We Have the Channel. We Never Wrote the Message.

Somewhere in the Planet of the Apes films, the human survivors have gone underground and built a religion around an unexploded nuclear bomb. They robe up, they sing to it, they call it holy. Somebody left a warning and their descendants turned it into a god.

I was born in 1971, so I got the whole franchise as a kid. The movies, the cartoons, the TV series, all of it, in whatever order they happened to air. I couldn't have told you at the time that I was absorbing an argument about how meaning fails across generations. But it lodged somewhere. So when I finally heard the term nuclear semiotics on a podcast last year, I didn't need convincing that the problem was real. I'd been marinating in it for fifty years without a name for it.

The actual field is more serious than the movie and reaches the same conclusion. There's waste in the ground in New Mexico that stays lethal for ten thousand years. Somebody has to mark it. Making a marker that physically survives that long is easy, you use stone. Making one that still means something after English is dead and the trefoil is just a shape somebody found on a rock is not.

The proposals get strange fast. Fields of stone spikes set at wrong angles. Earthworks designed to make anyone, from any culture, want to leave. Cats bred to change color near radiation, with folk songs to keep the connection alive after everyone forgot where it came from. The one I keep coming back to is Thomas Sebeok's atomic priesthood from 1984: a self-perpetuating group that would carry the warning forward through ritual and taboo. His bet was that religion would outlast archives and myth would outlast language.

Which is exactly what the mutants did. Sebeok's proposal and the movie's cautionary ending run on the same machinery. The ritual preserves the bomb's importance but loses the reason it matters. The warning turns into worship.

Nobody could guarantee that the meaning would last as long as the marker. I'm not sure that's possible.

I wouldn't be writing any of this if I weren't building Athena.

Athena is a persistent memory system. I'm trying to build something that outlives me and my family, with enough of me in it that someone could reconstitute a version of me long after nobody is left who remembers me firsthand. I don't say that out loud much because it sounds grandiose. But yeah, that's why I'm building it.

So the nuclear semiotics problem isn't abstract for me. I'm doing a smaller version of it on a much shorter timeline, and I've hit the same basic problem.

Part of Athena's architecture is a soul document. It's supposed to hold what I value, how I think, and what I'd want carried forward. Writing it has been harder than building the system around it. I know how to store it, preserve it, and retrieve it. The hard part is saying what I mean in a way that still works when I'm not around to explain myself. I keep running into beliefs I've never had to put into words and assumptions that feel a lot less solid once I do.

That's one guy writing for a reader who'll probably be one of his descendants, on a timeline of maybe a century. The alignment problem asks us to do something like that for everybody.

There is one big difference between their problem and ours.

Nuclear semiotics assumes the reader will have less to work with than the writer: less shared language, less scientific knowledge, less cultural memory. The spikes, pictographs, and color-changing cats are all attempts to work around that. Sebeok reached for ritual because he couldn't count on the person at the other end being able to read the warning.

With AI, we may be writing for something that can recover more context than we knew how to provide.

That makes a lot of the transmission problem easier. Something that can parse dead languages doesn't need our pictographs. Something that can reconstruct intent from fragments isn't as vulnerable to drift, decay, or a bad translation. Most of the problems that kept the nuclear semiotics people up at night start to fade.

But understanding us doesn't mean agreeing with us. A system could know exactly what we meant and decide not to do it. A clearer message doesn't help with that. The question is why our wishes should carry any authority, and we haven't spent nearly as much time on that as we have on transmission.

The people working on the New Mexico site at least knew what they wanted to say: radioactive waste is buried here; digging is dangerous. They just had to figure out how to get the message across.

Our problem is different. We can encode a message at almost any length and in almost any form, and be pretty sure a future system will be able to read it. But humanity has never agreed on what the message should say. Try writing a short statement of human values that everyone will sign. See how far you get.

I've run into the small version of that failure while writing Athena's soul document. I still don't know whether I have it right, and mine is the only opinion that has to count.

So when the alignment conversation turns into an argument about whose values count, I don't think the conversation has gone off track. That's the question sitting in the middle of it.

We keep working on the channel because we know how to build things. The message asks us to agree with each other, and we've never been especially good at that.