It feels like every year some new wave in technology goes into vogue. When we started OMX, it was all about niche marketplaces – usually the “eBay of something obscure and specific”. When I was on CBC’s Next Gen Dragon’s Den, it felt like every startup that pitched was the “Uber of something” such as food delivery, at home cooking and prescription delivery.

Learning and Intelligence

This year, undoubtedly, we have been hearing a lot about “machine learning” and “artificial intelligence”. The launch of the Vector Institute, a research facility devoted to AI, was announced in Toronto a month ago, and then Uber followed immediately after, announcing they would be opening an AI research hub for driverless cars, also in Toronto. Those terms seem to get thrown around a lot, but I think it is really important we get an understanding of how they can specifically be adopted into our processes and supply chain across the country to make industry more competitive and bring better solutions to end users in the defence sector.

Let’s back up a little. What exactly is “machine learning” for starters? Wikipedia, the source of all knowledge, defines it as “the subfield of computer science that, according to Arthur Samuel in 1959, gives ‘computers the ability to learn without being explicitly programmed.’ Artificial intelligence, however, goes a step further and is defined as “intelligence exhibited by machines.” In computer science, the field of AI research defines itself as the study of “intelligent agents,” which can be defined as “any device that perceives its environment and takes actions that maximize its chance of success at some goal.” Far too often the term “artificial intelligence” is applied when a machine mimics cognitive functions that humans associate with other human minds, such as learning and problem solving, which I believe is the root cause of a lot of people’s concerns about the field of AI and what this means for our future, especially if we hit that scary intersection of man and machine.

Singularity

The term “singularity” is also often thrown around, which is the hypothesis that the invention of artificial superintelligence will abruptly trigger runaway technological growth, resulting in unfathomable changes to human civilization. Essentially, if we develop machines that can learn on their own, they will far surpass human thinking and then remove themselves completely from human control.Stephen Hawking commented on Reddit that “if AI becomes better at designing AI than humans, we’ll hit an intelligence explosion that will ultimately result in machines whose intelligence exceeds ours by more than ours exceeds that of snails.” The father of the singularity hypothesis, Ray Kurzweil, estimated this would take place by 2045. This intelligence explosion would change everything about our world as we know it. And we would be stupid to ignore its possibility.

If the creepy “machine takeover” is too much for you, then there is this completely different world that lives somewhere in between today’s reality and one where we are snails in comparison to our computers. Besides the fact that there are many scholars who push the singularity out to up to 1000 years from now. As my hero, Walter Isaacson remarked: “the 0-1 (black and white) nature of machine programming is very complimentary with human thinking, and machines will still always serve humans.” Just because they will have more processing capability than us, does not mean they will develop consciousness and take over. I believe that in Canada, and in the defence sector, they have the ability to be our biggest competitive advantage. And I believe those areas of competitive advantage will be primarily in non-sexy, traditional areas of the supply chain. Areas that, with applied machine learning and AI, will catapult to increased efficiency and capability.

Local heroes

So, I sought out some of the heroes in Canada, those innovators working in this area of machine learning and artificial intelligence, and talked to them about how their technologies could be applied to move the dial in defence.

While most think of AI as being something new, it’s been over 10 years that Menya Solutions has been developing AI solutions for the Canadian Navy. They typically work in collaboration with key defence contractors such as Lockheed Martin Canada, Thales Canada, CAE, MDA, and Fujitsu Canada. Froduald Kabanza, a professor of artificial intelligence at the Université de Sherbrooke, founded Menya Solutions in 2007, and the company currently has 15 full-time employees in Sherbrooke, including artificial intelligence research scientists and developers.

Menya Solutions recently launched HybridLogic, a suite of artificial intelligent algorithms designed to help humans, robots and drones analyze a tactical situation and make timely decisions during defence operations. This technology includes intent, capability and opportunity analysis algorithms that process sensor data, human input, and a prior knowledge to predict threats, infer their goals and plans, and understand their vulnerabilities and the risks they pose to protected assets. HybridLogic also includes algorithms that generate a course of actions against recognized threats and support the execution of these actions by monitoring contingencies and revising the plan accordingly. The algorithms solve a variety of situation-understanding and decision-making problems to provide artificial intelligence capabilities to agents (humans, robots or drones), operating alone or in multiple coordinated units. The algorithms take into account various constraints related to defence operations in particular, including spatial constraints, communication constraints, computational constraints, and resource constraints.

From left to right: Julien Filion (AI scientist and developer), Francis Bisson (former collaborator, currently software developer at Google), Philipe Bellefeuille (AI researcher and scientist), Simon Chamberland (AI scientist and developer), Froduald Kabanza (Founder and AI scientist).

“Having humans collaborating with robots or drones in defence operations is no longer a science fiction concept,” said Froduald Kabanza, CEO of Menya. “The ability of drones to interact naturally with humans and to comprehend their environment will largely depend on their proficiency to recognize the goals and plans of other agents – a proficiency that technologies like HybridLogic are meant to provide.”

Kabanza went on to explain that, on the other hand, the unprecedented complexity level of defence operations brought by these technology developments is such that humans can no longer control everything without the aid of artificial intelligence algorithms at various levels. “Companies developing command and control systems for defence operations can use HybridLogic algorithms to automate the analysis of threats and support the awareness of human operators to monitored situations,” he said. Companies who are in the virtual reality training solutions area for defence operations can also integrate HybridLogic to present enriched training scenarios, according to Kabanza, “involving simulated entities that can make autonomous decisions.”

I’ve always said that data is very powerful, but it always comes down to “so what,” and using data as a lever to determine the answer to your problem and drive facts-based decision making.

Another company making waves in AI is Motsai Research, an electronic hardware design company, which extracts meaningful information from miniature motion and environmental sensors.  By combining the most advanced integrated electronic wireless technology and pushing the limits of sensor miniaturization, they create sensor solutions that acquire, process and classify motion patterns within the sensor itself, which incorporates custom electronic hardware, advanced sensor fusion algorithms and assisted machine learning to create powerful information-generating systems. This enables an important reduction in system complexity. Motsai is really focusing in on supporting the defence sector in areas such as soldier protection, machine operating conditions monitoring and advanced warning of abnormal situations. Novel miniaturization techniques, encapsulation methods and energy harvesting solutions allow sensors to be integrated inside objects and systems in ways that were not possible only a few years ago, which is driving growth in the Internet of Things market. But the next step is making sense of the data and using it to drive machine learning, which is exactly what Motsai is doing.

This image provides insight into the size of the sensor modules Motsai uses to collect data.

Sensors are at the frontier of cyber physical systems and will provide better situational awareness and improve decision-making abilities of future defence systems.  For this potential to be realized, future sensors must provide meaningful information and be fully integrated within larger systems using secure and robust communication mechanisms.  Machine learning and artificial intelligence can help leverage the potential of distributed sensors at many levels, and future systems will very likely integrate sensors in large scale deployments.

On the academic side, McMaster University has quite the history in the evolution of smart systems, including being creators of cognitive approaches to processing data from diverse sensors and the development of efficient optimization techniques. McMaster is working today to leverage this strength and bring a smart-systems approach to a broad range of industries within Canada in areas such as Nanoscale research, which uses machine and deep learning algorithms to provide quantitative insight into materials microstructures. They are also using machine learning to help with weapon detection, building more intelligent vehicles, and healthcare.

On the big data front, McMaster is working with IBM to improve big data quality by designing and building intuitive software tools to keep personal information private while automating this process for organizations, which saves them time, money and improves data analysis results. In the battlefield, they can apply big data to troubleshoot processing problems, and to collect and analyze data from a variety of sources, including temperature sensors, pressure-flow monitors, chemical-concentration indexes, and digital cameras. This data is then projected to a computer system that analyzes the information and changes the controls to ensure better products are created. It sounds like there are a lot of opportunities for this academic institute to collaborate with industry.

When I think about machine learning, I almost immediately think about autonomous vehicles. You may not have heard of this company, but Cohort Systems develops computing solutions that convert any vehicle into an autonomous vehicle. The solution is a mission execution system that supports deployment of hundreds of missions from which the autonomous vehicle will continuously review, select, and execute a mission that will achieve the most important objective for the robot, based upon the current knowledge of the robot’s status and the environment it is operating in. The company is initially focusing on perimeter security, where they can replace manned patrol vehicles with unmanned patrol vehicles, which is suitable for base security, border security, airports, or power stations.

It makes a lot of sense to implement machine learning technology for something like perimeter security as it is labor intensive, expensive, can be dangerous, and is a role humans do not do well at because it is just so monotonous. Unmanned patrolling costs less, is safer, provides a higher quality of service, and is more responsive to changes in the environment because the autonomous vehicle is always alert and communicates status instantly. Sometimes it just hurts to hear the robots are better than us in some areas, but we have to hear it.

RANK Software is a Toronto-based company that is pioneering the use of big data analytics for cyber security solutions. One of RANK’s differentiators is the use of deep learning. While the application of neural networks is not new, it is innovative in the cyber security field due to the limited availability of labeled/example datasets. RANK has developed an industry-first application of deep unsupervised learning to detect suspicious user strings.

Visualization of RANK software dashboard.

“Many security offerings provide the basics of anomaly detection and simple response whitelisting via Machine Learning,” said Mohan Rao, CTO, RANK Software. “At RANK, we are trying to solve the tough machine learning problems of true learned human-machine interaction, inferred decisions, and fuzzy intelligence. Our goal is to close the security skill gap through automated threat response.”

In the area of cyber security, machine learning and artificial intelligence technologies can help companies stay protected against ever emerging cyber threats. This is especially true where instances such as the recent spread of “WannaCry” ransomware have shown the limitations in traditional methods and processes for detecting and responding to cyber-attacks. Through its solution called VASA (Virtual Advisor for Security Analytics), RANK monitors user and machine behaviour across multiple dimensions in real time. This deeper visibility allows VASA to detect compromised behaviour and enable rapid response, even if the ransomware is mutating and changing rapidly.

As the defence industry and its supply chain face a global shortage in skilled security workers, they will need to trust technology to offer automated and/or prescriptive real-time response capabilities to keep up with the exponentially growing attack surface. Attack vectors are no longer external only; as seen with the Snowden document leaks, unauthorized internal access can be as costly and have longer lasting impacts than external single instance attacks.  RANK’s machine learning and artificial intelligence are the foundation of its threat monitoring capability, which has resulted in a partnership with Lockheed Martin Canada to create innovative cyber security solutions for future Canadian and international defence programs.

“The future of threat monitoring and investigations of all types relies on our ability to successfully develop and deploy solutions that leverage data science, behavioral analytics and machine learning to quickly identify and prioritize risk,” said Niranjan Mayya, Founder & CEO of RANK Software.

Lastly, a new company that I just met recently is Dat-uh, a Toronto based startup. The Founder and CEO, Humera Malik passionately explained that “Using the power of machine learning, we have been able to tackle the complicated challenge of making data science scalable for industrial companies with our automated analytics platform.”

Malik said that the Dat-uh platform is helping defence manufacturers accelerate their big data strategies with real-time predictions and insights that move past the limitations of human-managed data models and leverage automated Machine learning models that are self-learning, based on AI. Personally, I see the opportunity for this application to be used right across the manufacturing sector in day-to-day operations to create cutting-edge plant operations and production processes, which will ensure Canada’s place in developing the products and equipment of the future. Now that I find exciting and worth adopting technology for.

If you would like to connect with the companies mentioned in this article, go to theomx.com and search by company name.