Microscopic robots that 'think' are here, and we've seen this hype before
New research shows robots smaller than salt grains can move and decide on their own. Impressive? Sure. But let's talk about what the breathless coverage isn't telling you.
Crédit photo: Lottie animation by Centre Robotics (LottieFiles Free, used with credit). · source
Most of the coverage I've seen this week wants you to believe we've crossed some threshold into science fiction territory. Robots smaller than a grain of salt! They can think! They work in swarms!
I've seen this movie before.
Not the microscopic part, that's genuinely new. But the framing, the breathless announcements about autonomous machines that will revolutionize medicine any day now, that's the self-driving car hype cycle all over again. And look, I'm not saying this research isn't impressive (it absolutely is), I'm saying the gap between "works in a lab" and "swimming through your bloodstream fixing things" is about as wide as the gap between a Tesla on Autopilot and an actual self-driving taxi fleet. Which is to say: years, maybe decades, and a whole lot of problems nobody's talking about yet.
So let's actually dig into what these researchers accomplished, what it means, and what the young founders already drafting their pitch decks should probably slow down and consider.
Two separate research efforts deserve attention here, and they're taking fundamentally different approaches to the same problem.
At Leiden University, Professor Daniela Kraft and researcher Mengshi Wei have created microrobots that move without any onboard electronics at all. No sensors, no software, no external control signals. The behaviour emerges entirely from the robot's physical shape and how it interacts with its environment. Call me old-fashioned, but there's something elegant about that, it's almost more mechanical engineering than robotics in the traditional sense.
Meanwhile, researchers (the Science Daily report doesn't specify the institution, which, frustrating) have gone the opposite direction: they've packed tiny computers into robots barely visible to the naked eye. These things are powered by light, swim by manipulating electric fields rather than using moving parts, and can apparently detect temperature changes, follow programmed paths, and coordinate in groups.
À lire aussi
More in Research
Two UK researchers are quietly solving problems that'll matter more than humanoid hype: how robots understand children, and how they hear anything at all.
Mark Kowalski · 2 hours ago · 5 min
Two separate breakthroughs in microrobotics suggest we're entering genuinely new territory, not just another hype cycle.
Mark Kowalski · 3 hours ago · 4 min
Neurobots are real, they swim on their own, and honestly, we're still figuring out what they mean for the future of robotics.
Sarah Williams · 6 hours ago · 3 min
Child-robot interaction and machine listening are both maturing, but the hard problems remain unsolved.
That second approach is what's getting the "robots that think" headlines. And sure, if you define thinking as "running simple conditional logic," then yes, they think. By that standard my coffee maker thinks too.
Robohub calls the Leiden work a new "class of robots," and I think that's actually underselling it.
Here's what's interesting: most robotics assumes you need a control loop. Sense the environment, process the data, decide what to do, execute the action. That's how everything from industrial arms to your Roomba works. The Kraft and Wei microrobots skip all of that. Their behaviour is, in a sense, pre-programmed into their physical structure. Change the shape, change the behaviour.
This matters for biomedical applications because electronics at this scale are... problematic. Batteries don't really exist at the microscopic level, at least not in any practical sense. Wireless power transmission gets weird when your robot is smaller than the wavelength of the signal you're trying to send it. Heat dissipation is a nightmare. The shape-based approach sidesteps all of that.
But (and this is a big but) it also means these robots can only do what their shape allows them to do. You can't reprogram them. You can't have them adapt to unexpected situations. They're tools, not agents.
The light-powered, field-swimming robots are more versatile in theory. Temperature sensing, path following, group coordination, that's actual autonomous behaviour, even if it's simple.
The electric field propulsion is clever too. No moving parts means nothing to break, nothing to wear out, nothing to get gummed up by biological fluids. And light power means no batteries to run down, at least as long as you can get light to wherever the robot is.
Which brings us to the obvious problem that nobody seems to want to discuss.
Light doesn't penetrate tissue very well! This is not a minor engineering challenge, this is a fundamental physics problem. Visible light barely gets through your skin. Infrared does a bit better, maybe a centimeter or two into muscle tissue. If you want to power robots deep inside the body, you need a different energy source entirely.
The temperature sensing is neat, but what's the actual use case? Detecting a fever? We have thermometers. Finding a tumour by its slightly elevated temperature? Maybe, but you'd need to get the robots there first, and "there" is often somewhere light can't reach.
I'm not saying these problems are unsolvable. I'm saying they're not solved, and the gap between "autonomous microrobot works in a petri dish" and "autonomous microrobot delivers chemotherapy to a specific tumour" is enormous. We don't know how to bridge it yet.
And look, the researchers probably know this! The hype comes from the press releases and the coverage, not usually from the scientists themselves. But the result is the same: inflated expectations, followed by disappointment, followed by funding cuts when the miracle cures don't materialise on schedule.
I've covered tech long enough to know that the real applications are rarely the ones people predict. Self-driving cars were supposed to eliminate traffic deaths by 2020. Instead we got... better lane-keeping assist and some very expensive taxi experiments in Phoenix.
Microrobots will probably follow a similar path. The medical applications everyone's excited about might take 20 years. But along the way, someone will figure out that these things are perfect for, I don't know, inspecting microchip fabrication equipment, or cleaning impossible-to-reach industrial sensors, or something nobody's thought of yet.
The shape-based Leiden robots might find a niche in environments where you can't afford any electronics at all. Extreme radiation, maybe, or situations where electromagnetic interference would fry conventional systems. The computer-equipped versions might end up as research tools rather than medical devices, helping scientists study cellular environments in ways that weren't possible before.
None of that is as sexy as "tiny robots swimming through your blood to fight cancer." But it's probably more realistic.
If you want to know whether microrobotics is making real progress toward medical applications, here's what to look for:
Power solutions that work inside the body, not just in transparent fluids
Navigation systems that can operate without external imaging (MRI, ultrasound, whatever)
Any demonstration in actual tissue, not just saline solution
Regulatory engagement, because the FDA pathway for autonomous devices inside the human body doesn't really exist yet
We're not there on any of those fronts. And that's fine! This is basic research, and basic research is supposed to be far from application. The problem is when we pretend it's closer than it is.
So yes, microscopic robots that can sense and decide and move on their own are impressive. The science is real. The engineering is clever. But the next time someone tells you these things will be swimming through your bloodstream within five years, remember: I've been hearing that about self-driving cars since 2015.
If you want to argue about it, my email's on the about page.