Three years ago, I started this blog with a plan. I was going to write about Pepper, about NAO, about Sophia making her conference rounds with that uncanny almost-smile. About the films that shaped how we imagine artificial minds, and the researchers quietly building them into something real. That was the blog. That was the plan.
Somewhere along the way, I started writing about chatbots.
Not because I decided to. I didn’t sit down one day and think: this is the more interesting direction. It happened the way most drift happens; gradually, then completely. My feed filled up. The conversations around me shifted. I still follow Sophia on Instagram. I can’t tell you the last time I saw one of her posts.
This is not a story about disappearing
Boston Dynamics is still continuously improving Atlas. A new version, more capable than anything that came before, that most people I know have never heard of. Hanson Robotics is still working on Sophia. The researchers didn’t go home. The robots didn’t go anywhere.
We just stopped looking.
The novelty that talked back
The first wave of public fascination with physical robots ran on science fiction fuel. Nobody owned a Sophia. Nobody had a NAO in their living room. But we watched the videos: the parkour, the football, the eerily smooth movement. And we were enchanted, because this was the thing we’d only seen in films, suddenly real enough to point a camera at. It didn’t need to be useful. It just needed to exist.
Then came something that talked back.
Large language models (LLMs) didn’t just offer a new novelty; they delivered on a promise physical robots never quite managed. Ask Sophia something outside her parameters, and she loops. Open a chat window and ask almost anything, and you get a real answer, in real time, that actually helps. The magic trick landed. And suddenly the body seemed optional.
What followed wasn’t a technological defeat for robotics. It was a matter of noise. LLMs are accessible in a way physical robots have never been. All you need is a phone and an email address. They’re immediate. They’re useful for an absurd range of things. They develop fast enough that there’s always something new to react to. The robots couldn’t compete with that rhythm, so they dropped out of the cultural feed. Not gone. Just quiet.
The merger nobody is talking about
The irony is what happens next. In several labs right now, the two things are being merged. Boston Dynamics and OpenAI. Figure robotics. Apptronik. The embodied AI that physical robots always gestured toward but couldn’t quite achieve is being built by putting an LLM brain inside a body that can move through physical space, read a room, hand you something. It is not a distant hypothesis. It is already clunky and expensive and real.
Are we ready for the body?
When it arrives properly, and it will, I don’t think we’ll be ready. Not because the technology will be too strange, but because the strangeness will be the wrong shape. A year of talking to something that talks back well does a lot of psychological preparation work. The voice, the reasoning, the uncanny sense of being understood. We’ve been quietly getting used to all of it. But a body changes the register entirely. Something that moves through your space, that exists in the room with you, that triggers different instincts, older ones.
Then again, maybe not. Maybe the people who’ve spent years talking to AI every day have already done more of that work than they realise. Maybe the body will be less of a leap than it would have been in 2015, before any of us knew what it felt like to be genuinely heard by something that isn’t human.
I started this blog to write about robots. I ended up writing about chatbots, because that’s all I could see. Because that’s all any of us could see. The robots are still there, getting quieter and more capable in the background, waiting for the moment the novelty cycle turns again.
It always does.
If you want to read more about the uncanny vally of it all, may I suggest The line we needed to draw: sci-fi, robots, and the boundary we cannot afford to lose or The other we invented: the valley without a shape.