Dave Selinger, co-founder and CEO of Deep Sentinel, on VatorNews podcast

The company combines AI and human intervention to help detect threats and deter crime

Steven Loeb and Bambi Francisco Roizen speak with Dave Selinger, co-founder and CEO of Deep Sentinel, a security company that uses artificial intelligence to prevent crime. It deploys a series of cameras that are powered by deep learning, along with a real-time human element, to evaluate threats on a property within seconds.

The company has raised $23.4 million from investors that include Meyer Global Management, Nationwide Ventures, Basis Set Ventures, Lux Capital, Bezos Expeditions, Shasta Ventures, and Vishal Rao. 

Highlights from the interview:

• Selinger has been working in AI for the last 20 years, including running the AI team in the early days of Amazon with Jeff Bezos, who he says instilled three traits in him that have become core to his career, the first being drive. The second thing he learned is to trust and listen to data, even if you don't like it, and the third thing is the concept of applied artificial intelligence, meaning not in a lab or in a university setting, but using it to make money.
• Deep Sentinel came about because, about five years ago, there was a huge upheaval in the capabilities of AI. While AI has consistently over promised, and under delivered, for over 30, starting in around 2012 it finally started delivering, with a new form of deep deep learning. At the same time, Selinger's neighbor had a home invasion, causing him to look into the space and to realize that the existing solutions are ineffective, so he decided to apply AI to security.
• The company combines artificial intelligence with human intelligence, in which the AI filters out 99% of the threats, so the police aren't called by someone driving by or a dog or a tree blowing in the wind. If something is detected, the AI will inform Deep Sentinel's guards who work remotely. It's the AI's job is not just to make sure the guard is looking at the right things and the right events, and not looking at the wrong ones, but also to nudge the guard towards the right conclusion with each of the events that the guard reviews.
• It takes less than 10 seconds on average from the time a person enters the field of view of the camera to the time a guard is interacting with them. For comparison's sake, ADT won't even notice that somebody is on the property, typically until they've already broken the door or window and then typically it's between three and five minutes until they respond. 
• Small businesses are the most exposed and they are the ones who can't recover from a large break in. They also are the ones that have the least ability to afford a high end security solution like a guard sitting out front. Deep Sentinel has found that for small businesses it is the only legitimate solution for their problems.
• Selinger hopes to follow in the footsteps of Tesla, and their self-driving technology, when it comes to AI. Tesla announced a style of AI called self-supervised, or semi-supervised, learning, which means the AI can get better all by itself, instead of needing a scientist or an engineer. Semi supervised and self supervised learning are able to interact with the world around them and learn from it. That is the direction Deep Sentinel wants to go as well.
• Deep Sentinel measures ROI in two different ways, the first being dollar amount, and decreasing losses, which Selinger says is really easy to measure. The second way to measure ROI is the emotional side: if you own a business, and you have minor vandalism every single week, that's emotionally taxing, and it causes businesses to lose employees, or to have to pay additional hazard pay. The ability to transform that sense of fear from the employee base to confidence is soft ROI, but it's the ROI that customers talk to the company about most. 

Thanks to our sponsors:

Subscribe to our podcasts to get our interviews and shows as soon as they're published!