The January 6 riot at the U.S. Capitol has sparked discussion around various surveillance technologies and if they could have been used to defuse the mob before things got out of control. This is a good discussion to have, but it’s the wrong use case.
Now that Congressional hearings have begun on the matter, it’s pretty clear that anybody with functioning eyes and ears could have identified trouble was brewing at the Capitol that day. There was significant preliminary planning chatter on social media, the FBI issued a warning the day before that never made its way up the Capitol Police chain of command, and people showed up to the preliminary rally dressed in tactical gear, carrying everything from baseball bats and hockey sticks, to brass knuckles and bear spray, with minimal security response.
Anyone at the event (or even watching on TV) scanning the crowd and listening to its aggressive chants could determine it was a potential security risk. There really was no need for technology to identify the obvious. And the idea that facial recognition or, yes, gun detection could have played a meaningful role in stopping the riot is not founded on solid ground.
The use case in this situation is pretty straightforward: crowd control. But there were no processes in place to, well, control the crowd. And when humans fail to put that infrastructure in place, no technology can swoop in and “save the day.” As we now know, law enforcement was understaffed and unprepared for what happened that day, so technology-generated security alerts would have just added more noise to the chaos.
However, in cases where technology is used as part of a greater security and safety framework, it becomes a force multiplier for humans. For a better use case, let’s look at a large concert venue. Gun-detection technology at entry points enables guards to identify and mitigate potential threats before they are inside the venue, while also sparing everyone else from having to wait in line to go through the dreaded metal detector (and walking through again when they fail to remove the change from their pockets).
Inside the venue, artificial intelligence (AI) can proactively analyze video feeds to monitor the crowd to identify potential trouble spots, well in advance of humans reacting once the disturbance has occurred. By baselining “normal” and continuously correlating thousands of data points, the AI system can enable security personnel to be alerted to likely incidents early in their development, rather than discovering them after they’ve already exploded. For example, if people are suddenly moving away from or toward a single spot in the crowd, it’s likely some sort of disturbance has just started – a fight, someone having a medical issue, etc. Security staff can be alerted to this so they can intervene before the incident progresses into something more serious. Or, if a group of people wearing similar jackets or jerseys are lingering after the event has finished, it could be a gang or just a group of the opposing team’s fans that security should take a look at to determine if any action is required. And, of course, if someone walks into an area where they shouldn’t be, or behaves in a threatening way, AI can detect those anomalies as well and give security a heads-up.
You might ask, “Why use AI rather than just having security guards monitor video feeds?” Because AI is better at monitoring feeds over long periods of time. Humans are very good at understanding context and making fast decisions – for example, when someone drops to the floor in distress, it’s pretty clear it’s likely a medical issue and 911 needs to be called ASAP. Humans are very bad, though, at monitoring an overwhelming amount of information over long periods of time and identifying anomalous events. We all get bored and distracted very easily – in fact, a few years ago a Microsoft study found that humans have about an eight-second attention span – shorter than goldfish! This is where AI can be spectacular. It can consume enormous volumes of information (say, dozens of video feeds at a concert venue) and identify the “needle in the haystack” anomalous events. Then, it can notify its great decision-maker partner – venue security – to investigate and do the right thing.
By preventing weapons from entering the venue, combining the strengths of AI and people, and having pre-established protocols in place for investigating and responding to events, crowd control becomes a far more manageable exercise than simply having security personnel wandering the venue, hoping to be in the right place when trouble starts. If the pre-riot rally had these kinds of processes and protocols in place, with controlled points of ingress and egress, and large numbers of additional personnel available for security, then technology could have played a pivotal role in defusing the situation at the U.S. Capitol. Alas, none of these things were in place – so it’s the wrong use case for threat-identification technology. (And don’t let any technology company try to convince you otherwise!)