As governments attempt to mitigate the spread of the 2019 Coronavirus (Covid-19) by limiting travel and quarantining people, health authorities are following a well-known playbook. We’ve seen global outbreaks recently with the SARS outbreak in 2003 or H1N1 (swine flu) in 2009, which led to very different outcomes - SARS was almost entirely eliminated within a year due to an effective and coordinated program of isolating patients, while H1N1 settled and became a seasonal strain controlled by vaccinations with occasional ongoing flare-ups. But this time, the global effort has been emboldened by a new tool - Artificial Intelligence and Machine Learning are much more common now, and the advancements in these technologies are enabling better, faster responses.
For instance, take the development of vaccines, treatments and cures. Traditionally, this is a long and difficult experimental process that involves isolating samples of the virus and using trial and error with thousands of potential antivirals to see if they are effective. During the SARS outbreak, it took roughly 20 months for a vaccine to be developed and ready for clinical trials, by which point the outbreak was already over. This time, AI techniques are being used to find patterns between Covid-19 and other viruses that we already know about, so that the scope of any search for a cure can be reduced. This is dependent on understanding the virus and its genetic makeup, and we can now analyse that much faster as well. Baidu has reported that their AI tool has cut the processing time for analysing part of the coronavirus RNA from 55 minutes down to 27 seconds. Continuous sequencing is needed to help detect mutations of the original virus, so this isn’t a one-time task and time is critical. The most promising solution for developing a vaccine uses that genetic sequencing information to design a direct solution computationally, which also reduces the risks and logistical costs of obtaining and testing on actual virus samples. A couple of different labs say that they will have a vaccine ready for clinical trials within one to four months.
However, the challenge with vaccines is that trials, manufacturing, and deployment can still be slow. There are a few groups working on developing virtual clinical trials, where new medications and treatments can be tested in AI-powered simulations, but we are still a while away from having sufficiently accurate and reliable models of human bodies. Manufacturing of any vaccine will likely be limited to the hundreds of thousands or millions of doses per month, and a lot of coordination is required to make sure those doses get to the most needy areas and are deployed effectively. There are AI tools being developed that aim to help by optimising vaccination strategies to mitigate spread as much as possible, but this is still a long-term solution.
For now, the strategy has to focus on containing the virus as much as possible. With modern transportation moving people quickly, diseases can jump large distances and appear in unexpected places. By picking up cases early, authorities can act fast to confirm and contain, and if done effectively this can suppress the disease entirely. However, containment is disruptive and can have negative long-term impacts, so decision-makers need to be careful about where and when they decide to quarantine people. This is informed by surveillance efforts, with health workers on the ground reporting cases as they happen. AI tools are now augmenting this by monitoring news and social media, which can help capture cases outside of formal health systems and corroborate data collected from other sources. Artificial Intelligence may also be able to help model and predict how diseases spread, although this research needs a lot more data and it will be hard to know how reliable these tools might be. It’s important to note that AI isn’t doing this alone - human analysts are still interpreting the results of any algorithms and making the judgement calls, and balancing data collection and aggregation against privacy concerns.
Text mining of social media using natural language processing can also help decision makers monitor the morale of people in quarantined areas. Importantly, misinformation about the disease can also spread very quickly and lead to more chaos, so AI tools are being used alongside human moderation by Facebook to remove harmful content. Mapping tools built on top of AI analysis can also be shared publicly to help keep people informed about the spread of the disease, although this should be accompanied with some analysis and guidance to help people understand the risks and manage expectations.
Lastly, there are a couple of AI-enabled tools being used on the ground by frontline staff. Computer vision techniques can be used to process thermal imaging data to help detect individuals with virus symptoms, and the use of drones can help survey large numbers of people at once. Physicians can also use AI-enabled diagnostic tools to help differentiate Covid-19 from other types of flu or disease based on the symptoms and test results, although currently this is no replacement for lab-based testing of samples. The hope is that lab-on-a-chip diagnostic tools will be able to provide fast, remote results so that samples don’t have to be sent off site and healthcare workers can respond immediately.
None of this is to say that AI technology has provided a silver bullet that stops disease outbreaks and epidemics. The processes involved in controlling the spread of disease are still very human, but AI tools can help provide those humans with timely information for decision making. The spread of disease has significant negative impacts on people’s lives, so it is important that we can leverage all the tools we have available to mitigate the harm. In the future, we may see fewer and fewer outbreaks occurring as epidemic control protocols evolve to take advantage of new technology.
This Lunch and Learn webinar provides a detailed overview of reinforcement learning, its applications and use cases, and a selection of resources you can explore to learn more about reinforcement learning.
What is the edge? What are some of the key considerations to explore when you're planning an edge AI implementation? How can you best leverage the edge to give you the best business outcomes? Read on to learn more about how the edge brings AI to your fingertips.