LAS VEGAS – During last year’s Milestone Integration Platform Symposium (MIPS) in San Antonio, there was a lot of discussion about what had become the proverbial 800-pound elephant in the room for the video surveillance industry, which was product commoditization and the subsequent realities that must be confronted by manufacturers and systems integrators surrounding it. At this year’s symposium being held this week in Las Vegas, the company offered a deeper look at what could serve as a possible antidote to this margin erosion quandary: harnessing the power of artificial intelligence (AI) and other sources of intelligent data.
This is certainly not a new concept within the industry and the emerging capabilities presented by AI have been written about ad nauseam, however; the fact remains that very few end-users are taking advantage of these advanced feature sets or even some of the more basic tools that are widely available today to help people optimize the video data they are generating. According to Kenneth Hune Petersen, Milestone’s Chief Sales and Marketing Officer, who delivered a keynote address during the event, almost 90 percent of the company’s installations are security only deployments.
“There is so much more we can do,” Petersen says. “Why isn’t there, in every big shopping mall, smart parking assistant systems installed so when someone comes on Saturday morning and it is really busy, it could help them find a parking spot easily? Why aren’t there facial recognition cameras at the front of every store so we can use (demographic) data to enhance the customer experience? If you see these things are not installed, it is our obligation to talk about it.”
Although some people have written off technologies, such as AI, in the past as being unrealistic or not impactful enough for their industry in the near-term, Petersen pointed out that many of the futuristic devices depicted in the movie “Minority Report,” which was beyond its time when it was made in 2002, are outdated by today’s standards.
“What seems futuristic today can be reality tomorrow,” he says. “And the good news is we are part of this phenomenal industry where, if we work together, we can change the world. If we dare to dream together, we can make this a better place because I can guarantee you that what took us here today is not going to get us there tomorrow.”
Indeed, Petersen believes there is “massive business opportunity” for the industry in AI due to the fact that most of the world’s high-tech giants are investing heavily in the technology as well because only half a percent of the world’s data is currently being analyzed and used. In the same way that Amazon is using currently available technology to power its Amazon Go grocery stores, Petersen says integrators should be using the Milestone technology stack to build the video deployments of the future.
“I urge you… to see if there is gap out there that we haven’t found or seen and then go for it,” Petersen says.
Demystifying Artificial Intelligence
One of the highlights of the two-day conference was an opening keynote address from Tanmay Bakshi, who despite being only 14 years of age is already an accomplished computer programmer and expert in the burgeoning field of AI. Bakshi, who also works as a Watson programmer, sought to clear up some of the misconceptions surrounding AI and explained, in detail, many of the underlying technologies that comprise the field.
According to Bakshi, the classic definition of artificial intelligence as machines being able to think and make decisions like a human is not really what the technology represents today. Instead, he says it is characterized as any computer that shows a sign of being able to do something that most people would think requires human intelligence. Bakshi adds that machine learning most closely resembles this classic picture of AI as it features algorithms that enable computers to learn by themselves without any hard-coding or engineering on the part of humans, which is the way Watson operates.
Underneath that, there are neural networks, which essentially try and simulate the way the human brain works in a computer. There are several ways in which neural networks can be trained, according to Bakshi, including through genetic algorithms, which simulates natural selection or evolution in computers, as well as by using deep learning algorithms that are neural networks stacked on top of one another. Bakshi says that deep learning algorithms are already impacting hundreds of fields and hundreds of thousands of lives. And then there are deep reinforcement learning algorithms that try to combine how our brains work with the way we psychologically learn through rewards and punishments.
One of the biggest challenges as it relates to implementing AI and machine learning today, according to Bakshi, is a lack of labeled data. For example, take a video surveillance deployment at a large bank. Such a system recording footage from a multitude of branches would generate a ton of data within only a few days, however; no one has the time to go in and drag boxes around people, vehicles, objects, etc., to be able to classify them and thus train the system. And while some data can be crowd-sourced, other is confidential and must remain within an organization. Bakshi says there are also significant gaps between machine learning experts and domain (industry) experts. For these reasons, Bakshi says IBM is working on something called PowerAI that will empower domain experts to create more and better data much more quickly for the purpose of optimizing machine learning algorithms, reducing the time it takes to train these systems from weeks or months to a matter of days.
While some are fearful that AI will one day replace humans, Bakshi says those fears are misplaced and that the true goal of the technology is to augment human capabilities. “AI is not meant to replace us. In fact, the entire purpose of AI is that since it was created by humans, it should amplify our skills, augment our capabilities and allow us to do what we already do more efficiently, accurately and painlessly,” Bakshi says.
In reality, Bakshi says that AI is really starting to look like IA, “intelligence augmented,” by taking our ability to understand abstract concepts, such as emotion, communication and natural language, and combining those with abilities that computers have, including intense mathematical computation capabilities.
Industry in Transformation
Bjorn Skou Eilertsen, Chief Technology Officer for Milestone Systems, told attendees that it is “inevitable” that all devices will be connected and that all systems will use artificial intelligence in some way. In fact, he says the surveillance industry is already transforming as a result of these trends through three waves of innovation in the form of aggregation, automation and augmentation.
According to Eilertsen, aggregation refers to the increasing number of sensors and systems being connected with one another. “We will be creating almost incomprehensible datasets and we will be able to capture video and information from anywhere and about anything,” he explains. “In fact, we are creating the big data of video surveillance.”
With regards to automation, Eilertsen says we’ve already aggregated tremendous amounts of sensor information with thousands of cameras and other devices deployed in some cases within a single application, which makes trying to use such a system virtually useless without some kind of automation tool. “We need the systems to be automated, we need the systems to be autonomous and we need the systems to help us navigate all of that input and information,” he adds.
That leads to a third wave of innovation in augmentation, which is driven by the emergence of AI because human capabilities need to be augmented to handle these large datasets. “We need to make humans capable of navigating the systems of tomorrow and that augmentation is where we go beyond human capability and start connecting an aggregated dataset with an autonomous system and augmenting human capability,” he says. “That transformation is happening within the industry.”
Concurrent with these industry transformations, Eilertsen says that Milestone is also transforming the way it thinks about products and platforms and is investing to accommodate these changes as they permeate the market. For example, Eilertsen says the company will be advancing device drivers and developing a framework that enables integrators to connect not just cameras but all of the various connected devices deployed at an end-user's site to make improved data aggregation a reality.
On the automation side, the company has developed a Video Processing Service (VPS), which sits on top of graphical processing units (GPUs) to enable the company’s solution partners and integrators to build AI and deep learning algorithms directly into the Milestone VMS platform.
“We’ve been developing that for a year now and we’re going to bring it out later this year,” he says. “This is our answer for how we bring these autonomous systems into our world.”
Embracing Change
Of course, change doesn’t come easy for any industry, let alone physical security which has seen its fair share of overhyped solutions in the past. However, the technologies necessary to support this increasingly intelligent world of devices are in place and ready to support this brave new world of artificial intelligence.
According to Adam Scraba, Global Business Development Lead for GPU maker Nvidia, recent advancements in computing, particularly as it pertains to what can be accomplished now using GPUs, have implications for a number of industries, not the least of which is security.
“In security, I think we’re going to solve public safety and operational efficiency (challenges) in ways that are really impactful,” he says. “This isn’t a science project. These solutions are in production today and I think they are truly going to make the world a safer place.”
For 40 years, Scraba says the computing industry has been riding that incredible wave that is Moore’s Law, which states that every 18 months transistors will double in CPUs. So, every 18 months, computers have doubled in speed and performance. However, with bottlenecks in physics, Scraba says that wave is finally coming to an end and performance is plateauing. In addition, Nvidia realized about 10 years ago that the GPU, which was initially designed for the gaming industry, had hundreds of cores in it that could be developed into a supercomputer.
“Today, our GPUs have 5,000 cores in them. That’s a hundred teraflops of compute performance in every GPU,” Scraba says. “That not only makes up the deficiencies of what we lost with Moore’s Law, but it is also on an accelerated curve.”
This evolution in GPU computing power has thus enabled the advancements in neural networks that have made AI solutions a reality.
While Milestone currently invests significant time and R&D dollars creating features that help humans better understand their environment as part of an effort to improve the security of a facility or campus, Tim Palmquist, Milestone’s VP for the Americas, posed the question of what if the time comes when machines actually do all of this autonomously?
“The good or comforting news is none of that is going to happen yet. At least not tomorrow,” he says. “But what we can say is change is coming.”
He added that the security industry is no stranger to change and disruption as evidenced from the initial introduction of the IP camera in 1996 and the migration to network solutions that continues today. The next great industry disruption on par with this, according to Palmquist, is machine learning and that instead of fearing this change; he believes the market needs to embrace it.
“Ultimately, (IP) did disrupt the industry and for the last 20 years we’ve seen that innovation and we’ve seen it come from outside the industry. Companies from outside the traditional industry brought technologies in, new household names evolved and some household names didn’t make it over,” Palmquist says. “Change creates opportunities but it also creates challenges and I think we all together want to be one of those successful household names that continue on versus one that is referred to in the past tense.”
About the Author:
Joel Griffin is the Editor-in-Chief of SecurityInfoWatch.com and a veteran security journalist. You can reach him at [email protected].
