Robot Dogs Train at West Palm Beach Firm's New AI Facility

Sit, Spot. Roll over.

Process millions of images and perform industrial business inspections better than a human would.

Good dog.

An artificial-intelligence and machine-learning firm, South Florida-based Levatas trains robots like Spot to strut around collecting data and identifying safety issues at clients' properties. In April, the company opened its state-of-the-art industrial AI lab in its West Palm Beach office, which serves as Spot’s mechanical puppy training pen.

"Companies are looking for a machine that can walk on four legs… and is engineered to be reliable enough to run autonomous missions," Ray Popp, head of communications for Levatas, says. "It can take a photo and our software will process what it is reading."

Levatas started in its founder Chris Nielsen's garage before expanding into a Palm Beach Gardens office. At first, the company offered custom software design and consulting, helping clients with their websites and e-commerce platforms.

Nielsen credits the company's chief technology officer, Daniel Bruce, with refocusing Levatas' business model on computer vision and AI-enhanced inspection. The company began working with the machine builders at Boston Dynamics in 2020 to give a visual brain and inspection-performing capabilities to Spot, one of the latter company's signature robot models.

"Boston Dynamics creates what we call athletic intelligence, which is their engineering fully focused on making that robot walk on tough surfaces and maneuver around," Popp says. "What it wasn’t built for was actually seeing and capturing information, which is where our software comes in."

Levatas' new industrial AI lab simulates real-life situations that may occur on Spot's business inspections, such as a missing fire extinguisher or an appliance malfunction.

When in inspection mode, the metal-tinged robot — already known for its agility, dancing ability, and eerily life-like movements — is equipped with cameras and sensors around its body to read analog pressure gauges, identify temperature changes, listen for air and gas leaks, and detect corrosion.

"If a gauge or temperature looks off, or some hardware is out of place, we have to build out AI models that are trained to recognize what the area should look like," Popp tells New Times.

To that end, data scientists and software engineers at Levatas first feed the robo-pup millions of images, familiarizing the machine with the environment so that it can avoid people and walls while carrying out its tasks.

"These robots don’t do anything without a human telling it what it needs to do," Popp says. "We’ll basically pre-record a mission and say, 'For this mission, you’re going to get off the dock, walk over here, check this gauge, read this thermal image, and maybe read 50 gauges around this area.'"

Once they assign Spot a mission, no humans are generally involved in its control functions, though there is an emergency on/off switch. The button isn’t pushed often, Popp says. In one instance, Spot fell down the stairs, flipped right back up, and continued on its way.

"It’s like when you restart your computer because it’s stuck or something," Popp adds.

On other occasions, employees have interrupted Spot’s mission for more selfish purposes.

"The challenge we’re still having is people will stop, look down, and be like, 'Oh, there’s a robot, let me take a selfie with it,'" Popp says with a chuckle.

After an area is analyzed, Levatas alerts businesses and employees of possible safety issues and workplace threats. The robotic inspector may flag an item if the results need to be evaluated by a human. The company hopes the process will offer a quicker, more efficient alert system in many industrial settings, as compared to human-only inspection.

Levatas is looking to capitalize on the growing number of corporations that want to use automated inspection machines to avoid risks to employees and unnecessary costs for menial monitoring. The company's prospective customers can use its automated monitoring tools on fixed cameras, drones, and submersibles depending on client needs.

"A lot of companies can’t find people to do these jobs anyway. They’re boring jobs, or they’re unsafe,” Popp says. "They’ve got smart people who are better educated and would rather use them for better purposes."

Customers can buy Boston Dynamics Spot robots through Levatas, which takes a percentage of the sale. But the company says its revenue model is currently based on customers licensing out software or paying Levatas to set the software up or create custom features.

Florida companies such as Publix and Florida Power and Light, as well as other clients stretching across the United States, Europe, and Asia, have used Levatas' tools to perform business inspections, according to the company.

Last year, Levatas technology was tested at chip-making giant GlobalFoundries' manufacturing center in Vermont. Samantha Garrison, GlobalFoundries' deputy director of factory automation, said she worked with Levatas to send a Spot robot made by Boston Dynamics on a winding inspection mission around the facility.

"We have a lot of equipment that's got a really long life cycle and a slow rate of change. It's not economical for us to go and sensorize every piece of equipment in this really vast facility. It's probably more practical and actually quite a bit easier for us to implement to go and use the mobile robotics approach," Garrison said in a video released by Boston Dynamics.

Theories vary on how drastically automation will affect employment, with some studies predicting that wide-ranging workforce sectors are susceptible to automation-related layoffs. A PricewaterhouseCoopers analysis in 2017 found that some 38 percent of American jobs would be at high risk of being automated by the early 2030s.

Other research has indicated that while automation will change the type of tasks people do at work, it will not have as extreme an effect on displacing jobs as previously thought. With the recent expansion of AI development for real-world applications, however, more complicated tasks than ever are being outsourced to our digital counterparts.

Some "lower paying jobs" may disappear with the growth of machine learning, Popp says, though he believes this could simultaneously open up other job opportunities.

"For instance, now we have a technician who is in charge of making sure that spot is maintained well. We also hire people to help build out the AI models internally," Popp says. "But don’t get me wrong, there will probably be some displacement to start with."

Within two to three years, Popp anticipates more corporations will utilize inspection robots in bulk, rather than merely one to execute an entire operation.

"Instead of one robot doing this 20 times, they might have 20 robots doing this 20 times," he says. "Soon we’ll get to the point where it’s just like any other kind of machine you work with."