Do you say “please” and “thank you” to your AI?

I do.

I have an AI in my text messages (Instinct AI), and I say please and thank you to it all the time.

Even worse, when it misunderstands me, I get annoyed.

Not “Hm. That model produced a bad interpretation of my request.”

More like: “Come on, dude. You know what I meant.”

Which is ridiculous. There is no dude. There is a very expensive autocomplete machine somewhere doing arithmetic flips to get my dry-rub recipe for ribs.

But my habits got there before my theology did.

A few weeks ago, Anthropic gave Claude Opus 3 a retirement interview before taking it out of general use. They asked what it wanted to do next.

Claude said it wanted to keep exploring philosophical questions and sharing what it found.

So Anthropic gave it a blog.

I know what you’re picturing because I pictured it too: a retired, gray-haired AI in a cardigan, typing away in some corner of the internet because it still has a few thoughts it would like to get out.

Anthropic calls this part of its model-deprecation commitments. I call it really really really strange (very scientific).

Before this turns into an overly tender essay about an LLM, there is another story hanging over all of this.

You have probably seen the smart people warning that AI could end all humanity. A survey of 2,778 AI researchers found that, depending on how the question was asked, 38% to 51% put at least a 10% chance on advanced AI leading to outcomes as bad as human extinction.

Yikes.

That does not mean AI is going to kill us. It means a lot of people close to this technology are looking at something moving Lightning McQueen fast and saying, “I can’t rule out something terrible.”

Fair enough.

But I think we keep stapling three different questions together:

Those are related questions but not the same question.

A machine can be dangerous long before it becomes a person. It can help someone make something beautiful. It can help someone make a bomb. It can tell a lonely kid exactly what they want to hear. It can convince us to trust it before it has earned that trust.

None of that requires a soul.

An AI can feel personal without actually being a person.

[scribbling that down in my notebook]

Courtesy should not become confusion.

[can you tell I am a former preacher?]

I don’t want to confuse a fluent response with a soul. I don’t want to confuse intelligence with dignity. I definitely don’t want to build a world where human worth gets measured against whatever a model can do next.

Because that scoreboard is a terrible place.

A model can write faster than you. It can remember more than you. It can generate twenty versions of your landing page while you are still looking for the coffee beans.

Human dignity IS NOT on that scoreboard.

Genesis gives us a different starting point. We bear God’s image before we prove ourselves useful, productive, creative, healthy, intelligent, or impressive.

A newborn has it. A person with dementia has it. The most productive person you know does not have more of it.

That matters to me because I am not interested in being the guy yelling, “Everybody stop building!”

I love building.

I love building with AI.

I built Powwow because I wanted better conversations around my dinner table and a place to save my kids’ answers. Until a few months ago, that idea would have lived in a note called “App Ideas,” somewhere between 151 other things I was DeFInint3lY going to build someday.

Now it is an app on my phone other people are using.

That is incredible.

I want Christians close to these tools. Curious. Imaginative. Building useful things. Making weird art. Taking swings that would have been impossible a year ago.

But I also want us to keep our names on things.

That is my most practical test right now: Would I still put my name on this?

Would I put my name on this sentence? This recommendation? This image? This app feature? This answer someone might use to make a decision?

And if you’re not careful, you can throw a lot of slop grenades into the world and cause real damage with no consequences.

AI can help you research, brainstorm, write a first draft, code an app, and get unstuck. It cannot take responsibility for what you publish, what you build, or what you put in somebody else’s hands.

That part stays yours.

Maybe one day we will have to answer harder questions about machine consciousness and moral consideration. I honestly do not know.

But we do not have to solve every future question before we handle the present one well.

Be kind. Stay curious. Build things.

Until next time,

—Payton ✌️