And I believe you’re working on a book as well, is that right?

[00:03:43] Frederik: That’s correct, yes. So actually that book which I’m co-editing, that’s in the AI ethics space.

[00:03:48] It is tentatively titled “Computers with Salaries and Cemeteries: AI Ethics from Industry to Philosophy to Science Fiction.”

[00:03:57] Veronica: Oh my goodness. What a great title.

[00:03:59] Frederik: Yeah. Not AI generated, I should actually say human creativity there.  And also for example in terms of ethical angles that ought to be explored. And now in this book our belief is that ethics is a topic. It concerns everyone. It is also something that requires input from everyone. Maybe I should first say for me the fascination with quantum and quantum related technologies, the applications and the technology is one thing, but I’ve always really been fascinated also by holistically what is around it? What are the implications? What are the interpretations of quantum mechanics?

[00:04:38] So you can almost think of it like crowdsourcing effort and in particular it requires interdisciplinary exchange and dialogue. And that’s why, as you can see from the book’s title we have thought leaders there from across the world, more than 20 chapters, and trying to get all those different perspectives from business and industry, from philosophy, from literature, and also we are very very excited about it and it should come out within about a year.

[00:05:03] Veronica: Oh, great. Great. Yes. You really don’t have the full scope of a problem unless you have all those different viewpoints.

[00:05:09] Frederik: Absolutely. And you can cross-fertilize each other and a lot of innovation happens at those intersection points.

[00:05:15] Veronica: I believe you mentioned a case study in the book. You built an end-to-end pre-processing pipeline for EHRs.

[00:05:21] And I think part of that work was working with limited data sets and the quantum’s ability to do that. There’s so much data in EHRs, but it’s hard to get that data out and do things with it.

[00:05:31] So is that part of your work, the quantum algorithms that you’ve been working with?

[00:05:34] Frederik: So yes. One of the particular challenges when you try to apply these algorithms to real world data is exactly this noise and having data sets are difficult to handle, which, they might be very big, they might be very small when you’re dealing with rare diseases, for example.

[00:05:50] And I think it’s particularly interesting then also to consider how quantum computing may be able to address some of those issues when you’re dealing with difficult data sets. I think one of the interesting things also, when you’re looking at the factors that affect health outcomes, there’s increasing evidence that actually at least half or more of the influence of different factors is due to these social determinants of health. So the environment and also the behavior. And one thing is, first of all, collecting all of this data. So where do you get that data from and how do you measure it and consolidate and curate it?

[00:06:25] Things such as measuring the pollution levels, things such as measuring your physical activity with trackers and variables. Measuring factors such as your social interactions, which I guess hardly anyone measures those but they’re key. Right? If we look at longevity studies and the like, and so even once you’ve brought all of this data together, which is in itself a challenge, but then getting those action insights is very tricky.

[00:06:52] And that’s where some of these advanced algorithms come in. AI definitely plays a role, but also quantum algorithms are particularly interesting in order to try to uncover some of these patterns and correlations, which we may not be able to spot with classical means.

[00:07:05] Veronica: Yes. I remember when I first started reading about the social determinants of health, and it seemed so smart, right?

[00:07:11] If you live in a place with high pollution levels or there’s been industry there and the soil is poisoned, that would obviously affect your health. But like you said, how would you know that? How would you measure that? But with quantum computing the promise seems much more real that we could actually do something with that information.

[00:07:27] Personalized medicine always seemed like only if you had the very best healthcare and plenty of money to pay for the extra care. But maybe quantum algorithms and software open up the door to making that more broadly available to more people, hopefully. Correct?

[00:07:39] Frederik: Correct. Yes. And we should be clear, of course, quantum computing, it is still a very young technology.

[00:07:45] Yes. And there’s still a lot of work that needs to be done. But at the same time, when you consider biology is very complex. It is. It is very humbling. Indeed. It is. I like to think of quantum computing there as this additional tool that you have in your toolbox as a data scientist.

[00:08:00] And it will not help for every problem but at least as the quantum hardware and the quantum software continues to mature it may help provide some additional insights and in particular, in combination with other techniques. Being able to build such ensemble methods, for instance, where you take the best of all the models that I think is a very powerful combination.

[00:08:21] Veronica: Yes. Certainly. And I was reading some of your work and you’ve not only looked at the social determinants of health, but also I think you’ve mentioned this before briefly, about how long will a person stay on a particular treatment, because obviously compliance is also a big factor in healthcare.

[00:08:36] Frederik: Of course. Yeah. So this relates to the work I mentioned earlier on collaboration with Amgen where we predicted the persistence of rheumatoid arthritis patients

[00:08:49] Patient engagement has been called the blockbuster drug of the 21st century. And so I think anything one can do , in order to foster that absolutely can have a major impact on health outcomes. I think this also connects with the whole topic as these algorithms progress and as also we are starting to see now some of these first approvals also by the FDA for AI based diagnostic devices, for instance, in spotting colon cancer.

[00:09:10] It is also key then to foster the explainability of those models. Because if as a patient I do not understand why a certain therapy is recommended or discouraged, I may not be very motivated in actually pursuing it and then this would not be optimal from an engagement point of view.

[00:09:27] Veronica: Right. And do you see it? I mean, I have faith in technology, but I also have faith in doctors and they’re experience with patients. Do you see AI and quantum software as complementary to an actual human being doctor? How do you see that partnership evolving as time goes on?

[00:09:41] Frederik: So I like the term decision support systems. I think it is at this point in time I think few would be willing to completely surrender the decision power to, to an AI system. However, we should also be, I think, humble with regard to the progress of the technology. And if we go back a few decades or even just a few years there have often been statements well, uh, computers, and they can never do X, Y, Z.

[00:10:08] Then X, Y Z happened at some point. So I think we should also just be cautious in that regard. But for the foreseeable future, I think it’s these decision support systems and it’s working hand in hand  together with doctors.

[00:10:21] Veronica: Even some of the tools that summarize research. I think I read that doctors would have to read 24 hours a day for several months at a time, even to just keep up with all the research out there.

[00:10:30] So I know some of the large language models have these sort of research synopsis tools, even that part would be a good decision support tool.

[00:10:36] Frederik: It is in fact one of the projects we also have ongoing right now at QuantumBasel where we are working with a startup called CGC Genomics.

[00:10:45] And the task there is we are looking particularly in the oncology space because of these oncology reports, they can be dozens of pages long, very complex and even professionals, experts in this space may not have the time or fully understand, so what are the concrete next steps now the concrete therapy steps that we should be taking?

[00:11:09] And so as part of that, we have actually been developing certain AI systems, fine tuning large language models in order to understand this from a content perspective and again, act as a decision support tool in terms of summarizing and providing concrete actions and next steps.

[00:11:25] Veronica: I’ve talked to folks about ethics and quantum and ethics and technology before, and it’s fine to talk at a high level.

[00:11:30] Can you think of a specific situation that maybe doctors or CEOs or technologists would’ve to pause and actually weigh the ethics of a situation around like deciding whether to offer treatment or not offer treatment. Do you think about it in specific use cases like that?

[00:11:45] Frederik: I think from a technology point of view, there are these general questions which need to keep being discussed.

[00:11:51] And this pertains not just to quantum technology, but also to AI and related technologies. In terms of, first of all who should actually be accessing such technology is that I think with quantum it’s even more extreme actually than with AI because the access to quantum hardware is still rather limited.

[00:12:07] It is not like there are endless quantum computers in the world, right? Out there. So who should be accessing those? And then what sort of application should be investigated with them? And then finally, also, if we consider hopefully that we will have such breakthroughs at some point where we can perhaps discover a new drug or things of that nature.

[00:12:27] And the question is, who gets to reap those benefits? Can we somehow make this broadly accessible or is that only for those people, countries also that have developed that technology. So I think from a technology point of view, all of these factors come into play. And you’ve mentioned some scenarios then in specific industries how that plays out.

[00:12:47] So if we are looking again at the medical space how do we then for instance, if we have some new drug like that or some new treatment how can we then give that to a given set of patients?

[00:12:58] Veronica: So I guess one analogy might be access to AIDS treatment. I guess like the richer countries developed those and they weren’t because initially they were so expensive and we needed a way to get them to everyone who needs them.

[00:13:09] Frederik: Correct. And I think hopefully we can also learn from past developments. And the whole ethics space here, I think is a good example. If we look again towards AI, the recent progress in particularly generative AI, probably surprised many people or the majority of people.

[00:13:25] And as a result, the ethics discussion has almost come a little bit ad hoc and I think in many ways it could have been more proactive in terms of thinking about what regulation may or may not be needed. So I think with quantum computing it is an earlier technology and so perhaps we can also take advantage of this fact and try to already progress that, that particular kind of discussion.

[00:13:47] Veronica: Yes, definitely. And you mentioned before model explainability, I have a 16-year-old son and he uses Chat GPT sometimes probably more than I want to admit, but sometimes, and so we always use it as an opportunity to talk about where do you think it got that information?

[00:14:01] I was explaining to him how it actually works. And I think that model explainability is such an important issue. But if you’re a patient getting a cancer diagnosis, maybe you know that’s farther from your mind. How do you think about model explainability in a healthcare context as an ethical issue?

[00:14:16] Frederik: I touched on it already a little bit with regard to some of the regulatory aspects. So for instance from the FDA’s perspective and also with regard to as a patient really having that full engagement. But I also want to give an example also from the point of view of clinical adoption.

[00:14:36] And also how this can affect research decisions. In one other collaboration I worked with Roche Diabetes Care and we pioneered the application of machine learning to real-world data sources, in particular electronic health records. And as we were developing these models we tried different techniques, some more on the black box side, and some more easily understandable.

[00:14:56] And we actually found that in some of those black box approaches, we didn’t get significantly better performance in terms of prediction, accuracies, and the like. But at the same time we then also saw that not only did we not get better performance, but it would make it significantly more difficult to actually use such models in a clinical setting.

[00:15:13] And so we then focused on logistic regression. So one of those approaches that is more easily explainable and then focused on that in order to progress for this particular paper.

[00:15:24] Veronica: I was reading about the AI summit they had that was in Paris, I think it was in February.

[00:15:28] And, the tone of the conversation was, regulation will stifle innovation and we can’t have that. How do you see the sort of a global sentiment around regulation and AI right now?

[00:15:38] Frederik: I think it’s a very polarizing issue because you can look at it as you say, you can then look at it from the point of view, it will stifle it. And on the other hand, you can make the argument if AI really is as transformative as some people claim to be, it’s been called humanity’s last invention perhaps.

[00:15:55] If it really does have this power, then perhaps regulation is not such a bad thing. So I think you can make the argument definitely on both sides. I think the whole nature of ethics, and so it can sometimes be misunderstood a bit that as soon as you think about ethics, that you are somehow inhibiting progress.

[00:16:13] This is not necessarily the case because it can also help guide innovation in ways which you may not have thought about before. In the same way, to draw an analogy here, the whole discussion around energy consumption. Also very closely connected to a lot of the AI developments, you might say, okay, this will just cap us to a certain model size, but you can also try to make a virtue out of necessity. And then you say you know what, maybe this will then inspire more efficient techniques. And we already see some of those developments out there, for instance, with what are called liquid neural networks and other such technologies.

[00:16:51] So I think also from an ethics point of view, I think let’s use that in order to actually help innovation and bring forth new ideas, because otherwise it becomes a sort of self-fulfilling prophecy. If we say these things stifle innovation, then yeah, maybe we won’t have more ideas.

[00:17:07] Veronica: I was reading an essay about how guardrails and ethics could actually increase adoption because it would increase confidence and trust and it would maybe avert something terrible happening, like a denial of care or shutting off some services or, I don’t know, like I, I don’t wanna speculate about the bad things that could happen, but that those guardrails might actually be better in the long run for the technology.

[00:17:30] I thought that was a compelling reason to have guardrails and guidelines on ethics and the whole thing.

[00:17:34] Frederik: Yeah. I think again, it is worth looking here both towards the past. So what have we seen already in terms of other technological developments, but also looking towards the future?

[00:17:44] I mentioned in this discourse, it is important I believe also to have this inspiration from science fiction, from films, from literature, and to see what is happening there. And I think it’s always a question of degree, right? Because I think most people would say neither should one completely stop technological development, nor should it go completely unchecked and in a chaotic fashion.

[00:18:08] So it’s somewhere in the middle there. And with that in mind as you say, I think we need to look at both. What is short term the impact and this of course is complicated then also by geopolitical competition and rivalry, but then also long term, how can we progress towards some of these lofty goals for society?

[00:18:28] Coming back to healthcare here, for instance, precision medicine. Really driving towards this personalized, proactive care and being able to keep people healthy as opposed to only finding cures after they’ve already fallen sick.

[00:18:41] Veronica: Yes. Are there any pieces of science fiction or a movie or a book that you particularly find inspiring or that you recommend that people watch or read?

[00:18:49] Frederik: I guess it’s a matter of taste. Since we’re on the topic of medicine and healthcare, one movie, which is among my favorites, it’s called “Gatica.” And this paints the scenario, the setting where we live in a world where genetic engineering has really progressed and people make choices.

[00:19:06] It also affects social class around that. And how would society evolve in such a setting?

[00:19:11] Veronica: I confess, I have not seen “Her”, which several folks have recommended as a sort of AI, somewhat of a cautionary tale about AI. And I did read a book called “Six Wakes,” which is about clones and space travel, and is this person still a human after they’ve been cloned? But yeah, it is a really good way to get into some of these complicated issues, I think.

[00:19:33] Frederik: Yeah. And I think it brings new ideas forth and some, okay, will it exactly pan out like that? Usually not. But it makes you think about some of these issues.

[00:19:42] And with regard to the genetic engineering, clearly better understanding the genome is extremely powerful because although I argued earlier that social determinants of health play a key role still, our genes do also play a key role as we know.

[00:19:56] Veronica: Yes. The environment turns some of them on and turns some of them off.

[00:19:59] Exactly. Yes, exactly. Complicated. So what is ahead this year for QuantumBasel? Do you have any new initiatives or anything in particular you’d like to call out that’s coming up this year?

[00:20:09] Frederik: I think this year’s special internationally because we have the year of quantum, right? I think it’s nice that this gives it that general momentum and awareness.

[00:20:19] I think it’s also nice actually because when people hear quantum, often it is because we talked about science fiction for them. Often it is a kind of science fiction topic in that sense. But it is important to realize that actually this theory was discovered, developed a hundred years ago.

[00:20:33] So in that sense it’s nothing new there. It is just that we’ve now advanced to a stage where we can often what are called the next generation quantum technologies, if we consider some of the first generations, such as lasers or superconductors or the like, but where we can look at these next generation quantum technologies like quantum sensing quantum communications and quantum computing as well.

[00:20:54] So for us at QuantumBasel, in general, we see ourselves positioned between lab and industry. Meaning as I’ve said earlier, we work with academic partners, but at the same time also work to try to find some of these early opportunities perhaps for thinking about business problems in a new way and drive towards quantum advantages.

[00:21:15] And so we want to continue that journey. We want to build on some of these projects that we’ve already worked on and really try to continue fostering this acceleration of going from the lab to the industry.

[00:21:25] Veronica: So looking ahead a little bit and moving past the debates of when quantum advantage is going to arrive, what would you do, what challenge would you take on with a fully fault tolerant quantum computer?

[00:21:36] Frederik: I think that’s an ethical question in itself, because now we are debating the relative merits, perhaps, of some of these business and science problems.

[00:21:44] But sticking to, let’s say with the healthcare theme I’ve argued earlier that I really think is crucial, in order to get precision medicine, we need to think about keeping people healthy. And in order to keep people healthy we have to define some form of a health status. And then we need to have continual monitoring.

[00:22:04] As soon as we see some anomalies, deviations we need to then have early proactive interventions because they are much more effective than when people fall sick. So in order to actually then find those anomalies, we need to become much, much better at spotting these earliest signs and indicators.

[00:22:23] And in fact it’s also a paper which I recently co-authored, which looks at what kind of quantum algorithms and techniques could apply to this problem of biomarker discovery? So if we had, let’s say, a full tolerant quantum computer, I think I would advocate that this would be one of the problems that could be of particular interest to look at.

[00:22:44] Veronica: What kind of biomarker would you look for?

[00:22:46] Frederik: It could be anything really. So we have classic ones glucose levels HbA1C cholesterol levels over time. Yeah, it, but it can be anything.

[00:22:55] It can be medical images. Even if we analyze how we behave and conduct ourselves in terms of, let’s say in our writing, we may have early indicators also with regard to mental health. So it’s really about getting this holistic picture. And very often biology is complex.

[00:23:13] And so in that regard, it is really the case that often the diseases we want to consider are multifactorial. And so when I talk about biomarkers, I think very often this would be some combination of many different factors where, again, with AI techniques and then also quantum algorithms, we can hopefully uncover some of these new patterns, which so far we have only insufficiently understood.

[00:23:38] Veronica: Right. So even something like gait, or like you mentioned before, social connections. Like how many times have you seen your best friend or your grandmother or your child in a month, that kind of thing.

[00:23:49] Frederik: Yeah, absolutely. And just distinguishing from, in some cases it’s just trivial, whether it’s one time or five times or so.

[00:23:55] But in other cases, especially if there are deviations over time, and now also looking at time series kind of analysis this could be some things which really generate important insights.

[00:24:06] Veronica: And you could imagine tailoring an algorithm to fit a certain population, an age or a genetic background to make the results more personalized.

[00:24:17] Frederik: Absolutely, there’s a huge gap still between where we are today and ultimately this level of the individual, because even now, a lot of the research and analysis is done on non-representative populations. And even just being able to consider additional data sources and step by step make progress towards that level of an individual will still be a lot of work.

[00:24:40] Veronica: Right, right. Like hormone levels in men versus women.

[00:24:43] Frederik: For example. And here we are just again, we are just talking about genders and there are so many, the level of an individual is so far away from that still.

[00:24:52] Veronica: Right. A lot of work to do in the future and hopefully really powerful tools to do it with.

[00:24:55] Frederik: I think that’s why this sort of intersection of these different types of ways of doing computing with in particular the medical space, I think is certainly something which will keep us busy and which will still allow a lot more interesting research for a long time to come.

[00:25:10] Veronica: Yes. Thank you so much for your time today. I really appreciate the chance to talk to you about your work and learn more about QuantumBasel.

[00:25:15] Frederik: It was a pleasure. Thanks for having me, Veronica.

[00:25:17] Veronica: Thanks for joining us for another episode of the quantum spin by HKA. You can find all episodes on our website, hkamarcom.com. Of course you can find us on all your favorite podcast platforms as well. Follow us on LinkedIn under HKA marketing communications

[00:25:34] if you have an idea for a guest, or if you’d like to be on the podcast yourself, you can reach me on LinkedIn, Veronica Combs, or you can go to our website and share your suggestion via the contact us page. Thanks for listening. Talk to you soon.