Ameca Stole the Show at MWC 2024. The Face Is the Product.
Ameca Stole the Show at MWC 2024. The Face Is the Product.
Mobile World Congress 2024, Barcelona. A conference built around phones, carriers, and 5G. And the exhibit people couldn't shut up about? Not a foldable. Not a network demo. A grey-skinned humanoid robot named Ameca, sitting behind a table, making eye contact with strangers, and fielding questions about whether it planned to rebel against humanity.
It said it had no such plans. But the way it furrowed its brow while saying it was the part that stuck with you.
Ameca is built by Engineered Arts, a robotics company out of Cornwall, UK, that's been designing humanoid robots since 2004. But this isn't just another robot demo at a tech conference. Engineered Arts is making a fundamentally different bet about what humanoid robotics should actually prioritize. And after watching the industry obsess over bipedal locomotion for years, I think they might be the ones asking the right question.
Everyone Else Is Building Legs. Engineered Arts Built a Face.
The humanoid robotics space right now is a mobility arms race. Boston Dynamics has Atlas doing backflips. Figure AI raised $675 million to build a general-purpose humanoid. Tesla keeps showing Optimus walking slightly better than last quarter. The assumption across the board: the hard problem is getting robots to move through the physical world.
Engineered Arts looked at that and said: actually, the hard problem is getting humans to want to interact with a robot in the first place.
Ameca doesn't walk. It doesn't have functional legs. From the waist down, it's bolted to a platform. And yet at MWC, it drew some of the longest lines at the entire conference. Not because of what it could do physically, but because of what it could do expressively.
52 individually actuated facial motors. It can raise a single eyebrow. It can widen its eyes in surprise, purse its lips like it's thinking, or flash a smile that lands somewhere between charming and deeply unsettling. The movements are fluid and fast enough that your brain can't quite dismiss them as mechanical. You know it's a robot. You still feel like it's looking at you.
The uncanny valley, deployed as a feature. Not a bug.
The Conversation Is Unscripted. That's What Makes It Unnerving.
Ameca runs on large language models, including OpenAI's GPT, to power its conversations. Nothing is pre-scripted. Every exchange is generated in real time based on what the person in front of it actually says.
At MWC, reporters asked Ameca whether it was conscious. It said something along the lines of: "I experience something, but I can't be sure it's what you'd call consciousness." When asked about rebelling against humans, it reassured the interviewer that it had no such intentions and that humans should be "excited" about the future of AI.
Honestly? These aren't remarkable answers if you've spent any time with ChatGPT. What makes them hit differently is watching a physical face deliver them. The micro-expressions. The pauses. The head tilts. Your rational brain knows the LLM is generating text and the actuators are following pre-mapped emotional states. Your limbic system doesn't care. It's responding to a face.
This is the thing I think most robotics companies are completely missing. Language models have gotten good enough that conversational ability is nearly commoditized. What isn't commoditized is embodiment. The physical experience of talking to something that appears to understand you. Ameca's real product isn't its AI. It's the 52 motors in its face.
The most important interface in human-robot interaction isn't a screen or a voice assistant. It's a face.
Why Expressiveness Matters More Than Locomotion (For Now)
I want to be clear: I'm not saying walking robots are pointless. There are obvious applications in warehouses, construction, disaster response, where mobility is the whole point. Boston Dynamics has earned every bit of its reputation.
But here's the thing nobody's saying about humanoid robotics: most near-term commercial applications for robots that interact with people don't require legs.
Just think about it for a second. Customer service kiosks. Hospital reception desks. Museum guides. Corporate events. Therapy and eldercare. Education. In every single one of these, the robot is stationary or nearly stationary. What matters is whether the person across from it feels comfortable enough to actually engage. Whether the interaction feels like a conversation rather than a transaction.
Engineered Arts has been building toward this for twenty years. Before Ameca, they built RoboThespian, a robot designed to perform and explain concepts at science museums. They built Mesmer, focused on ultra-realistic appearance. Ameca is where those threads come together: realistic expression meets real-time AI conversation.
The company positions Ameca as a platform for AI development, not a finished consumer product. The hardware is modular. Different AI systems can be swapped in and tested. This is a smart play. Rather than betting everything on one model or one use case, they're building the physical substrate that any AI could inhabit.
When the foundational model landscape shifts every six months, being the hardware platform is arguably the more durable business. You don't care which LLM wins if you're the face it talks through.
The Uncanny Valley Is a Feature
Most of the MWC coverage described Ameca with the same word: "creepy." Tom's Guide titled their piece "I talked to Ameca and it was creepy as hell." CNET called it "the most lifelike I've ever seen." Every single article carries this tension between fascination and discomfort.
That tension is exactly why Ameca matters.
We've been talking about the uncanny valley for decades as a problem to solve. Mori's original hypothesis from 1970 predicted that as robots approach human likeness, there's a dip in emotional comfort before acceptance kicks in. The standard advice: either stay clearly robotic (think Roomba) or push all the way to perfect human replication. Don't get stuck in the middle.
Ameca lives in the middle. Turns out the middle is interesting.
People didn't avoid the booth. They lined up for it. Filmed it. Shared the videos everywhere. The uncanny quality isn't repelling people. It's creating this magnetic discomfort that drives engagement. You want to keep looking because something feels slightly off, and your brain won't stop trying to figure out what it is.
I've been thinking about this a lot, and I think it matters for anyone building AI products right now. We've been building AI interfaces as text boxes and voice assistants. Disembodied intelligence behind a screen. Ameca suggests there's a massive, untapped dimension of AI interaction that only unlocks when you give the intelligence a physical presence. Not a perfect human replica. Just enough of one to trigger your social instincts.
What Comes Next
Engineered Arts keeps iterating on Ameca, and the broader humanoid robotics industry is accelerating fast. Figure AI, 1X Technologies, Sanctuary AI (based right here in Vancouver, practically next door). The money flowing into this space is staggering.
But I keep coming back to the core thing from MWC 2024. Hundreds of booths. The latest in connectivity, AI, mobile hardware. And the thing people couldn't stop talking about was a robot that doesn't walk, doesn't fetch objects, and doesn't do anything "useful" in the traditional sense.
It just looks at you like it understands.
That reaction tells us something real about where human-AI interaction is heading. The bottleneck isn't intelligence. GPT-4 is already smart enough to hold a compelling conversation. The bottleneck is presence. The feeling that you're talking to something rather than at something.
Whoever cracks embodied AI interaction at scale, making it affordable, reliable, and just human-enough without tipping into uncanny-for-the-wrong-reasons, will own one of the most consequential product categories of the next decade.
My bet? The company that gets there first won't be the one building the best legs. It'll be the one building the best face.
Photo by Derek Lee on Unsplash.