Roomba creator Colin Angle unveils Familiar, a furry AI companion designed to connect, not clean

The creator of Roomba is back with a furry robot companion

Colin Angle, the maker of the Roomba and the man who helped put 50 million household robots into people's homes, is back with a new robot. But this one is designed as a companion, not a cleaner. The first robot from Angle's new company, Familiar Machines & Magic, is a dog-sized robotic pet that resembles a cross between a bear, a barn owl, and a golden retriever. It has an expressive face, with movable eyebrows, ears, and eyes, and the company calls it a "Familiar," a name meant to evoke folklore around the idea of a supernatural companion. Based on a demo video I saw ahead of its appearance at the WSJ Future of Everything conference this week, the quadruped robot can move around your home on all fours independently, like a pet. The Familiar is a "physically embodied AI system" that will use generative AI, via an on-device model, to engage with its owner with the intent of forming an emotional connection and develop "a distinct personality," Angle told me in an interview. Robots that can react and respond to humans should, in theory, be more effective serving in what Angle calls "high human connection roles," such as companionship, entertainment, hospitality, smart home, eldercare, and parental support. "The next era of robotics is not just about dexterity or humanoid form -- it's about machines that can build and sustain human connection," says Angle. Internally codenamed Ami, the first Familiar won't be available to buy until next year at the earliest, and will cost "around the same as pet ownership," Angle said. Its exact features are also still under wraps, but Angle says initial use cases are focused on families with young children, companionship for the elderly, and addressing the global loneliness epidemic. It's a bold move for a man who built a career on robots that clean floors. Angle says his entire three-decade career in robotics has led to this moment. The original name for iRobot, founded in 1990, was Artificial Creatures Inc. But back then, the tech to create artificial life didn't exist. "Finally, I get to do what I originally set out to do. It's not just about building cool animatronics. Now is finally the time where the tech exists -- if properly and responsibly used -- to start creating Familiars." After iRobot's failed sale to Amazon, Angle stepped down as CEO in 2024. In the time since, he and Familiar Machines cofounders Angle, along with cofounders (and iRobot veterans) Ira Renfrew and Chris Jones, have assembled a team of roboticists and engineers from Disney, MIT, Boston Dynamics, Amazon, Bose, and Sonos. Their goal is to create a robot that is not just a toy or a chatbot slapped into a piece of plastic, as we saw everywhere at CES this year. From the get-go, the team dismissed the idea of a humanoid robot, believing it unnecessarily complicated for their purpose. Ami is a deliberately unidentifiable creature because, Angle says, if you create a specific animal or form-factor, people will have preconceived ideas about its abilities. They also decided not to have the Familiar talk. Instead, it will make nonverbal sounds -- the two units shown off at WSJ Future of Everything made meowing and purring noises. "By design, it will avoid giving factual advice about things that maybe it shouldn't be giving factual advice about," says Angle, with a nod toward the hot water LLM-powered chatbots keep getting themselves in. Primary communication will be through expression and body language, aided by a camera-based vision system and microphone array. With 23 degrees of freedom, the robot can move its head, neck, ears, eyes, and eyebrows, and walk at a slow human pace, but it can't grip things or climb stairs. Its four legs provide stability, which should help with concerns around the robot falling and damaging property or injuring people. Familiar Machines' goal is to use AI to create a robot that can learn from its owners, remember patterns, and adapt to their routines, with the aim of fostering long-term engagement, says Angle. They want to avoid being stuck in a closet or falling victim to the fate of dozens of earlier home robots (see Jibo, Aibo, Vector, Astro, etc.). "If this is a toy, we've failed," says Angle. "If this is a creature that you want in your world, then we've knocked it out of the park. It's kind of one way or the other." The Familiar is powered by Nvidia's Jetson Orin chip. "Its onboard edge AI stack is powered by a custom small multimodal model optimized for social reasoning, combining vision, audio, language, and memory to create socially responsive behaviors in real time," says Angle. It doesn't require an internet connection, although it can be connected, and it doesn't stream audio or video to the cloud, a purposeful design decision to protect privacy and improve latency. However, it's still a device with cameras and mics in your family space. So, why would anyone want this robot in their home? While Angle is coy about specifics, since the robot is still in development, he says an AI-powered companion could help address the growing problem of loneliness and provide an alternative to technology that keeps us glued to screens. "If your Familiar gets you up and out of your room and walking around -- that's a real way to try to address isolation and loneliness," he says. Most attempts at companion robots so far fail the "plate of glass" test, says Angle. "If a sheet of glass between you and the device wouldn't change the experience, it should just be a screen." This is why Ami is designed to interact with you, including nudging you and hugging you. It also has a "luxurious" touch-sensitive coat, says Angle. The video demo I saw of Ami showed a boy putting down his tablet to pet it, a man deciding to stop doomscrolling and go to bed after a nudge from the robot, an elderly woman walking Ami, and a younger woman doing yoga next to it. The idea is that if it encourages you to do things other than being on a screen or sitting alone at home, you're more likely to engage with other people. While more non-screen-based interactive technology could be an antidote to our screen-obsessed society, it's an extremely tenuous link to more human interaction. It's going to take a lot more than a robotic pet to unglue my teenager from TikTok. And it is still a robot, not a human you are interacting with. By offering a substitute for some aspects of human companionship, it could just as easily become as big a barrier to real social interaction as screens are. Assuming Angle and his team can pull off the robotic and personality parts, success will still depend heavily on how consumers respond to bringing a robot into their homes. And how much it costs. The first Roomba was a hit because it cost under $200 and vacuumed the floor. "Pet ownership costs" is a dizzyingly broad range. While Angle claims interest in the Familiar is "higher than what we saw with Roomba," the lack of a clear purpose feels like a problem. The strongest use cases Angle hinted at are as a parent support tool -- a device that interacts with your children when you can't and is better than a tablet or TV screen -- and as elder support, helping alleviate loneliness and managing routines for medication and movement. The latter is similar to one of the few successful companion robots so far, Intuition Robotics' ElliQ. (Angle is on the company's board). Many people will look at this Familiar concept and say, "I'd rather have a real dog or a cat." As a devoted pet owner, I know I would prefer to give my cat snuggles, walk my dog, or hang out with my chickens than spend time with a machine. Angle points out that there are many reasons people can't have pets, sharing a statistic that claims pet ownership declines to just 9 percent after age 68, when it can become harder for someone to care for an animal. For those who want an animal companion but for whatever reason can't have one, the Familiar could be an interesting alternative. Angle demoed two Familiars on stage at the WSJ conference, showing them moving around, walking, and interacting with people, and making meowing and purring noises. The units were partially operator-controlled and partially autonomous. Angle didn't say to what extent, but he says the Familiar will be fully autonomous by the time it launches next year. "This is a demo, but it's a demo on its way to a product; we're already in factories," he told the audience. But even then, it's not clear how close it will be to Angle's vision. These are very big promises in a field filled with failures and sky-high expectations, and all we've seen so far are tightly controlled demos. What is clear is that Angle believes his team has already made significant strides towards their vision of creating artificial life. "It's not a toy. It's a real robot," he says. "There is enough to it that it's beautiful, wonderful to pet and give a hug to, and it can keep up with you. It's the agency. For some definition of alive, it's alive."

'This is the robot I wanted to build forever': former iRobot chief on his extraordinary 'Familiar' AI companion

Twice now, over the last 24 years, Colin Angle has surprised me. I'm not talking bemusement or nodding appreciation. Think jaw-dropping moments when I felt the earth shift a bit under my feet. The first time was in 2002 when he unveiled the iRobot Roomba, a rather basic-looking but shockingly effective robot vacuum, and the second was today, when he showed me a video of his new companion robot, the dog-sized Familiar. Angle and I have spent decades talking about the robot revolution, or rather, having frank discussions about the reality of that not-quite-yet-here robot uprising. He's the one who cautioned me years ago that functional home humanoid robots wouldn't arrive until 2050. And when he rang me up to talk about his new product under the banner of "Familiar Machines and Magic", I wondered if perhaps he, too, had drunk the Kool-Aid and now believed that humanoid robots like Tesla Optimus, Neo Beta 1, and Figure 03 were all ready to join me in my kitchen or living room. I needn't have worried. "Guess what I'm not building? Humanoids. I'm sure you're shocked by this revelation," Angle laughed. "I think humanoids make a lot of sense in industrial settings -- I joined the board of directors of Boston Dynamics, ...I believe the humanoids have a place. But [it's] going to be a long time before that place is the home," said Angle. Angle had yet to reveal the product, and now I thought it might be a desk-bound puck that you could talk to -- I could hardly hide my disappointment. Then Angle showed me the video. But before he played it, her offered this clarification, "The robots here are real -- none of this is CGI. Just as a disclaimer, because it's necessary. " Seeing the astonishing Familiar for the first time In the video, a medium-sized dog-like creature peers from around a wall into a living room where an elderly mother and adult daughter sit. It's black eyes blink expressively, and its pointed ears wiggle. Then it pads over on large, almost bear-like paws. The neutral-colored fur moves naturally, and its motion is fluid and lifelike. One woman in the video pets the robot. In another scene, the robot reaches its paw up and pats the owner in an effort to get his attention. I'm both excited and concerned. I've seen robots like this before, though perhaps never at this scale. Where the Sony Aibo is the size of a large chihuahua, this bot, which Angle says is called a "Familiar," is the size of a collie or smaller golden retriever. The video ends, and Angle excitedly starts rattling off specs. "[it has] 23° of freedom. Got an [Nvidia] Jensen Orin processor on it. It has vision, it has array microphones. It is running full-stack AI with reinforcement learning at the bottom layer. It has memory formation capabilities." Included in that are a variety of AI models, which will give the Familiar the ability to take its base personality and learn and grow with you. When I ask Angle which models, he tells me it's a mix of homegrown and third-party, but also acknowledges that it's far from final. "Every month, there's a best model, and it has changed month-to-month, so I actually can't answer that question yet," he added. All the benefits of rapid change Things, though, are moving fast, and like much of the rest of the robotics industry, Angle's company is benefiting from "AI Time." "What you can do with a small software team and AI programming tools. It is remarkable...what we're doing was, in fact, impossible 6 months ago," said Angle. Angle's Familair robot doesn't cook, clean, fold laundry, or talk. The goal isn't removing human home maintenance drudgery. It's connection. "Our goal is to build physical AI solutions that involve human connection and the challenges that we're working to solve, aligning the physical embodiment with the expectations that are needed to create value," explained Angle. The focus on value goes back to Angle's days as co-founder and CEO of iRobot. Basically, any robot must provide more long-term value than it costs. This is how it avoids ending up in the closet. Angle worries that humanoid robots, which are currently extremely expensive, "set expectations insanely high. There's an expectation of dexterity that's necessary. There's an expectation of understanding and comprehension that are well beyond the capabilities of AI and robotics today." The Roomba robot vacuum, which initially sold for just $199, was the perfect blend of value and utility, and it met expectations. Of the millions of units sold, few were relegated to the closet. They were never pretty or expressive, but they sure could clean a rug. "One of the things that made Roomba successful was it found a way to get out of the closet and be used on a routine basis, and thus it had enduring value and was respected as such," said Angle. When Angle thought about how to bring value to the robot/AI/human equation, he and his team of former iRobot employees and advisors with backgrounds from places like MIT and Boston Dynamics, thought of pets. He reminded me that people pet their animals for more than an hour each day. "So pets had, in fact, succeeded in finding a long-term role. Now, of course, they're not robots, but, you know, it's an interesting prove point." It's not quite a dog The Familair, as we'll call it, is far closer to a pet than a humanoid robot. Its triangular face evokes a cat, but its substantial body is of another species. Angle told me a lot of thought went into that design, "Obviously, we didn't want to be evocative of humans. We also didn't want to be evocative of a dog or a cat, and so ...actually, abstract bear is what we were going for." Of course, a robot of this size can be heavy, loud, expensive, and such a massive drain on battery life that it's scarcely useful as a toy or companion. Angle holds up what looks like a piece of plastic, telling me it was one of the first things his company created and is, in a way, the foundation of the Familiar Machines & Magic AI companion. "This is a 3D printed object...is an actuator with an encoder and a gearbox on it that was manufacturable at prices that would allow this Familiar to exist. The entire scale of the robot is based on this." It's a critical and apparently affordable piece of quickly manufacturable hardware to build and scale the bot, and let it move and act in ways that elicit a reaction from anyone who sees it. "It's kind of like taking something beyond the animatronics you'd see at Disney, making it not man in the loop, but actually autonomous and then selling at consumer price points," added Angle. Of course, it's not just the movement, but the entire look of the thing. The white and tan fur (an early choice, and there will be quite a few coat color choices in the future) looks soft, malleable, and takes that collection of 3D-printed motors and makes a symphony of artificial life. "You wonder why all these robots are hard. It's because putting a fuzzy skin on them is an incredible challenge. This is a 3D knit coat, where we're able to digitally specify the shape. How much plush there is. Where is it nice to pet? Where do we need airflow to come in, to make sure it stays cool?" said Angle. Part of the solution for mass-producing this kind of artificial skin and fur came from, of all places, the shoe industry, which showed them how to "extend into really weird directions for creating mass manufacturable, materials of arbitrary shape and physical characteristics." Don't expect a chatbot Unlike many AI companions and assistants, the Familiar never speaks, only making vaguely animal sounds, an unsurprising choice when you consider how you'd feel if your Doberman started chatting you up. Still, Angle told me this robot will be, like any good pet, emotionally aware. "Within the first few minutes of experiencing it, you should be able to understand that it is expressing understandable behavior. If it's trying to get your attention, if it's trying to get picked up, if it's trying to get you to take it for a walk," said Angle. And, yes, this robot, which should be lighter than a comparably-sized dog, can go for a walk inside or out, though keep it away from the rain. Battery life is unspecified beyond: it can do a walk and will go and charge itself when necessary. "It is always awake and alive even when it's recharging and has a substantial duty cycle during the day," explained Angle. Despite that always-on (and possibly a little creepy) nature, stereo cameras, and microphones, the Familair will not be a home-surveillance robot, at least not for now. "Certainly, that's not something that is viewed as a launch feature. This is not marketed as a security solution," said Angle. A robot companion of this magnitude, one that can connect with you at a deeper level, remember your interactions, identify family members, and grow and change with you, is a tall order. Would it cost thousands? Angle refused to be pinned down on price. Again, though, he returned to the concept of a pet. "This is not something that is designed only for the highest bracket income. If you can afford a pet, you can afford this," Angle promised. A Familiar coming-out party This week, Angle and his Familiar will make their public debut at the Wall Street Journal Future of Everything Conference in New York City on May 4. Angle told me it's likely that what conference attendees see and what I saw in the video a few weeks ago might be quite different. The team plans to keep working on the robot right up until the conference. After that, expectations will be set, and while Angle has known massive success with iRobot (before Amazon tried to buy them, failed, and almost destroyed the company -- Angle left), there are no guarantees here. When I ask Angle why the Familair might succeed where others have failed, he responded with some enthusiasm: "The fact that this isn't a toy. The fact that it is a supportive, aware presence in your home, which looks valuable, is expressive, and knows enough what's going on so that if you come home from work stressed out, it actually can come over and try to cheer you up." Angle has literally been dreaming of making the Familiar for longer than I've known him. "This is the robot I wanted to build forever," said Angle, adding at another point, "This thing has been kind of in my mind, under construction, for 30 years." Perhaps Angle's willingness to take the largest swings in a market littered with the corpses of failed AI and robot companions is rooted in how he sees himself. Angle told me his favorite character, a hero of his, is Dr. John Hammond from Jurassic Park. Right, the guy who brought back the dinosaurs and created the world's most dangerous theme park. Angle, though, sees a different lesson: Hammond just wanted to make his dream, his fantasy, real. "You know, he had lived his whole career with smoke and mirrors, and Jurassic Park was trying to give people real, and I think to, maybe, without the 'Don't build dinosaurs' -- that was his big misstep -- but I think the chance to build physical AI that is actually capable of satisfying and enduring human connection is now possible." Not only can Angle now make his how dreams real, but he thinks he can do it better than those who have come before it, who have relied too heavily on handing the reins over to AI and letting it operate in some "very challenging and questionable arenas," like privacy and security. "We can avoid all of this, and make something wonderful that allows the world to be a little bit more caring...it's just as concrete, needed, and valuable as Roomba ever was." Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button! And of course, you can also follow TechRadar on YouTube and TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.