At CityAge’s Crossover: AI events, with 800 delegates from around the world, we learned that AI is already being used from the genome to the waiting room, to keep us safe.
During this pandemic, the use of AI is pervasive. And after the pandemic, Ai’s application in health care, and pretty much every sector — is going to accelerate, probably exponentially.
You can read some insights on what AI and the future holds for us and how it is making contributions to human wellbeing in this transcript from CrossOver: AI featuring:
- Purang Abolmaesumi, Professor of Medical Imaging, Image Guided Therapy and Applied Machine Learning, University of British Columbia
- Handol Kim, CEO, Variational.AI
- Helia Mohammadi, Chief Data Scientist | Healthcare and AI Lead | Canadian National Healthcare | Microsoft
- Panel Chair: Federica di Palma, Chief Scientific Officer and Vice President Sectors, Genome BC
How AI is Beating COVID-19
Federica:
Thank you Naomi and welcome to our panel. Today, we will be discussing with our experts the role of artificial intelligence in post-pandemic growth. Of course, with its abilities to learn from data partners, patterns, machine and deep learning have proven to be essential in this pandemic, particularly in a number of ways from identifying the first disease clusters back in December of 2019, to actually contributing to the screening and tracking of patients, to managing the measures for reducing transmission. And there is no doubt that it will continue to be at the center of attention in the post pandemic world, probably with the role that it will play in vaccine development and drug targets development. AI benefits, however, were obvious and have been shown pre-pandemic. And so really, these events have shown us that crises have a potential to accelerate technology and innovation uptake in a tremendous way. And so, I would really now start the conversation with our panelists to ask my first questions to them and that is to provide from their perspective specific examples of how artificial intelligence is being used to tackle this pandemic at present. Purang, would you like to start please?
Purang:
Sure Federica. Thank you so much for chairing the panel. In BC, thanks to the generous support and funding provided by Canada’s Digital Technology Supercluster, we at UBC in partnership with Providence Health Care, Rural Coordination BC, Vancouver Coastal Health, as well as local industry, including Clarius and Change Healthcare, have developed IM POCUS, which is an initiative and a collection of 80 point-of care ultrasound imaging devices that are distributed across our province. These devices allow the physicians across our vast community to scan patients and monitor the effect of pandemic, including COVID-19, on the patients at large. And specifically to provide timely advice in terms of the clinical management of patients. These decisions can be especially very critical when we have patients in rural regions for which receiving the results of the PCR test, for example, can be very long. So in this specific case, AI is actually providing decision support for these physicians to acquire the right data that can be used for action that can be very critical for patient management.
Federica:
Thank you. Helia?
Helia:
I’d like to add to Purang’s points. Over the past few years, we’ve seen a rapidly growing number of healthcare organizations looking not only to deploy new technologies but also to develop their own digital solutions that use data and AI. There are a number of applications that I can think about, applications of AI during pandemics such as this one. Solutions from triaging and symptom checking by health bots, for example, which are text-based or voice-based chatbots, all the way to genomics research on the virus and vaccine development. The recent solution that we worked on in partnership with Providence Health Care and UBC is around PPE monitoring to identify the individuals for not wearing their masks or not wearing it properly. And also crowd counting where you can actually use IoT technology in combination with AI to identify the number of individuals in the facility, for example in the ER waiting room, and this is while the privacy measures are also in place.
Helia:
This practical solution is using Microsoft IoT solution and it enables this type of analysis on edge, which means the computation and analytics happens at the IoT device or the camera in this case. Another great solution that I can think of is we are working on our text analytics for medical domain. Using the solution, you can comb through large volumes of research papers and documents and find keywords and do knowledge mining through natural language processing. And this is particularly important during this pandemic because there are massive volumes of documentations and research papers that physicians and researchers and healthcare responders, they have to go through on a daily basis. And this can save them a lot of time.
Federica:
Absolutely. Handol, over to you.
Handol:
Thank you very much. At Variational AI, we’re actually working on developing a therapeutic to directly address SARS-CoV-2, which is the virus that causes COVID-19. So what we’re doing is we’re using AI to discover new molecules, small molecules, that will bind to the site that will hopefully lead to a biologically favorable reaction, so no less than trying to find a cure for COVID. It’s a very different effort than looking for a vaccine and the wonderful progress that’s been made by players adopting completely new technologies and new modalities, whether it’s mRNA or different technologies in developing a vaccine in record time, is absolutely breathtaking and is tremendously inspirational.
Handol:
We’re working on the therapeutic side and on the small molecule side as opposed to, let’s say AbCellera, they’ve done some tremendous things leveraging AI as well as microfluidics and various other platform technologies to bring a therapeutic out in record time. So what we’re seeing right now is the age of rapid responses driven by this massive global pandemic and crisis forcing innovation to happen in a timescale that we’ve not really seen before since maybe the Manhattan Project. And this, hopefully, is a much more peaceful use.
Handol:
What we’re doing is we’re directly looking at novel molecules and using a generative machine learning approach that we take is we’re training on data that was actually designed for the SARS-CoV virus, or what we call SARS classic, which ravaged parts of the world about 15 years ago. But we stopped, but we have a lot of data that’s been available that’s in the literature and freely available, we train on that. Plus in partnership with the Vancouver Prostate Centre and our PI, Art Cherkasov, as well as with adMare BioInnovations, we’re all working together, and this project is being funded by the Digital Technology Supercluster. So we’re seeing some really, really great results and we’ve already generated some candidate molecules, which we’re now ordering and sampling and we’ll be testing biochemically. And the idea is to have a viable small molecule therapeutic ready to move forward in development.
Federica:
Thank you Handol, this is really interesting and you made some very good point actually because, certainly, this pandemic has demonstrated to us that even though, yes, there’s been an acceleration in innovation, such as the vaccine development, which is unprecedented and spectacular, the reality is in order for us to cope efficiently with such a virus whose transmission is so high, it’s really a number of approaches, which is, of course, the vaccine development, but the therapeutic as well as the public health measures that we need to adopt in everyday life and that. So that was great touching on that. So these were fantastic examples of how artificial intelligence is really helping tackle this pandemic now. Thinking a little bit fast-forward in the future, I would like to ask you what is your opinion with regard to what sort of application you see artificial intelligence dominating and driving in a post-pandemic world, not at present, but in a post-pandemic world. Purang?
Purang:
This is a very interesting question. I mean, many of us have observed the radical change in our lifestyles, as well as changes that digitization has brought to us in a very accelerated way over the past nine months. Many technologies that were planned to be happening over the next 10 to 15 years had to happen in a very, very rapid pace. And as a result, we don’t have the luxury of time where we were originally planning to transform our healthcare at a much slower, I guess, way than what we were anticipating today. One of the things that I can see happening is a huge demand for access to care in general. And the care that I personally anticipate would happen post-pandemic is a shift from centralized healthcare to more of a community-based care or patient-empowered care. And this is very critical because of the rapidly growing, I guess, strain on our healthcare budget that cannot sustain the way it’s currently being operating. And this transformation, or basically digital transformation of care, and a distributed healthcare system will be a key to actually open new pathways to improve patient care and also increase patient outcome.
Federica:
Very interesting. Handol?
Handol:
Just to elaborate, and I’m in full agreement with Purang, and I’d like to actually take it one step further. My background is in information technologies and we saw what we call the consumerization of the enterprise, and so email, phones, computers, et cetera, et cetera. And we see the rise of the cloud, and I’m sure Helia will address that from the Microsoft perspective. But if we could look at the consumerization of health, we’re already in an age where we have multiple devices, whether it’s a watch or a phone or these other wearables that will track a staggering amount of data over time on an individual basis that’s used by the individual. And I think the individual is the ultimate arbiter and steward of their own data and how that’s used, whether it’s specifically for them or the ability to pool in the context of public health.
Handol:
Then to enable different treatment options, I think what we’re seeing is no less than the growing disruption, potentially, of the way that therapies and treatments and patient care is both planned, executed, and distributed. And at the center of all of that is data and the ownership of the data that’s generated. And it’s no longer the touch point when you’re talking to a clinician, but it’s essentially on an ongoing basis. And I think that will enable the explosion of the use of artificial intelligence or machine learning or various other data science approaches to potentially unlock insights and options on a far more economical basis for the greater good of the taxpayer and also as us in society. And I think that is a really exciting potential trend that I hope will continue to accelerate even after the pandemic is over.
Federica:
This is great. I love this take on the shift towards patients who have been much more active in the participation of their healthcare. Helia, I would love to hear your take on this.
Helia:
These are fantastic points raised and I can see precision medicine continuing to evolve and grow. The start was not after COVID, it was decades before COVID. Genomics research has been a great influence and we have great examples across the country, UHN and BC Cancer are great examples of bringing precision medicine to the care. And I can see with, for example, genomics data, you require large volumes of data storage capabilities and compute. So each of the genomic files, you can look at gigabytes of data per genomics file and you need thousands of core hours to process that data. So just for one patient, so comparing to when you want to look at population health, you need to have massive amounts of data from across the world. And that is really enabled through the power of cloud because you have basically unlimited resources, unlimited storage, high-performance storage, unlimited high performance compute, access to clusters of CPUs and GPU-powered machines. And that’s really one of the enablers of precision medicine being continued after this pandemic.
Helia:
I can also talk about another application of AI that could also be important post-pandemic. AI can help governments and health organizations determine rapid growth areas and prepare the Canadian workforce with skilling and direction to meet those growth needs. For example, LinkedIn Economic Graph is a great resource of data with analytics across all of Canada on job seekers, open jobs, skills inventory, which are being updated real-time by the labor force themselves. So AI across citizen and healthcare services data could help us potentially be more proactive in terms of planning services and resources to match needs and deliver in more efficient ways. I can think of population health and syndromic surveillance, these could be other great areas of growth, and with more and more connected data sources, researchers would be able to study the population and find clusters with similar symptoms, and potentially predict another pandemic. Hopefully that never happens, but this is really what basically are required to stay after this pandemic within the healthcare organizations.
Federica:
Great. This is really interesting actually and I saw Handol nodding all the way through it. Would you like to comment Handol?
Handol:
I think, and Federica, you know this in your role at Genome BC is, I mean, you’ve got 3 billion base pairs. You’ve got a massive amount of data and you’ve got all of these omics and multi-omics datasets. And then you layer in all of this other information and I’m willing to talk about the fact that we need data because I work in machine learning and if you don’t have data, you can’t do anything, so the digitalization like Purang was saying is the first priority. But then, at the same time, we can’t draw this false equivalency that just because there is a lot of data that this is a problem that’s right for disruption by machine learning.
Handol:
Once you have a massive amount of data, then the real work starts happening and that’s where we need that innovation on the algorithmic side. And this is where a strong requirement of an academic and a research-based community, such as at UBC or SFU, in machine learning is really, really important because now that we have all of this data, there’s a lot of noise out there and it’s really hard to derive that signal. And you’re going to burn up the planet running GPU hours to find nothing, right?
Handol:
We have to take a balance and especially us in the AI world, we kind of talk about the automation and a gradient descent-based form of finding insights, but we also have to make sure that the experts and the human experts are there and working hand in glove. Otherwise, what you find is that a lot of AI projects, you do all of the stuff and you say, “Here’s the result.” And someone who’s skilled in the art will go, “I could have told you that.” And we have to make sure that we don’t waste our time doing that. And so data, a lot of data is great, but simultaneously or slightly lagging, we need to invest in that algorithmic excellence in order to develop the next wave of models and approaches.
Helia:
Maybe I can add one more to this challenge. This is fantastic point Handol. So not only having that data, but clean and usable data. I can also add another layer to it when to process this type of data, you typically are relying on your local HPC, high-performance computing environments, which oftentimes are not enough for, for example, genomics researchers to leverage and use. And you have to wait in very long job queues, especially researchers or PhD students or postdocs who don’t have enough time to run their algorithms, which takes a year.
Helia:
An example I can provide is one of our researchers in the Canada Research Centres, they actually leverage cloud to reduce the computation time from almost a year to 17 hours, so this is really one of the enablements. So not only the data and the innovations and the algorithms that the researchers are using, but also the tools that are required for them to do it in much more accelerated way. So when we think about AI and applications of it, there are so many different layers that we can think about, such as privacy and security, but at the same time, compute, the data and how relatable and how, basically, clean and usable that data is.
Federica:
We are getting into really a data conversation now, which I was looking forward to, which is all about the data that makes this artificial intelligence really possible. So I’m going to start with Purang, but obviously I really love to have a bit of a conversation on the challenges of data because, of course, we do know that in order to work effectively and efficiently, as Handol has already alluded to, we need good quality data. We need lots of data, but we need also good-quality data. And so there are some challenges, and I was hoping that we could perhaps go into a little bit more details about this.
Federica:
Before we do that, I just want to make a point that I’ve heard in this conversation from all of you, especially when you were giving the examples in the last conversation, that you’ve made a point that is quite important, I think, and that is the collaboration, the early collaboration, between academia and industry for this algorithmic development, and particularly for this innovation to actually develop. And I think, for these early stages of innovation, I think this is a point I just want to make. Purang, where do we stand with regard to access to the quantity and quality of health data in British Columbia, Canada in general possible, which are necessary to really leverage artificial intelligence potential?
Purang:
This is a very interesting point, and I also want to elaborate on the points that were brought up by my other panelists in a sense that what pandemic showed was that our healthcare system is a bit fragile. I mean, data silos all around the province that need to be integrated so that we can have a very rapid response to a crisis such as what we are facing today. As a result of this need to harmonize and integrate all of those data silos, what also pandemic showed was that people were willing to talk to each other in a very collaborative and collegial and partnership way so that a lot of barriers in terms of coming up with actionable predictions for our society could be eliminated. And what I hope is that this kind of collaboration and communication across several parties within the province and at the national level would continue.
Purang:
I do hope that the data, including the healthcare data, as there’s many other data sources that are required to benefit our society, would be harmonized with proper regulations shared with people, including scientists at UBC, as well as industry and other partners, so that we can actually make a better society. One key advantage of BC itself in this whole journey is that it is not a very large province in a sense that healthcare system is very radically distributed. And at the same time, it’s not as small so that the validation of many of the healthcare technologies cannot be meaningfully performed. So I think we need to build on the trust that we have built with each other within the journey of the past nine months and actually work together to come up with proper regulations and proper ethical and security discussions that can allow us to share data for the better good of the society.
Helia:
I cannot agree with you more Purang. We all know that data exists in silos, but we recently have been working and having great efforts and we see big shifts to more consolidated data repositories. Examples are the Rare Disease Data Consortium in Ottawa, one petabyte of genomics data for cancer research from BC Cancer, DNAstack’s COVID cloud, which includes sequenced genomic data of the coronavirus from across the world that are currently deployed on Microsoft Azure. However, we need to approach this thoughtfully with new technologies that allow us to share and collaborate in new ways while maintaining privacy of health data. And full understanding, and impact assessment must take place, which includes security and policy and privacy of stakeholders around the table together with the technology providers to ensure common goals and common understanding of risks, and to really define steps to take to reduce risks to acceptable levels based on each of the use cases. We actually have a responsible AI leader for Canada, John Weigelt, who has these conversations directly with government and healthcare organizations to help them move forward.
Federica:
Thank you very much Helia. Handol, a final comment?
Handol:
I think that I’ll go a little bit more radical, which is the ownership of the data about you and your genome or your health conditions should be owned by you. And I think that if we take an atomistic view that you are, again, the most reliable shepherd of your own interests. Now, that said, we have to have literacy in terms of what data means and what health means because you have literacy around your credit rating, you have literacy around not posting so much stuff on Facebook because they’ll take that and sell it to advertisers, so you don’t want to have the same thing happen with your health data. But also, that’s balanced against the potential benefit to you as an individual, but also to the healthcare system and the ability of public health to go in and recognize and identify trends.
Handol:
I think there’s a bit of a balancing act and there’s no easy answer, but I think that the average person needs to get more literate about both data and that is not just for healthcare, but in general, right? And I think, what we’ve seen in social media and all of these other blowbacks on GDPR and all of these other things, I think there is a sense of privacy being a very important concern, but we also need to balance that against like, “Hey, there’s some benefit over here as well.” But we have to be able to draw a straight line. And I think that the individual person is generally going to be the most reliable steward and people are going to overshare and what have you, but I think if we can work on that, use methods like federated learning or different other methods to anonymize the data, there are a lot of ways and tools that we can use in order to get the best of both worlds.
Federica:
Thank you. I think the entire panel agrees with all of the things that have been said and with what you’ve just said. I think it’s so important, so let’s just shift the attention on data but also becoming more literate about data, which I think everyone across different fields has to do, not just the artificial intelligence field really. It’s a collaboration, it’s a multidisciplinary collaboration, but we all have to be literate about data, data standardization.
Federica:
The way we are generating data, specifically within this pandemic right now, it’s all in different ways. And we know that different type of datas ultimately are going to cause even more noise and confusion when the algorithm are going to try and extrapolate any meaning from it. So we do need that interoperability, standardization, we need sharing for sure. And those are great points and I just wanted to outline them again. And moving just to the last couple of questions because we are almost running out of time, but I really want to hear from you, what can we do to accelerate this, to remove these barriers and accelerate the potential and the promise of artificial intelligence and its benefit to healthcare?
Purang:
I think, as it was brought up, one of the key things about creating an impact within the healthcare industry is really to have a partnership. And the partnership in BC is actually fairly uniquely positioned in a sense that we have to bring in fundamental AI scientists, applied AI engineers, the local industry, healthcare authorities, as well as the regulators all together within a very unified platform so that the individuals such as physicians who are in the field can come up with the right requirements for a healthcare product. And those set of requirements can be filtered all the way to a fundamental AI scientist who can then create technologies that can rapidly integrate it. So this whole harmonized ecosystem of innovation is very critical for creating a very accelerated impact. And I think BC is very uniquely positioned given, again, its size and the trust that everybody has in the community with each other to actually make such an impact at the provincial and the national level.
Federica:
Fantastic. Thank you Purang. I think that’s a very important point. Helia?
Helia:
I can say an accelerated change can be done through education and skilling to understand what is possible through AI. And this means for all stakeholders, not just technical individuals. To unlock the promise of AI, we need to understand art of the possible in context of business or organizational goals and really all functions. And define it in our own terms, business value of AI. And when I look at every success story across Canada, I see a common contributing factor. There’s an internal AI champion who’s not only understanding the domain, but also is familiar with what AI could empower them with.
Helia:
We encourage skilling and education for all levels, specifically for decision-makers. And we have complementary training resources, such as our AI Business School modules, which includes building an AI strategy, cultural readiness for AI, responsible adoption of AI, and AI tech overview for business leaders, these are just to name a few. And I can add that experiments are interesting, but unless we can define real business value in our use cases, it will not catch on and we will lose momentum. We all know that success breeds success and if one organization can define and prove out a highly valuable use case, even a small technically enablement-oriented or a small AI use case, more people in the organization will start to see AI as an enabler and look to implement that particular solution into their own functions as well.
Federica:
Thank you. I love this and I’ve heard everybody at the table, every stakeholder at the table, education, skills. Handol, do you have, very, very quickly, a few more to add?
Handol:
I think that we have to remove the fear of AI. I mean, people think that there are robots coming to take their jobs. And that goes for PhD researchers that we talked to who are firmly in the domain of analog, right? And I think a lot of that is generational, but it’s very similar to the internet. In the late ’90s, you can replace the internet with AI and it’s exactly the same thing. We’ve seen this movie before. So I think that there’s a generational change, there’s change management that needs to happen. And then just in terms of people who are going to retire. Okay boomer, your time is up, now the Generation X people and then the Millennials will come in and they’ll have a different interaction with the technology and the adoption will happen almost organically.
Handol:
Yet at the same time, there is this tremendously potentially disruptive thing about AI, which really could potentially take a lot of jobs in different industries. And it’s not like, necessarily, every single job is going to be automated away, but there is a real potential for that and we have to be sensitive to that. And we have to make sure that the improvements that we adopt are, on the net, beneficial for society. And that’s a responsibility of the industry as well. We’ve seen with big tech right now, you can’t absolve yourself from your social impact, especially if you’ve become extremely successful. So we have to be mindful of that as we introduce this technology.
Federica:
Thank you very much for your take. This is a really interesting question and a complex problem. And thank you for bringing it up because this is going to be, really, my last question really for this panel. And that is that, yes, with all of the benefits that we have spoken about, unfortunately, artificial intelligence is also met with and seen as a threat and as a risk possibly for these job losses. But I would like to propose to look at AI as a complementary, as an added value. And as you say, that will require and necessitate a change of skills, education, but also a change of management, a change of leadership. There are certainly qualities that are very human, like critical thinking, strategic thinking, creativity, that cannot be replaced by AI. So I would like to hear finally from Purang and Helia on this topic before we close our panel. Purang?
Purang:
I couldn’t agree more, Federica. This is a very interesting point, and at the same time, as you mentioned, a very sensitive issue. One thing though I can emphasize, especially in the context of the healthcare, is that even though there are predictions that certain jobs could be automated, but the need in our healthcare system for a much larger workforce across various healthcare occupations is actually significantly outpace the pace of the automation. But, as was mentioned in terms of education, what we anticipate to happen is that several of the jobs that exist today, they need to be restructured and possibly talked about in a new way so that the physicians are more empowered by the use of the AI, giving them faster access to data. And support them with AI implementation actually can help them faster stratify patient care and, at the same time, provide an environment that continuous innovation actually would lead to improved patient outcome.
Federica:
Thank you Purang. Helia?
Helia:
As we all know, AI can be one of the world’s greatest tools for good. And it requires that we approach this with the ability to work with others, build on what AI has to offer, and really address the problems at hand. And this is why we introduced AI for Health, along with other programs at Microsoft. And we really believe that the development and deployment of AI must be guided by the creation of an ethical framework. When we include AI-powered solutions, there comes the responsibility to be ethical. We need to study the social impact and potential human harm. We need to develop a more transparent solution and be able to question the machine learning models. And at the same time, think about accountability, to be able to undo unintentional harm that may have been as a result of biased data or data that wasn’t inclusive or private and secure.
Helia:
We actually have a dedicated team specifically to ethical AI to support our customers and partners with designing fair, transparent, accountable, private, and ethical AI. But this journey must be a collaborative discussion between government, healthcare, industry around each use case so that we can fully understand potential positive and negative impacts, privacy impacts, and so on and so forth, and work together to mitigate risks while also drive transparency with users of AI and also beneficiaries of AI, so that they build that trust. And in short, healthcare professionals and organizations, of course, they see risks everywhere, but we need to fully understand all of the solutions available to reduce or mitigate risks. And this doesn’t happen until we work through very specific details of each use case together so it’s really an ongoing collaboration and education and learning from one another while taking an ethical AI approach that can mitigate concerns and risks.
Federica:
Thank you very much, this was a fantastic closing statement. And yes, as this pandemic has already shown us, for every global challenge we like to tackle, we can never be myopic about it, we have to consider the multi-dimensional aspects of the problem in a holistic way, really address this. And I think this was really well said by Helia. You have been extraordinary. We have come to the end of our time and thank you very much, it was an excellent conversation. Naomi, back to you. Thank you.
