3 Tips on how to think about robotics for your business from the CEO of Embodied Systems

Staff Writer Jan 05, 2023
Alexa Fund Robotics Amazon CES 2023

Paolo Pirjanian is the founder and CEO of Embodied, a next-generation robotics startup that is a part of the Alexa Fund. The company’s Moxie robot was voted as among the top 100 inventions of 2020 by TIME magazine. Moxie is a lifelike, believable robot that interacts with children using empathetic eye contact, realistic body language and facial expressions, and is designed to foster a sense of connection. 

“The robot is designed to develop ‘EQ’ skills in children,” says Pirjanian. “Through play-based interactions, missions and tasks, Moxie helps children do things like label their emotions, regulate their emotions through breathing exercises, and express gratitude.”

MOXIE’s therapeutic framework that leverages established and evidence-based therapeutic strategies has been shown to have a positive developmental impact on children with mental behavioral developmental disorders (MBDDs) in as little as six weeks. 

Pirjanian has been involved with robotics for all of his career that has spanned nearly three decades. He has conducted research for NASA’s Jet Propulsion Lab, in addition to serving as the CTO of iRobot, the company behind the revolutionary Roomba vacuums.

In this article, Pirjanian shares three tips for business decision makers looking to build or incorporate robotics and artificial intelligence into their operations, products and services.    

Paolo Pirjanian, CEO, Emodied

1. Design for different modalities of human-computer interaction

“Robots are often depicted in popular media as autonomous machines that imitate and often exceed human intelligence – not always with the most pleasant consequences. However, it is more practical to think of robotics as a human-centric engineering discipline, one where technology doesn’t supplant human intelligence, but complements it by helping make up for our limitations and biases.

This human-centric view drives every facet of product development for Moxie, especially when it comes to enabling multi-modal interactions. 

Moxie’s goal is to help develop and empower mindful and confident children. For this to happen, we had to ensure that Moxie is able to make eye contact with children. 

We didn’t want to model behavior where people are not able to make eye contact. 

As a result, we developed computer vision algorithms to ensure that Moxie was able to locate the child – which can get tricky when the child is seated in a group – and make eye contact. However you don’t want Moxie to maintain eye contact for an excessively long period of time, which could make for an unpleasant and almost dystopian experience. We developed a whole grouping of animations where Moxie looked at you, turned away for a second and then looked back again. 

Voice is another important facet of developing multi-modal interactions. Here, we had to leverage the state-of-the-art in natural language processing to develop a system that enables Moxie to identify where the speaker is located, distinguish between pertinent and background noise, and understand the context of the conversations.

We recognize that in a majority of use cases, robots don’t have to be as attuned to the intricacies of human emotion as Moxie. However, business and technical leaders should always view robots within the prism of human-computer interaction, and not as autonomous entities. Even warehouse robots have to interact with humans as they carry out their tasks, and thoughtful human-centric and multi-modal design will help robots carry out their duties to the best possible extent.” 

2. Focus on what’s important to your customer

“It’s vital that you zero in on the “must haves” when it comes to your customer. For Moxie, it was imperative that we kept the cost low. And because we work with children, it also meant that we could make no compromises when it came to privacy. 

Our refusal to compromise on cost and privacy meant that we had to overcome significant scientific and technological challenges that we could have overcome by adopting more standard solutions. 

For example, we had to be thoughtful and clever when it came to selecting the optimal number of motors responsible for gestures and body movements. Incorporating a high-powered GPU from an external vendor would have increased the cost of the robot to over ten thousand dollars. As a result, we had to invest engineering resources to optimize our software to run on a low-end, low-cost, “off-the-shelf” processor. 

Our focus on privacy meant that we had to overcome several software-related challenges as well. We invested resources to make sure that every piece of data is encrypted. Because every image taken by Moxie stays on device, we couldn’t use many of the standard cloud-based computer vision algorithms. That meant that we have to invest resources in developing mechanisms that help ensure images never leave the robot. This is an incredible challenge to overcome cost-effectively  because computer vision can be highly compute-intensive.  

As anyone who has developed products and services for children knows, you can’t even remotely have anything inappropriate. This includes obvious things like the usage of inappropriate words. But it’s important to remember that we live in an incredibly diverse world. A Bart Simpson reference might be totally acceptable in one family, but totally unacceptable in another. This meant that we had to design and train a neural network on all of the topics that might be considered inappropriate.”

Focusing on areas that were important to our customers allowed us to understand how and where we should allocate our valuable technical resources.” 

3.  Leverage innovations in the broader scientific community

“Many of the most significant advances in the field of artificial intelligence have been made possible as a result of collaborations between academia and the industry at large. There are several startups, enterprises and organizations that have leveraged innovations from initiatives like the ImageNet Large Scale Visual Recognition Challenge or the development of the Robotics Operating System (ROS). 

You should build your own custom in-house solution when there’s no alternative. However, when possible, you should also work with the best provides in the world and utilize their services.

Voice is one example of an instance where you don’t have to reinvent the wheel. When we launched Moxie five years ago Alexa didn’t have some of the capabilities we were looking for. However it’s an entirely different story today – there’s such a strong community of scientists, businesses and developers building on Alexa. The service keeps getting better over time. So it makes a lot more sense to harness all of this innovation, and free up your resources to delivering a great customer experience.” 


Also at CES

Alexa Innovators: Endel
How Spinn is elevating the coffee experience and the next generation of hardware
Tripp CEO Nanea Reeves is building the largest meditation community in the metaverse