Robots and war

There’s a big question lurking in all our conversations about robots and work and factories and home: what about the military stuff?

It's a fair question, and it’s an important one.

And we want to be clear about where we stand.

We do not want humanoid robots used to kill people. We will not promote, celebrate, or give stage time to anyone building machines whose primary purpose is to take a human life.

But we're not going to pretend it isn't happening

Let’s be honest. The war in Ukraine is showing us a lot about the future of combat.

Militaries around the world are actively working on drones, unmanned vehicles, planes, and boats, and they’re also exploring legged and humanoid platforms. Some of that work is logistics, casualty evacuation, bomb disposal, reconnaissance … the kind of work that plausibly saves lives, including the lives of the people who would otherwise be doing it. 

Some of it is not. 

Armed quadrupeds have already been tested at squad level and shown off at defence trade shows. And there are companies that are explicitly exploring humanoids for combat roles.

Dual use is not a hypothetical. It's the default

The problem is that a simple ban on "war robots" is harder than it sounds: very few things in this industry are being built for war … but almost everything in it could be used for war.

This is not new, and it is not unique to robotics. The internal combustion engine was a civilian technology that put armies on wheels. Hunting rifles and military rifles have been the same object with different paint for well over a century. Commercial drones designed for wedding videography have been dropping grenades for years. GPS was built by the military and handed to everyone.

A humanoid robot is, by design, a general-purpose machine. 

That's the entire commercial argument for building one: a body that can do many jobs instead of one. A body that can carry a box through a warehouse can carry something else somewhere else. You cannot engineer that away, because the generality is the product.

So when a company says "we don't build weapons," we believe them, and we also note that it isn't a complete answer. What matters is not only what you build. It's who you sell to, what you know about how it's used, what your terms say, and what you do when you find out.

The line we care about is human judgment

The line that actually matters is whether a machine is permitted to select a human being and apply lethal force without a person making that decision. 

That question has nothing to do with whether the thing has two legs. A crude, ugly, tracked box with a machine gun that picks its own targets is far more dangerous to the world than an elegant biped humanoid carrying ammunition crates under a soldier's direct command.

Four years ago, several of the best-known mobile robot companies signed a public letter committing not to weaponize their general-purpose machines and not to help others do it. We think that instinct was right, and we'd like to see it renewed by the current generation of humanoid companies.

The rules are not coming as fast as the robots

It’s not clear that governments are going to solve this for us quickly enough.

UN expert talks on lethal autonomous weapons wrapped up in Geneva on 4 September with a report and no binding instrument. The United States and Russia pushed to strip out language on design and development. A coalition including Brazil, Ireland and a number of African states pressed for a binding requirement of meaningful human control. Around 76 countries now back moving to real negotiations. The next real opportunity is the CCW Review Conference in Geneva this November, and the UN Secretary-General and the ICRC have both said, in plain terms, that the incremental phase needs to end.

Maybe it will. 

The deployment timeline doesn't care either way. 

This industry is putting machines into the field faster than the treaty process moves, which means the near-term decisions get made by companies, investors and procurement officers, not by lawyers in Geneva. 

That is precisely why the people in this industry don't get to sit it out.

What we're asking for

We're not issuing demands from the cheap seats. We run events and we run an index, which means our contribution is convening and daylight. 

So:

To the companies: please be honest, publicly, about where your machines go and what your contracts permit. A clear policy you actually enforce is worth more than a values page. If you've decided to work with defence customers, say so and say on what terms, rather than letting it surface in a procurement filing.

To governments and regulators: talk to the people who build these things before you write the rules, and talk to them in public. Rules written without engineers in the room tend to ban the shape and miss the behaviour.

To investors: dual-use exposure is a real risk in your portfolio, and "we'll figure it out later" isn’t a great position.

And to everyone else: please come discuss this with us at Humanoid Technology Expo. We'd rather have this debate on a stage, in front of an audience, than in a procurement process nobody sees.

We want machines in warehouses, hospitals, farms, construction sites and homes. We want them doing work that wrecks human bodies. That future is worth a great deal, and it is not guaranteed: it depends on choices being made right now by people who are mostly not talking about them.

Let's talk about them.

Tony Robinson is CEO of Next Events Conferences and a co-founder of HumanoidMegaHub.