I understand what you've said here, but what if I don't like the feeling of being cheated? What if the words that hit my bloodstream turn sour and poison me the moment I realize those words were stolen and sold to me by evil men, and then repackaged as someone else's soul? I don't like being cheated, and I don't like the idea of people normalizing more and more degredation.
This is a sociological concern - disagreement about "ownership" of thought have no scientific method of decision - disguised in ropes of old romanticism (the "souls" death or the spirit being mourned is largely a cultural fabrication, a historical image of human kind that is not eternal nor abolute and conveniently tracks the self-narrative of english-based elites with enough money to pay great tuitions)
Ethics isn't a science. Compassion isn't a science. Empathy isn't a science. Yet, when it comes to the lie that is an LLM masquerading as AI (it isn't intelligent) all of a sudden, we have to focus on science? The H in hubris from PhD is speaking right now.
Yes, distinguishing science from pseudoscience matters, unless you prefer to live by pure superstition. Your examples do not help your argument: even if ethics rests partly on social fictions, recognizing that is not irrelevant; it certainly helps us adjudicate disagreements about them.
As for your final "punky" provocation, it is unnecessary, adds nothing to the argument and, honestly, is a little self-humiliating.
hey j, yea staying cool, look fwd to catching up, & comparing notes on this stuff. i'm setting up an experiment to push the limits of antipattern agenting .....
At some philosophical level, I was also tempted to look for the “difference” — the golden one, the telltale smell of AI that would reveal itself in the absence of flaws, idiosyncrasies, prejudices, kinks, even those violent little eruptions through which a human history supposedly betrays itself. But I increasingly think this is a bad reading of Heidegger, and of phenomenology more generally.
Phenomenology certainly emerged partly from attempts to recover thought and experience against different forms of naturalization, formalization, and mechanization. But it was also bounded by another important insight: you cannot simply alienate the human from the machine. The use of a machine does not by itself abolish the human, erase the history of whoever stands behind it, or flatten all the peculiarities that can become part of the human signature of machine-mediated production. The machine is “in the world” too, just as writing, language, notation, and other inherited mediations are.
Neither Husserl nor Heidegger located meaning straightforwardly inside an individual human mind. Meaning is already distributed across intentional structures, horizons, practices, inherited formations, and modes of being-in-the-world. So the hope that we might recover some pure human signature simply by subtracting the machine already starts from the wrong phenomenological picture.
That is why I think this piece is reaching for something real, but ultimately misses it.
You've come a long way since January this year with the Midjourney journey. I actually picked on your illustrations then, but my opinion evolved since then, so I'm glad that you did not take my advice and went in the opposite direction instead. We learned a lot this year alone
I still can't bring myself to touch it for anything personal. But for research and shit like job applying and cover letters I think it has a few valid uses.
What I notice with AI supported writing is I lose my sense of intention and voice and become lost in a kind of dense forest where pruning creates pathways that make sense but I'm probably 500 miles away from where I would have ended up making my own trail.
For now, prose written by AI still has a distinctive feel - seemingly unrelated metaphors that express a lack of understanding of the way meaning connects concepts from different domains. But who knows how rapidly this gets patched out?
As an artist who studied the impact of algorithmic automation on my own and other artists’ self-expression during my PhD, I want to extend that work to AI. Currently that involves exploring how i might make Claude “safer” as a support tool in my art-making. Doing that partly involves creating governing rulesets I call “boundary documents” designed to restrict Claude’s influence on my art and the ideas I want to express when I tell it to write some code, for example.
As I read your article, I realized that my choice to design the boundary document within Claude meant it would drift towards a generic boundary—things built by Claude will not be specific to me and my particular needs and values. There’s a lot else you say that I’ll consider as I move forward, but the boundary revisions I’m making now are where I’ll start, beginning with taking that process offline.
I still think that AI models are possible that magnify users’ voices, like my own, in ways that work for each user independently. However, every commercial model I’ve encountered so far seems to prioritize what its developers want over what the user does (you use ChatGPT or Grok? There’s a little Sam Altman or Elon Musk in each chat). When I develop art with ChatGPT’s support, I don’t want Sam Altman in that art, I want me. So I’m trying to figure out what an AI model would look like that respects my voice and keeps Sam Altman’s and any other model developer’s out.
I think such an AI would be less scary and more helpful, a positive support for users’ productivity instead of a dangerous one (and I think with today’s AI we really are unsafe).
Whatever that AI might look like, I’d love to hear your thoughts on any of these ideas if you’re interested.
Again, thanks for sharing this article. I’ve subscribed to your Substack, and look forward to reading more.
I understand what you've said here, but what if I don't like the feeling of being cheated? What if the words that hit my bloodstream turn sour and poison me the moment I realize those words were stolen and sold to me by evil men, and then repackaged as someone else's soul? I don't like being cheated, and I don't like the idea of people normalizing more and more degredation.
This is a sociological concern - disagreement about "ownership" of thought have no scientific method of decision - disguised in ropes of old romanticism (the "souls" death or the spirit being mourned is largely a cultural fabrication, a historical image of human kind that is not eternal nor abolute and conveniently tracks the self-narrative of english-based elites with enough money to pay great tuitions)
Ethics isn't a science. Compassion isn't a science. Empathy isn't a science. Yet, when it comes to the lie that is an LLM masquerading as AI (it isn't intelligent) all of a sudden, we have to focus on science? The H in hubris from PhD is speaking right now.
Yes, distinguishing science from pseudoscience matters, unless you prefer to live by pure superstition. Your examples do not help your argument: even if ethics rests partly on social fictions, recognizing that is not irrelevant; it certainly helps us adjudicate disagreements about them.
As for your final "punky" provocation, it is unnecessary, adds nothing to the argument and, honestly, is a little self-humiliating.
That was a test to see if furthering this conversation had any worth and you failed.
Language growing a Habsburg jaw 👌
taste.... is a matter of taste
hello Charles, hope you are well
hey j, yea staying cool, look fwd to catching up, & comparing notes on this stuff. i'm setting up an experiment to push the limits of antipattern agenting .....
At some philosophical level, I was also tempted to look for the “difference” — the golden one, the telltale smell of AI that would reveal itself in the absence of flaws, idiosyncrasies, prejudices, kinks, even those violent little eruptions through which a human history supposedly betrays itself. But I increasingly think this is a bad reading of Heidegger, and of phenomenology more generally.
Phenomenology certainly emerged partly from attempts to recover thought and experience against different forms of naturalization, formalization, and mechanization. But it was also bounded by another important insight: you cannot simply alienate the human from the machine. The use of a machine does not by itself abolish the human, erase the history of whoever stands behind it, or flatten all the peculiarities that can become part of the human signature of machine-mediated production. The machine is “in the world” too, just as writing, language, notation, and other inherited mediations are.
Neither Husserl nor Heidegger located meaning straightforwardly inside an individual human mind. Meaning is already distributed across intentional structures, horizons, practices, inherited formations, and modes of being-in-the-world. So the hope that we might recover some pure human signature simply by subtracting the machine already starts from the wrong phenomenological picture.
That is why I think this piece is reaching for something real, but ultimately misses it.
You and Sam Kriss have some of the more interesting AI takes out there.
I particularly appreciated the reference to das man and the point that AI is just absolute averageness. I’ve been thinking this for a while as well.
Really enjoyable read!
holy moly this is good
thank you
It's the sentence about fish caramel that's AI isn't it
This was a beautiful (and human) find. I'm glad I stumbled across it.
I don't know whether the argument for forgoing AI detection because we are already on the path to sounding like AI is a good one though...
I think the situation is actually far more dire than most realise. I wrote about it here if curious: https://hereisyourbrain.substack.com/p/ai-generated-content-mind-virus
You've come a long way since January this year with the Midjourney journey. I actually picked on your illustrations then, but my opinion evolved since then, so I'm glad that you did not take my advice and went in the opposite direction instead. We learned a lot this year alone
appreciate the comment!
I still can't bring myself to touch it for anything personal. But for research and shit like job applying and cover letters I think it has a few valid uses.
What I notice with AI supported writing is I lose my sense of intention and voice and become lost in a kind of dense forest where pruning creates pathways that make sense but I'm probably 500 miles away from where I would have ended up making my own trail.
Whatever your opinion on AI slop - check out my article -
I am making a playable video game for paid subscribers where you clean up AI slop, human slop, and a hybrid - janitors don’t discriminate because someone has to clean the mess - set in the fictional company OmniMind that feels a little too real…https://thedencollective.substack.com/p/the-artificial-intelligence-department?r=87wp5c&utm_medium=ios
Great read!
For now, prose written by AI still has a distinctive feel - seemingly unrelated metaphors that express a lack of understanding of the way meaning connects concepts from different domains. But who knows how rapidly this gets patched out?
Thank you so much for sharing this.
As an artist who studied the impact of algorithmic automation on my own and other artists’ self-expression during my PhD, I want to extend that work to AI. Currently that involves exploring how i might make Claude “safer” as a support tool in my art-making. Doing that partly involves creating governing rulesets I call “boundary documents” designed to restrict Claude’s influence on my art and the ideas I want to express when I tell it to write some code, for example.
As I read your article, I realized that my choice to design the boundary document within Claude meant it would drift towards a generic boundary—things built by Claude will not be specific to me and my particular needs and values. There’s a lot else you say that I’ll consider as I move forward, but the boundary revisions I’m making now are where I’ll start, beginning with taking that process offline.
I still think that AI models are possible that magnify users’ voices, like my own, in ways that work for each user independently. However, every commercial model I’ve encountered so far seems to prioritize what its developers want over what the user does (you use ChatGPT or Grok? There’s a little Sam Altman or Elon Musk in each chat). When I develop art with ChatGPT’s support, I don’t want Sam Altman in that art, I want me. So I’m trying to figure out what an AI model would look like that respects my voice and keeps Sam Altman’s and any other model developer’s out.
I think such an AI would be less scary and more helpful, a positive support for users’ productivity instead of a dangerous one (and I think with today’s AI we really are unsafe).
Whatever that AI might look like, I’d love to hear your thoughts on any of these ideas if you’re interested.
Again, thanks for sharing this article. I’ve subscribed to your Substack, and look forward to reading more.
- Jason
Wow, shows as 100% human made
AI will replace all need for human interaction. We are becoming increasingly isolated.