Thought Leaders in Artificial Intelligence: Sonny Tai, CEO of Actuate AI (Part 2)

Sramana Mitra: Does it work in alerts?

Sonny Tai: We send alerts directly to the customer in any way they want. We integrate with their existing security protocols. It depends on what the customer security system looks like. It could be that they have a security operation center which means that they typically have a staff of a few guards who are monitoring intelligence in a room full of big-screen TVs pulled up.

We also have some smaller clients like schools and other organizations. They don’t have a security operations center. We send text message alerts to their security staff.

Sramana Mitra: Where are you seeing traction?

Sonny Tai: The most traction we’re seeing so far is with a segment called remote video monitoring. It’s a pretty new segment. As you might imagine, if a company wants to provide video monitoring service to customers, it’s impossible for a human being to monitor numerous cameras at the same time. AI is really changing that paradigm. Strong computer vision technology now enables you to identify when a feed has an object that has a high probability of being a threat.

A big part of our business right now is selling to these types of customers. We help them use fewer people to monitor more cameras, which is a strong economic value proposition to them. They could provide a better service to their customers.

Sramana Mitra: When you say remote monitoring industry, what do they monitor?

Sonny Tai: They monitor anything. It runs the entire gamut.

Sramana Mitra: You are not selling directly to the end customer. You’re selling to the service provider.

Sonny Tai: We do sell directly to the end customer sometimes. It depends on their security protocols. A lot of them are increasingly using remote monitoring companies. It runs the entire gamut. Some of them are schools, construction sites, or car dealerships – anywhere you want an immediate response without the expense of having many security guards.

Sramana Mitra: Every time we talk to an AI company, there is this chicken and egg. An AI model needs to be trained on real-world data. For you to be able to develop the model, you needed to access real-world data.

Sonny Tai: We started as a gun detection company. We built AI models to identify when a person pulls a gun out. In the US, there are often these mass shooting incidents. When we started looking for data online, if you went on Google, you would get photos of guns, but they come from stock photos or action movies. It doesn’t replicate what a security camera would see.

We had to create a lot of our own data and scrape a lot of our own data. The first thing I did was to partner with a YouTube influencer. He had a crowdsourced repository of 2,000 gun crime incidents. He passed it over to me. Our data science team parsed them and annotated them. We also create a lot of our own data.

This was all early-stage hustle. We took a lot of photos from security camera-type angles of us holding fake weapons. One of our angel investors has a personal collection of 200 guns. He let us borrow several of those. Just by doing that, we had a model that ended up working pretty well.

As we landed more and more customer deployments, we were able to collect more and more customer data. We’d use that data to re-annotate the data and retrain our model. Nowadays, we don’t create our own data anymore. It’s all sourced from deployments.

This segment is part 2 in the series : Thought Leaders in Artificial Intelligence: Sonny Tai, CEO of Actuate AI
1 2 3 4