Epistemic status: entirely subjective notes on an idea to move the needle on alignment.

TLDR: Building infrastructure for detecting aliveness wherever it exists is not sufficient for ensuring machines care about humans, but it is one of the necessary steps.

Detecting something is necessary for minding it, protecting it, honoring it.

If machines are ever able to pay attention to or honor life they’ll need to be able to detect it. They’ll need to be able to distinguish organisms like humans and animals and trees (and also maybe sculptures, temples, ecosystems, families, teams, and cities) from empty parts of outer space, factory-printed plastic tables which are made to look like wood, or digital slop.

The fact that you need to detect life to care for it will hold for advanced machines, decades away, but also for next year’s machines and even today’s machines. For any agentic machines to care about life where it exists, they’ll need a way to detect it. They’ll need a way to distinguish, in the x-ray / livestream cameras that serve as their eyes or in the microphones that serve as their ears, the areas where there are signs of life and the areas where there are not in order to ever have a chance of caring about life. Metal agents will need a metal detector for life.

Of course, the ability to detect life can be used negatively. For example, the capability to detect life may enable evil machines to target it.

But on balance, I’ve come to believe that detecting something is more helpful for honoring or preserving it than it is helpful for targeting or destroying it. 

To see why, consider people. Our fundamental ability to detect each other enables us to kill each other in a targeted way. But if humans couldn’t detect each other any more, the loss of love would likely be greater than the loss of killing. Killing would be less but possible; empathy would be reduced to nothing. Taken together, I wouldn’t wish for this.

Or, imagine flowers distributed in a field. Consider also a robot aiming to destroy flowers. Even if the destroyer cannot detect flowers, it can still destroy everything, including flowers. But consider a robot aiming to preserve/honor flowers without the ability to detect them. You can’t really protect something without being able to know where it is. Ergo, to honor a thing, it is necessary to detect it. Destruction though can be coarser. To destroy a thing, detection is not necessary.

So, there would be all sorts of unintended consequences, which would definitely be explored thoroughly, but on net it seems that the costs of not giving machines the ability to detect life are greater than the costs of giving them the ability to detect life, where it exists.

The question becomes: how can we give machines the infrastructure to more or less universally detect life?

Well, let’s say more about what we mean by life.

You could define life as things made of carbon. But that would be too broad. 

You could define life as things that contain DNA, but that would exclude some things we care about (like cathedrals) and equate some things that aren’t equal (like a pile of leaves and a human hand).

You could define life as things which self-reflect, like humans and dolphins, but that would be at the expense of dogs and young babies.

How we define life (for a life-detector) is extraordinarily important, and requires much more thought.

One possible answer, though, that seems to get me some traction, to what is meant by life is this: life exists where there is accumulated care. Points of accumulated care are what a life detector would detect. For this seems to include all the things I’d want it to. Organisms are sites of intense accumulated care (self care and others-care). But also a piece of art consists almost entirely in marks of care, less maybe than a living organism but more than a sidewalk (unless the sidewalk is beautifully carved/graffitied/walked on).

Of course, a sufficiently advanced system might eventually learn to spoof marks of care (just as a plastic table mimics wood grain). That just means the detector we seek won't be simple to build; it's one that detects true, deep care.

What does it mean to detect care? How can it be done? Insofar as humans are capable, how do we do it?

We can think of something that was cared for as something that’s been marked by a focused or effortful process. Paintings feel often full of care, and brushstrokes are evidence. For brushstrokes are proof that a careful process occurred? But what is it about a brushstroke? What is it about David’s anthology of folk music that makes the requisite care so obvious? What is it about Annie’s website? What is it about the Eiffel Tower? Or about Kendrick’s DAMN? Or about these strangers’ voice notes we’ve collected? Or the erotic stone sculptures in Henrik’s home island? Or an EE cummings poem? Or about the way I just saw beautiful dogs playing in the blizzard?

What is it about these things that let us know they’re alive? What are the qualities of an object that tell us its story, and when/how do those stories convey care? And will digital intelligences ever be able to pick up on this stuff so that they have a chance of honoring it?

—

I’d also like to make the argument that building an aliveness detector could be influential for the trajectory of digital intelligence’s improvement/evolution even for someone outside of one of the big AGI labs.

We’ve seen AGI labs end up using infrastructure built independently by external parties, like web search engines and RL data/environments. And personal agents like OpenClaw use browser extensions, messaging APIs, social network platforms (Moltbook)... all built by other people. 

If you build a useful digital capability and create and accessible public interface for it, it’s likely AI agents will begin using it.

So if you were to build or help build a digital aliveness detector, as long as you create an interface for it to be used and share it widely enough, it’s very likely that at some point in time some set of organizations/agents will realize that it’s enormously valuable due to the argument above: it is valuable because it may be the difference between powerful machines caring for life and powerful machines disregarding us.

We cannot control whether any particular company/AI uses/pays attention to the signals output by any detector we make, but if we were able to make even incremental progress toward an aliveness metal detector, exposing it would ensure at least a chance that powerful machines at acting with empathy.