Features
A reality check for AI

If you have visited a security trade show, attended a vendor-led seminar, or read a security industry trade publication such as this one at any point in the last few years, you will have encountered the term Artificial Intelligence (AI) and heard about all of its benefits.

Is AI a marketing slogan to entice integrators and end users, a catch-all term for a range of recent innovations, or a genuine boon to the security industry? The answer is almost certainly yes to all of these, but how to effectively move beyond these surface evaluations is a bigger question with multiple answers.

The market may be muddled

The term AI has been applied quite broadly to recent security technology; it is loaded with meaning and can have a major impact as a buzzword. However, it’s misapplication may have led to some confusion in the market.

“Generally, it is not well understood. AI is a broad umbrella term which is over-marketed,” said Josh Woodhouse, lead analyst and founder at Novaira Insights. “It is hard to ‘explain AI’.”
Woodhouse, who responded to SP&T’s questions via email, said clarification may be required, depending on how AI is being applied.

In terms of surveillance technology, AI is most often being used in three main areas, he said: to improve the performance of video analytics and video quality; to apply intelligent automation to security systems; and as an assistive tool (more on “generative AI” below).

Mike Jude, research director, video surveillance and vision applications, IDC, agreed that users generally want their applications to be AI-enabled, “but when you drill into it, it’s not clear that they understand what that means,” he said. “They simply want to see it as a checkbox on the things that they buy.”

AI may come with a big wow factor, but in some cases “it literally has no context or no meaning,” he said. “That’s kind of the challenge to the industry — providing concrete examples of what they mean by AI and how that reduces to benefits to the end user.”

Jude said AI has thus far proven itself to be most beneficial to physical security systems in terms of detection, interpretation and presentation. “Our perception is the market is getting kind of saturated on the terminology, but there are real benefits from AI approaches,” he said. “We’ve reached the point where the technology generally is mature enough that you can bake it into various things and achieve some measurable improvements.”

Game-changer

Dean Drako, president and CEO of Eagle Eye Networks, said AI represents a “complete game-changer” for the security industry, helping to usher in a new age where cameras can offer a lot more to the end user than they historically have as static surveillance tools.

“We have millions and millions and millions of cameras but nobody watches them. With AI, you can actually start to watch all of the cameras all of the time. That makes them more valuable,” said Drako.

He pointed to a range of recent technical developments and new applications that are helping to expand the security industry in a relatively short period of time, including gunshot detection, and parking lot and traffic management tools. “There’s a lot of things going on right now that are going to grow our business,” he said.

End-user reception

Ultimately, the end user is going to focus on the results they get from the technology they’ve invested in. Users will often narrow their concerns down to two main areas, said Jude: cost and complexity (is this going to make my life easier or harder as a security practitioner?).

“We’ve seen a willingness on the part of end users to pay a premium for simplicity,” said Jude. “If AI results in a measurably simpler interface with the technology… then that’s going to be very well received.”

One of the most successful applications of AI in the security world is the use of “AI-powered” video analytics, which utilize deep learning to train their models, said Woodhouse.

“This technique has led to more accurate performance compared with rules-based video analytics historically used,” said Woodhouse. “The ability to run advanced deep learning video analytics on a surveillance camera is becoming ubiquitous … the majority of new camera models shipped will have this capability within the next two years.”

Woodhouse remarked that the increased availability of computing power in recent years has supported this transformation.

“This also means the cost of running advanced, power-hungry algorithms for video analytics and other security system-wide intelligence has reduced,” he said.

Leap forward

A quantum leap forward for technology also requires a period of adjustment for both the end user and the integrator installing the upgraded systems. In some cases, AI may be a hammer in search of a nail.

Speaking at a Security Industry Association “state of the industry” media briefing held during ISC West, Mike Mathes, president of global growth at integrator Convergint, likened the situation to the introduction of IP-based camera systems a few decades ago: the technology was revolutionary, but the network infrastructure necessary to adopt it took a while to catch up.

“Today’s infrastructure for AI is data,” he said, noting that end users may not currently have the resources to cull or aggregate data such that it takes full advantage of AI-based systems. “We need the user base to evolve in a direction that allows them to be able to utilize AI.”

Woodhouse noted that integrators “are having to navigate a lot of customer education and confusion around AI for their security customers. However they are also able to offer higher performing security systems which leverage the technology.”

Next steps

The potential for AI is unquestionably vast. Eagle Eye’s Drako estimated that AI will eventually double the size of the security industry. “You need a lot of computers, you need a lot of engineering… you need a lot of work to make it happen. But people are going to want it and they’re going to demand it.”

A very recent innovation is the integration of natural language or generative AI interfaces into security systems.

“Generative AI is beginning to be used in nearly all industries including consumer tech. Security is no different and we’ve seen the first integrations of popular generative AI interfaces such as ChatGPT into security systems,” said Woodhouse. “These allow the end users and installers to interact with security systems in a more natural way even without needing to be fully trained to operate the security system.”

Drako said AI’s greatest benefit to security technology is its capacity to analyze data at scale but acknowledged generative AI’s potential to improve the user experience. It’s “the icing on the cake,” he said.

AI has seemingly been a staple of the security industry’s product development and marketing efforts for many years now but we may only just be getting started.

In baseball parlance, Drako estimated we’re currently in the second of nine innings, and there’s another decade to go before we see his prediction of a doubling of the security industry come true.

For now, at least, we should appreciate how far AI has come, strive to understand its true potential and be ready for even greater advancements in the not-too-distant future.