6 AI Startups Doing Physical Security
Table of contents
Last year, a BBC journalist put China’s massive surveillance camera system to the test. The country has about 175 million cameras, powered with artificial intelligence technologies like computer vision that can recognize faces and keep an eye on activity. Once the journalist’s photo was uploaded to the system, it took authorities about seven minutes to track him down in a city of 3.5 million people. In 2014, machines surpassed the identification accuracy rate of the human eye, according to a co-founder of Sensetime, one of the top AI startups in China that develops facial recognition technology used in those closed-circuit TV (CCTV) camera systems. Artificial intelligence isn’t just for tagging photos on Facebook these days. Physical security systems are increasingly using AI to watch people and protect high-value assets like your vintage Star Wars collection.
It’s a lucrative market, of course. The worldwide value of physical security products in 2016 reached about $28 billion, for example. China plans to add nearly a half-billion more cameras to keep its 1.3 billion people “safe” in just the next few years. Computer vision is one of the key technologies enabling much of the AI physical security tech today and many companies are working on various solutions. It’s already ubiquitous, from the creepy prompts on Facebook to tag your so-called friend Steve to the ability of your phone camera to pinpoint a smiling face.
We tasked our most paranoid MBA, locked safely in his hotel room after a recent visit to the cannabis clubs in Europe, to track down some of the most interesting startups doing AI physical security.
What’s Under the Hood
Founded in 2016 with offices in Tel Aviv and New York, UVeye raised $4.5 million in Seed money last July. You would expect a company from the Startup Nation, surrounded by less-than-friendly neighbors, to come up with a physical security system that scans the undercarriage of vehicles for threats like bombs, weapons or non-Kosher food. High-resolution cameras create a 3D image to detect any anomalies as small as a USB thumb drive. Advanced image processing and machine-learning algorithms go to work, producing results in seconds. The UVeye system is available as both a stationary platform for places like border crossings or a mobile platform for ad hoc events like presidential inaugurations. The company has even leveraged the technology into a secondary market for automotive shops to detect damage or problems with vehicles.
Eyes in the Sky
Founded in 2015, San Diego-based Shield AI raised a $10.5 million Series A last April, to bring total funding to $13.1 million. The recent round included top VC firm Andreessen Horowitz, and their tech also earned them a spot of the CB Insights AI 100 list. Shield is building an artificially intelligent drone for surveillance and reconnaissance. Among the founders is a former U.S. veteran who served in Afghanistan. He thought a robot capable of navigating different environments, from buildings to caves, without a remote pilot would be a useful tool on the battlefield to protect service members or civilians. That has scored Shield contracts with the Department of Defense and the Department of Homeland Security.
The Hivemind Nova is a quadcopter capable of doing that and more, using its AI software to constantly improve, based on lessons learned from individual drones, the drone collective or through external sources such as video feeds and even social media. Shield AI has also developed a trainable neural network (Hivemind Argo) for real-time detection in video of specific people, vehicles or other objects. Yes, the machine wars are coming.
Caught in the Act
Another Series A startup to make it on the CB Insights AI 100 list, Deep Sentinel out of San Francisco raised $7.4 million last year from Shasta Ventures, Lux Capital and Amazon’s Jeff Bezos. Founded by serial entrepreneur David Selinger, who did a stint with Amazon back in the day, Deep Sentinel is developing a home camera security system that detects crime before it happens. It uses the same sort of AI tech behind self-driving cars and Facebook facial recognition algorithms to spot suspicious activity and sound the alarm. For example, it might spy a guy in a hoodie getting ready to make off with your Amazon package sitting on the front porch and send an alert. The system is reportedly 90 percent accurate, with plans to add some sort of deterrent. Like maybe a killer drone from Shield AI.
Walk this Way
Founded in 2013, Evolv Technology out of the Boston area is on a mission to protect the world. Toward that end, the company has taken in nearly $30 million, including an $18 million Series B last year whose investors included Bill Gates and Lux Capital. The company has developed a body scanner it calls the Evolv Edge, which can detect firearms or explosives without all of the fuss involved in going through security at the airport.
Each machine can process 600 people an hour, which frankly doesn’t sound all that fast to us if you want to use it at a subway station during rush hour. It does come with facial recognition technology that can alert security personnel when a person of interest walks through the scanner. The company is also developing a more robust security system for your home compound or evil lair, the Evolv Mosaiq, which combines cameras, sensors, and biometrics such as fingerprint scans for real-time security.
Check this Out
You might have seen the headlines recently about the new Amazon Go store in Seattle where you scan your smartphone, shop, and walk out without ever having to interact with a human. No more cashiers in this future vision of grocery shopping. Hundreds of cameras follow your every move, virtually putting items in and out of your online shopping cart as you move through the store. The charge hits your Amazon account after you walk out the door. Obviously security is a big deal, and while we don’t know much about Amazon’s system, there are startups out there that have developed AI platforms that don’t miss a single transaction.
Everseen out of Ireland, founded in 2007, has raised $13.5 million in two separate rounds last year. The company’s AI platform, timi.ai, watches every transaction at the point of sale to detect fraudulent behavior or errors, which it says costs retailers up to $40 billion per year. The company spent six years training timi on millions of transactions and in-store customer behavior to understand the “DNA” of the retail and checkout environment. The company’s claims its solution is used by five of the world’s 10 largest retailers.
Another player in this space is Cambridge, Mass.-based StopLift, which raised money privately, so no word on investors or total amounts. The company’s checkout vision system analyzes video in real-time and compares it to transaction data at point-of-sale. The system can see when someone doesn’t scan an item in self-checkout or spot a practice called “sweethearting” where a store employee does the same thing or covers up the barcode to help a friend get free stuff without paying for it. ShopLift came out of a Harvard Business school project, leveraging computer vision research from MIT to automate surveillance checkout.
Conclusion
A few things become obvious as we start to look at the future of AI and physical security. First, there’s nowhere to hide. Seriously, we’re screwed when the machines begin to hunt us down. Second, while physical security may be increasing with AI, job security is decreasing. One obvious threat is to the 3.5 million cashiers in the United States or the 1.1 million security guards staring blankly at video surveillance monitors. And we haven’t even re-visited the issue of robotic security guard startups like Knightscope, which despite a recent PR disaster success involving terrorizing homeless people in San Francisco, raised $25 million this month. Third, we’re seeing some very reputable investors pour money into the space, not to mention the nearly $640 million China’s Sensetime took in last year alone. You don’t need AI to tell you to follow the money.
Share
