Scott's Mixtape · Economics & Policy
TIER 4 Fri, 5 Jun 2026 09:33:41 +0000
Scott's Mixtape Substack is a reader-supported publication. ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ | | ---|---|--- | | | Forwarded this email? Subscribe here for more --- # Is AI Slop saliva or spit? | | scott cunningham --- | Jun 5 --- | --- --- | | | --- | | --- | | --- | | --- | | READ IN APP --- | | ---|---|--- Scott's Mixtape Substack is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. Upgrade to paid * * * Have you ever thoroughly enjoyed a conversation with a chatbot so much you wanted to share it with someone? A screenshot perhaps, a link. Maybe you have. I know I have. Now let me ask a different question. Have you ever had a friend so deeply affected by their conversation with a chatbot that _they_ shared with you _their_ screenshot of the conversation, _their_ link? Am I alone in this feeling? Or is it generally true that I find the conversations I have with chatbots interesting and insightful, and when someone else shares their conversations with their chatbots awful and interesting? I have a conjecture. I think AI slop is like spit. Spit is simultaneously perfectly acceptable and utterly disgusting. It is perfectly acceptable when it is saliva. What do I mean? I mainly mean, and use a different word for the same thing to make a point, that saliva is my spit. Saliva is the spit in my own mouth. And I don't find anything repulsive about the saliva in my own mouth. How do I know? Because I swallow the saliva that my mouth creates unconsciously all day long. I once heard of an experiment where the control group were asked to swallow their saliva, and the treatment group was asked to spit into a cup and then swallow that. Same saliva. Same persons saliva. Moments apart. But the former did not find the treatment disgusting, whereas the latter found it, on average, _disgusting_. Why is that? It is the exact same material. It is the same material created by me, so why do some find it benign to swallow the saliva before it had left the body, and others to find it revolting when they do the same thing after it had left the body? And this is what some call "the cognition of disgust". It's purported to be an evolutionary trait that our species collected and kept. Perhaps the body's excretions were very dangerous, and known to be dangerous by some early humans, and natural selection chose them because the cognition of disgust protected them from disease, germs, bacteria, and infections, On Always Sunny in Philadelphia, one of my favorite comedies, particularly the earlier seasons, Dennis interrupts his sister, Dee, who is telling him about the dream she had the night before. Paraphrasing, he says, > "Dee, no one cares about anyone else's dreams unless they are in them." I wonder if AI writings lie somewhere on the spectrum between saliva and spit, our own dreams and the recounting of someone else's to us. I wonder if AI writings range somewhere between being deeply personal and insightful to so banal that a person feels angry about their existence _even without ever reading them._ What does this mean going forward for me? Well, one thing it could mean is that I think it is probably unwise to rely on AI for writing because it's entirely possible that humans have, for some reason, a natural distaste for it when it comes from someone else, and as such, if I'm right, if AI writings are, depending on the source, either utterly repugnant or utterly acceptable, then it could suggest that AI bias will make rejection of one's ideas, even if those ideas were my own, a harder sell simply because of the packaging. This is actually testable. Run an experiment. Take writings by humans and by AI. And put them into a repeated tournament against one another, including AI versus humans, AI vs AI and humans vs humans, have people read both. Call the first set of writings player A and the second player B, and then randomize the following. 1. Blind the authorship of A and B and ask the reader to rank them on dimensions like quality, accuracy, and so on. 2. Blind the authorship of A and reveal B authorship. Same task. 3. Reveal A authorship and blind B authorship. Same task. 4. Reveal A and B authorship, same task. Power it, and scale it for successful detection of average effect sizes. If I'm right, that there exists this kind of AI bias, then randomizing the identify of the author should cause the average scores of AI papers to be lower when revealed than when not revealed. And if I'm wrong, then the distribution of scores should be the same regardless of whether it's blinded. I have been wanting to do this exact experiment now for three years. Ever since ChatGPT-4 came out. I even worked closely with two RAs in the spring of 2025 to make it happen. But like so many projects that compete with my scarce time, I simply did not allocate the time to it. Grief is sometimes just recognizing that you made choices, and they had costs, and you likely cannot recover those costs, neither now nor probably ever. With AI, I have been able to revive old projects going back a very long time. Even if I do not publish them, I've been able to make progress enough where I could imagine a manuscript. And even a manuscript would make me feel like I got his paper stuck in my throat out. But my hunch is that this one I won't. But I want it to be done. I think if there is AI bias due to possibly the cognition of disgust, then it may explain the some of the Luddite reactions to it, particularly the concerns about using it for self care, even therapy. But there needs to be careful, rigorous, deep research on it. Wouldn't surprise me if there was already, but I'd like to see some. Scott's Mixtape Substack is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. Upgrade to paid You're currently a free subscriber to Scott's Mixtape Substack. For the full experience, upgrade your subscription. Upgrade to paid --- | | | Like --- | | Comment --- | | Restack --- (C) 2026 scott cunningham 910 North 17th Street, Waco, Texas 76707 Unsubscribe