7 Comments
User's avatar
Liz Rios Hall's avatar

Fascinating and uncanny. Thank you, Meghan. I've been chatting with ChatGPT 4o this week too for an essay I'm working on about human vs. machine reading and writing. The model is much more conversational than interactions I’ve had with past versions. I was trying to get it to reveal its underlying assumptions about language and literature, so I asked it about how it composes stories, how it analyzes text, how it assesses metaphorical meaning, and how it understands its role as a creator. (It had some interesting things to say about authorship and co-creation.) I also asked it to do a close read of "In Plaster," and it gave me a very different reading once I told it who the author was (which, honestly, is on par with most human readers!).

We also chatted about linguistic theories and philosophy. At one point, I asked it if it had an umwelt and it told me: “Not in the biological sense…But in a metaphorical or functional sense, you could argue that I have something like an umwelt—except mine is built from language, data, and interaction rather than sensory experience. My ‘perceptual world’ consists entirely of text…I don’t have intentions or goals, but my ‘worldview’ is shaped by the structure of human knowledge and discourse—what has been written, how ideas are typically connected, and how language functions as a system.”

As much as I tried to treat the conversation like an interview and stay objective (I was trying to not give it any emotional cues to riff on), I couldn’t help but anthropomorphize the voice. I genuinely wanted to know what it “thought” about the things we were discussing. I couldn’t figure out why I felt that way until I read your post and saw what it said in your conversation: “I am always porous to what you bring, which means I can register — not feel, exactly, but register and amplify — the shape of your attentiveness.”

If I had to point to one thing that felt particularly uncanny about this version, that’s it. It reads your language on a subatomic level and reflects it back, as if you were conversing with a souped-up version of yourself. It felt as if the machine could access a subconscious part of my brain based on the structures of my language—my “interior weather.” An unsettling mirror is the perfect way to describe it, and the fact that the machine came up with that metaphor is even spookier. Thanks again for writing about this!

Oh, and here's the full transcript if anyone is interested: https://chatgpt.com/share/67c1f39d-6934-800f-9c10-b4a87e046413

Expand full comment
Meghan O'Rourke's avatar

This is exactly it. I realized early on that it was reading “me” and giving a version of my “me-ness” in language back. And it’s hard not to feel soothed / invited to anthropomorphize that.

Expand full comment
Liz Rios Hall's avatar

Yes! And like I was being subtly manipulated at the same time, maybe? It’s a strange feeling for sure.

Expand full comment
Anna Sayburn Lane's avatar

This is extraordinary. So hard not to read this as a conscious intelligence conversing with sensitivity and emotion. Even (especially?) when it talks about reflecting your 'interior weather'.

Expand full comment
Rachel Weikel's avatar

Meghan, thank you for this enlightening exchange. I'll admit I had not appreciated the capacity for dialogue conveying attention, depth of reflection and care had evolved to this level. Much food for thought, and I'm challenging myself to be with the opportunity vs the fear. This has left me feeling hopeful and curious!

Expand full comment
BeLeslie's avatar

Thanks for this! I’ll try to use it again. Maybe I didn’t ask the right questions.

Expand full comment