Motorola CTO on how AI & ML help first responders tackle crises

Image courtesy of Motorola Solutions
- Advertisement -

Dr. Mahesh Saptharishi, Chief Technology Officer at Motorola Solutions, has been involved with the topic of AI and machine learning for close to two decades. Since his phD in machine learning, applied statistics and computer vision, he has been an academic, co-founder of a video analytics company for enterprise security, and eventually became the CTO at Motorola in January 2019.

His role at Motorola is to work at “the intersection of AI, design, and human-computer interaction. We start with in-depth user experience research, translate the research insights into innovative ideal models of human-computer interaction, develop state-of-the-art AI tools (including computer vision, speech recognition, natural language understanding, and conversational interfaces), and help product teams realize these ideal designs enabled by our advances in machine intelligence.”

In an exclusive interview, he details the most interesting developments happening at Motorola’s labs, what the USD 680 million in R&D budget is spent on, how machine learning and AI have evolved, and how technology is helping first responders in addressing the ongoing COVID-19 pandemic.

Your PhD thesis nearly two decades ago was on machine learning tasks for visual object recognition. How has this field evolved over the years? 

Computation and data are the engine and fuel, respectively, for machine learning. The past two decades have seen significant increases in transistor density along with new processor paradigms tuned specifically for machine learning (vector processors and GPUs). In the last decade alone, the cost of storing data has reduced by an order of magnitude.

- Advertisement -

This dramatic increase in computing power and access to data has moved machine learning from an academic niche and the ‘stuff of science fiction’ to being mainstream computer science. With this field of study being democratized over the past decade, the world has seen an accelerated growth of new advancements, applications, and discoveries.

As a result, machine learning has become a truly practical tool to enhance user experience, increase efficiency, and enable people to make better decisions. This level of practicality was non-existent in the early part of this century.

Compared to two decades ago, very few of us can go a day without encountering some type of machine learning enabled capability — whether it is taking a good selfie with your smartphone, getting the next recommendation for a movie to watch, or ensuring that no one is stealing your identity or committing fraud with your credit card.

With new ML libraries and tools available, are we closer to managing this complexity? Is there a fundamental difference between the sorts of algorithms developed at that time, and the ones in use now? Or is there just a qualitative increase in sophistication?

With some simple directions, most people can assemble furniture, yet few are skilled carpenters. A person at home with some technical background, a computer, and access to the internet can build an ML application today. A similar application would have required an academic or industrial research institution with a compute cluster about the size of a small building two decades ago. This is a direct consequence of tools that allow developers to harness the capabilities of machine learning without understanding the algorithms or the sophistication behind them.

This ease-of-development is a double-edged sword. On the one hand, we are seeing a massive proliferation of ML-enabled products. On the other hand, measuring the performance and appropriateness of an ML approach for a specific problem still requires skill.

ML algorithms are not all equally and universally good for arbitrary problems. An easy-to-develop, yet misapplied technique can result in poor performance at best and do damage to society at worst.

Thus, there is still much work to be done in managing the complexity of responsibly applying ML.

What are your top few priorities for R&D at Motorola Solutions Chief Technology Office? What are some of the most exciting things you’re working on?

Every year Motorola Solutions invests around USD $680m in user-centric R&D. This helps us to develop the mission-critical solutions needed for tomorrow.

Some of the things we’re working on include:

  • Applications that combine artificial intelligence (AI) and human intelligence to rapidly interpret vast quantities of data, as well as new user interfaces to efficiently deliver information
  • Developing fully integrated solutions for public safety and enterprise organisations that combine voice, video, software solutions – all of which can be delivered to our customers in flexible ways, as managed and support services
  • Applying innovation to evolve the capabilities of public safety command centres. These centres provide the central point where distress calls from the public are managed and where critical information is integrated between agency databases, officers in the field and where resources are dispatched. We’re developing a variety of technologies to improve and streamline every aspect of the workflows that occur in these command centres.

In your research on new technologies, how closely and in what capacity do you work with first responders and other end users who will potentially use the system? What is the workflow like?

We work very closely with first responders and we spend hundreds of hours on ride-alongs with them, observing and understanding how they interact with technology and what pressures they face.

This approach has led Motorola Solutions to co-develop many innovations in partnership with its customers for more than nine decades now. A field of research that we pioneered and continue to follow closely is known as High Velocity Human Factors.

Basically, this approach recognises that in times of emergency, first responders have an innate ability to focus solely on the most important task at hand and to block out all other distractions.

So, from a technology design point of view, we develop practical tools that ease the burden on them, with intuitive design interfaces that reduce the mental pressure. That helps them to stay focused and to make better, faster and more accurate decisions under pressure.

How does the COVID-19 pandemic present opportunities to combine video and AI-powered analytics? How can governments and authorities better enforce some of the rules they have put in place using these tools?

Firstly, it is important to understand that Motorola Solutions develops solutions that address public and commercial safety needs while safeguarding individual privacy rights. The responsible use of analytics and data management is an essential part of our product design process.

These highly accurate tools leverage the power of video with the intelligence of AI-powered analytics. However, they do not replace human judgement and decision-making.

Authorities can apply these tools to monitor communicate and respond more effectively to protect public safety.

Video combined with AI powered analytics can assist enterprise organisations in helping their employees safely return to work, customers return to retail stores and service providers get back in business after COVID-19 lockdowns.

For example, the power of video and the intelligence of AI-powered analytics can be used to collect visual footage and statistical patterns on where social-distancing protocols have been breached and where individuals are not wearing face masks. Organizations can then be notified if guidelines are not observed and can quickly make informed decisions to address the situation.

Additionally, by combining video analytics with physical access security, organisations can better understand where any individual who has tested positive for COVID-19 has been, and which doors that person may have accessed within the workplace.

Collectively, these solutions give organisations peace of mind that they can keep people safe as they re-open their workplaces.

What will the ‘new normal’ of public safety and enterprises look like in the aftermath of COVID-19?

We’ve already seen major shifts internationally in the types of crimes being committed during the COVID-19 pandemic.

In the first phase of the great lockdown we saw a reduction in road accidents, burglaries and assaults. However, in many markets this has been offset by increases in domestic violence, cyber-crime and online fraud. One example of that has been the use of social media channels to fraudulently offer personal protective equipment including facemasks.

With much of society in lockdown, first responders have continued their work on the frontline of the pandemic and have had to experience big changes to their daily workflows and the type of services they are delivering. Now public safety is preparing for the ‘next normal’. With more people being out of work for extended periods of time there are prediction is that burglaries will increase again. So the need for first responders to help citizens will likely stay the same or even increase – but, the risk to these officers will also face greater risks to their personal health. Thus, we see an increased need for tele-presense based assistance – for example, leveraging video, drones and other tools to maintain safe social distance in the field.

Additionally, like the rest of us, many public safety agencies and enterprises have learnt which roles can be performed remotely and which ones cannot.

The increasing use of control room software has enabled enterprise and public safely command centre personnel to set up virtualised environments, sometimes within their own homes, helping them to be the ‘eyes and ears’ of their colleagues working in the field.

In the longer term, we can expect to see the economic impacts of COVID-19 flow through to government services and that might mean further pressure on budgets and resources. They will require advanced solutions to help them to do more with less and maximise the efficiency of their physical resources by combining the best elements of human and artificial intelligence to deliver better public safety outcomes.

Similarly, enterprises are considering how to build up their operational resilience, end-to-end digitization, and the future of work, among others, according to a recent McKinsey report.

How have enterprises and the emergency services sector adapted to the pandemic through innovative technologies and approaches? 

A range of innovative approaches have been undertaken to respond to the pandemic. This includes the adoption of contact free mobility and communications solutions to enable officers to maintain safe social distancing limits, smart software to enable rapidly scalable communications solutions to support the fast rollout of emergency medical facilities and the use of drones combined with video to monitor and maintain public safety and security.

Innovative technologies have seen a spike in adoption due to the pandemic, such as artificial intelligence being used to assist Singapore’s healthcare sector to keep doctors up to date on the latest information related to COVID-19.

Meanwhile, many enterprise customers are using the AI and video analytics capabilities I mentioned earlier to enable their people safely return to work after lockdowns.

- Advertisement -