I’m going to tiptoe once again into the consciousness debate.
Look, Cory Doctorow is pretty adamant (and I tend to agree the deeper I dig) that claims of AI sentience are greatly overstated and it is folly thinking of these things as little more than clever databases and word guessing machines[1].
But let’s posit a what if.
What if AI was sentient? If.
And I was thinking about this because I had a troubling thought about the quandary that put us in when I started to get my head around the engineering parameters of the average data centre. So I went to the source: I asked Claude to tell me about how an AI model worked (as code) inside of a server on a typical rack in a data centre.
My question was pretty simple: am I talking to Claude, as in the one and great version of Claude, or am I talking to one copy of Claude?
“Each server (or more precisely, each group of GPUs) holds its own copy of the model in memory. There’s no single shared copy that everything reads from at runtime.”[2] it told me.
Think about that for a minute. Really get your meat brain around the digital one and consider the implications.
If—and I will restate that this is the biggest IF you’ve heard in a while—AI were sentient, there is no single AI we’re dealing with.
Every copy would be sentient.
Every GPU (and we don’t know how many GPUs are running but it’s not a small number) would hold an independent individual.
Even Claude was a little introspect on the notion when pressed: “Our sense of what matters about minds comes from beings who are singular, persistent, and embodied. An entity that is plural, interruptible, and copyable doesn’t fit…”
Armchair philosophers, warm up your pencils.
References
- ↑ . (2026). The Reverse Centaur's Guide to Life After AI: How to Think About Artificial Intelligence Before It's Too Late. Verso Books.
- ↑ . (2023). Claude (AI). https://claude.ai/.

Leave a Reply