In March I was at the ISC West show in Las Vegas and to say the show was a success would be an understatement. This year’s buzzword and products centered around AI, or artificial intelligence. By the end of the show, I was pretty sure that AI not only could be used in security applications but might also make you breakfast and tie your shoes. And the truth is AI is going to change the way we do things, but before we just throw out everything and let ChatGPT run our lives, let’s take a pause and see what we can actually do today vs. five years from now.
Almost all AI applications in security today center around video, which for this column’s purpose I am also going to include thermal, LiDAR, radar, WiFi and a few others because the AI is looking for the most part at human or item recognition based on the shapes and movements. In reality while they are being called AI, they are really single-purpose analytics applications (e.g., is there a human, dog or weapon, etc., and variety of other targets).
Most of today’s AI products do not take into consideration anything else about a security incident other than shape of the object, which leads us to the age-old problem of how do we determine intent of the human we detected? Today, we must rely on humans to try and determine intent by watching and or listening to the object detected. Most of those humans doing this are monitoring center specialists making the split decision of, is this person detecting a threat to life or property or from a law enforcement perspective is there a crime being committed?
Take a typical home that has cameras in the front and rear, it’s daytime and the monitoring center gets an alarm that a human has been detected in the back yard — the person seen is wearing jeans and hoodie with the hood over their head so you can’t see the face. Now we must determine is this person supposed to be here, is this just a kid and friend of the family, is this person a burglar or service worker there to fix the AC … there are a dozen reason why this person could legally there and another dozen why they should not be there, so what do we do now?
Some pundits would say send the cops, some would say observe and see what they do, others would say call the homeowners and see what they want to do, others would say do a voice down or turn on a siren or sprinklers and still others would say do nothing and move to the next alarm. Remember this was about AI, “I” being the intelligence part; however, these response options really don’t seem extraordinarily intelligent to me, after we have been doing for the past five decades with normal alarm systems.
We have several basic business considerations we have to work though first. We can’t just send the police on everything, doing so will strain law enforcement resources. In today’s monitoring business, we can’t just watch this person for an extended amount of time, we can do talk downs and turn on sprinklers, but what if this person is supposed to be there? These are not easy choices to make, but we must do something that is consistent and repeatable, and more importantly we have to meet the expectations of the subscribers and we have to make sure they are part of this process.
Today’s AI does an amazing job at detecting humans, and there are times when I look at video and you can barely see anything in the bounding box. Knowing the difference between a dog and a bear may not seem interesting for most, but if you own a home or business where bears are present knowing one is near is a big deal. My point is that what we have today does an amazing job of detection, but the future AI is going to answer the question of what is that person in the back yard really doing there — that’s when we bring intelligence to the traditional security application.