Colin Angle

Are we building robots to do everything for us, or to help us do more of what makes us human?

Last week, I had the opportunity to speak at the "Physical AI: Minds in Motion" Summit at @MIT where I reflected on a question that has become increasingly important to me over my 35+ years in robotics: Just because we can automate something, does that mean we should?

Today, we have an extraordinary new toolkit in AI and robotics. But as we race to automate more of our lives, I think we need to ask a more fundamental question: Are we using these technologies to replace human capabilities, or to expand them?

Imagine a robot that cooks dinner for someone who can no longer do so themselves. That's useful. But imagine a robot that helps that person continue cooking for the people they love. That's profound.

At @FamiliarMxM, we're exploring what it means to build robots that act as catalysts for human connection, agency, and purpose, rather than simply replacements for human effort.

The goal, after all, is to enable human connection, not replace it.

Thank you to John Werner and the @imaginationxyz team for bringing together such a thoughtful group of people to explore the future of physical AI.

Full talk: youtube.com/watch?v=31iHQc…

I believe one of the most important capabilities for the future of Physical AI is the ability to understand and interact with people.

A robot can have extraordinary physical capabilities. But if it can't read the room, recognize people's needs, or behave appropriately in a social environment, it's going to struggle to find a place in our daily lives.

At @FamiliarMxM, we're building the capability to address this challenge. Our technology is designed to enable robots to observe, understand, respond, and remember, so they can interact with people in socially and emotionally intelligent ways.

Our first application is a Familiar, a robotic companion I've dreamed of building for years. A presence that gets to know you, supports you, and becomes part of your daily life.

But the underlying capability has much broader potential, from robots supporting families and older adults, to machines that work alongside people in homes, hospitals, and workplaces.

I talked about the origins of the Familiar and the thinking that led us here at TEDxFargo.

Watch the full talk here: youtube.com/watch?v=obDJOg…

I've always believed that some of the best ideas come from questioning the conventional way of doing things.

When we were building Roomba, the idea of having a robot autonomously clean your home seemed pretty unconventional. Today, millions of people live with one.

Now, at @FamiliarMxM, we're exploring another possibility: what if robots could do more than perform tasks? What if they could understand and respond to people, develop relationships over time, and become a meaningful presence in our lives?

It's a very different way of thinking about what robots are for.

I talked about this approach to problem-solving at TEDxFargo: youtube.com/watch?v=obDJOg…

Meta’s Muse and OpenAI’s Dots show AI starting to move beyond answering questions and into taking action.

Put that kind of agency into a machine that shares our physical world, and the challenges and risks increase further.

A robot in a home, hospital, school, or workplace has to read people. It has to understand tone, attention, mood, relationships, timing, and personal boundaries. It has to know when to approach or stay back, and how to behave differently around a child, an older adult, or a group of people.

A lot of that judgment requires social and emotional intelligence.

There’s enormous investment going into teaching machines how to perceive, reason, move, and manipulate the physical world. Yet the industry is still underestimating something fundamental: machines that operate around people need to understand people.

That’s the AI platform we’re building at Familiar Machines & Magic: social and emotional intelligence for #PhysicalAI.

We believe it will be critical for the category to succeed.

Data centers may be leaving earth.

Every major technological era eventually creates pressure on infrastructure that was built for the previous one. Electrification changed grids. Cars changed cities. The internet changed telecom infrastructure. Now, AI is changing compute and energy. 

With #Suncatcher, @Google is exploring putting AI compute into orbit, using solar-powered satellites - next week, we'll see the first in-orbit test of AI chips.

There’s a lot of invention still to come around the systems that make AI possible. Our own AI team at @FamiliarMxM has been rethinking how to maximize performance for #PhysicalAI running on the edge.

I’m excited to see what other assumptions get challenged as we figure out what it takes to bring Physical AI into the world at scale.

Image credit: NASA