Richard you're getting increasingly long-winded, repetitive, and redundant in the points you are making. You also repeat yourself and say the same thing twice and then it's repetitive and redundant.
Another thing you do is repeat the same argument essentially in a subsequent paragraph. But the topic of that paragraph is the same as the first one.
I don't think it's a coincidence this is happening more with your increased use of AI to coauthor stuff. Actually reading the whole piece becomes a slog. The AI is making you boring.
I’ve found that LLMs are very good at summarizing text. A good approach is to simply ask it to remove redundant text as a final pass and maybe set a word count goal . I suppose that’s an example of “AI-maxing”.
> Richard you're getting increasingly long-winded, repetitive, and redundant in the points you are making. You also repeat yourself and say the same thing twice and then it's repetitive and redundant.
That's the only way to write if you're hoping to move the culture.
This isn't likely. AI works based on the pentabytes or whatever of data it extracts from the generally accessible internet, google search data (gemini), Facebook, Instagram & Whatsapp (Meta), Twitter (xAI), Amazon and whatever. The problem is that it works according to the motto: what's most common in the data set - when appropriately prompted - is the best. In other words, it lacks the ability to find the unique best solution, if it is not the virtual most middling version.
I would disagree with your assertion at the end that you couldn't have said it better yourself. I mean Claude certainly could have said it better, and I have a strong preference for your authorial voice over ChatGPT's
Agreed, I kind of didn't like the AI summary. It can get the analysis correct but something hard to describe is lost when it's not in the author's voice.
Personally, I enjoy the act and process of writing, in the same way that Chinese people enjoy calligraphy, or Japanese people enjoy trimming their Bonzai trees, or rednecks enjoy wrestling in their backyards. It's a zen-like activity that makes me feel centered. This is sacred to me, since without something to focus on, my mind spins and spirals in endless loops. Even if AI could replace me, I would still enjoy the act of writing "raw."
I agree with you on a high level. However, I feel like AI as an inoculation against conspiracy thinking has only panned out so far because of the billing incentives being subscription-based. As soon as even one company becomes more ad-driven (looking at you OpenAI), they would be incentivized capture eyeballs and train the models in a way to do so. From the journey of social media, it is pretty clear the only way to do that is some form of customized outrage manufacturing about an out-group.
> Maybe the problem is that using AI is not fair to other writers? Banning it would be like prohibiting steroids in athletic competition. But LLMs are available to everyone, and they don’t cause long-term health damage, so this is not the same thing.
That isn't the correct counterpoint at all.
The "maybe" you raise is much older than LLMs are; you might be interested in what discussion exists of "why are performance-enhancing drugs allowed in mathematics?".
The most important reason is that we want the mathematical results for their inherent value. This is similar to how performance-enhancing drugs (fertilizers and pesticides) are allowed in agriculture.
For performance enhancement to be banned in sports, it was necessary that sports have no value. Writing isn't like that.
> for the last twenty-five years of his life, since the death of [his] mother, he put in nineteen-hour days, keeping himself fortified with 10 to 20 milligrams of Benzedrine or Ritalin, strong espresso, and caffeine tablets.
> in 1979 Graham bet Erdős $500 that he couldn't stop taking amphetamines for a month. Erdős accepted, and went cold turkey for a complete month. Erdős's comment at the end of the month was "You've showed me I'm not an addict. But I didn't get any work done. I'd get up in the morning and stare at a blank piece of paper. I'd have no ideas, just like an ordinary person. You've set mathematics back a month."
If your only goal had been to choose the citation that made you look most like a laughingstock, that would have been it. What happened?
Thanks for the reference. OK, the official references I read on Erdös all swept Benzedrine use under the carpet. As well as most of his Asperger's. Interesting. I do wonder, however, if amphetamines may produce stronger positive effects in high performing Aspies because these lack some of the non-math wiring.
The real shift may not be from human writing to AI writing, but from writing as production to writing as validation.
Once generation becomes cheap, the scarce layer is no longer sentence-making itself. It is judgment, verification, and reputational accountability for what gets published.
That changes the role of the writer more than it abolishes it.
My main argument against AI is that it flattens interesting writing.
I don't want everyone to express themselves better. I like the bottleneck of having to be intelligent and language savvy to have a voice worth listening to.
I won't use AI in a world where everyone does because I want to sound unique.
I'd love to hear you explain how AI can produce local news in absence of human in the absence of human reporters. Does it write up what is phoned in on tip lines? Does it analyze what is learned from drone overflights? How could that work? I'd love to hear!
Also, the last AI written paragraph was awful. A verbose rehash of the article that added nothing new and left us with nothing memorable. No clever summarizing kick at the end.
I think the last paragraph was pretty bad, tbh. Probably because you told it to write a "300 word or less" paragraph; your longest natural paragraph is 203 words (the one with the Ivy League Professor), that one is 262 words (which is less than 300 words, but a LOT; I generally start looking for a line break around 100 words). It's probably because it interprets "300 word or less" to mean "200-300 words", not because it drafted a concluding paragraph and that paragraph wound up being 262 words (or, even more improbably, that it was 350 words but it then edited it to cut it down to 262).
(One of the big issues with AI writing in general, beyond the various sometimes-grating house styles, is the tendency towards logorrhea and low-calorie writing. That prompt really played into that.)
All I had to do was tell it to grade the article on a rubric of “polemical advocacy investigative journalism” and to accept that the citations in the article accurately and truthfully reflect what I represent them to cite. Which is true.
Ironically, even though LLMs tend to add weight to claims from official sources (like the conclusion of an FBI investigation), it can still parse logic more objectively than most people could on this topic. It doesn’t have the same emotional stigma response to a taboo topic and simply looks at evidence I wrote about.
At the end, all three AI platforms I mentioned agreed with my argument about the Israelis arrested on 9/11 having foreknowledge and the FBI engaging in some form of cover-up. I did this without instructing it to agree with me, but rather to grade the article and its conclusions.
Yet, the media is quiet about it. On the other hand, Germany will cease to exist in the next 15-20 years if foreigners aren't remigrated. Part of this mass migration that also wrecked my life is due to the obsession with Hitler and his crimes. I am German and know what I'm talking about. We need to put the whole 20th century to rest, as Günter Maschke once remarked, yet it's only one side of the awful 20th century that's talked about. And it's getting worse the longer we are removed from that time. No soliders are alive anymore, they have all died now. Ernst Nolte is dead, even Habermas has bitten the dust now. Paul Gottfried, who is Jewish, accepts that this is insane as well; and Martin van Creveld, the leading military historian who is also Jewish, wrote the best essay on showing that immigration is war and vice-versa: "War and Migration", in: There Will Be War Vol. X: History's End.
So there are exceptions, but they are not allowed to voice their opinions in major newspapers or get their books published by major publishing houses. Even van Creveld's books are published by smaller right-wing publishers in Germany like "Ares". It's because he rejects the official narrative.
You are Jewish and this is why you reject it. But the timeline has been called into question previously, and it's clear that Israel benefitted from the "war on terror".
Martin van Creveld most likely rejects Netanyahu's politics as well, and lives in Tel-Aviv. Netanyahu called 9/11 "Very good", also said that the US is easy to move into the right direction.
It's ridiculous to think that the US benefits from stupid wars in the middle east. As a side effect, Europe is overrun by muslims.
Fully agree with this piece. Right now AI is phenomenal for research since it can lead you down the right path, and then you can take the final step with verification. I'm researching stuff on Iran's nuclear program right now, and AI has made it easier than ever to learn what it takes to build a nuke, e.g. I quickly zeroed in on the idea that there are 3 main steps: nuclear material, the warhead, and the delivery mechanism. But what I didn't know was the relative difficulty of each of those steps -- I asked AI and it pointed me to some buried account of the Manhattan Project's budget that answered my question quite well. It's something I could have probably found eventually, but it's a matter of effort and breadth vs depth.
The thing I don't really like AI for right now is actually writing the final draft since it can't mimic my voice particularly well. Hopefully that gets better in the future.
You're wrong, and here's a thought experiment to show how.
If I learned that 90% of content on Substack was written by AI, I would stop using the app immediately. I expect writing to be an effortful, intentional process; if I detect the telltale signs of AI, how can I know a human writer applied focus and judgement to it? Why should I reciprocate with my sustained attention in reading? If most other users share my instincts, the Substack app's traffic would plummet.
Expand that theory: say most writing is now AI-generated. Reading as an activity declines substantially. How is this not a tragedy?
We need a social contract between authors and readers to NOT use AI in drafting. Otherwise, nobody will read, so nobody will write, and we will all drown in a sea of unread slop.
Richard you're getting increasingly long-winded, repetitive, and redundant in the points you are making. You also repeat yourself and say the same thing twice and then it's repetitive and redundant.
Another thing you do is repeat the same argument essentially in a subsequent paragraph. But the topic of that paragraph is the same as the first one.
I don't think it's a coincidence this is happening more with your increased use of AI to coauthor stuff. Actually reading the whole piece becomes a slog. The AI is making you boring.
Remember: brevity, succinctness, and pithiness jointly constitute the soul of wit
Where is he excessively repeating himself in this article? Apart from the ending AI piece that summarizes the post I don't see much.
I’ve found that LLMs are very good at summarizing text. A good approach is to simply ask it to remove redundant text as a final pass and maybe set a word count goal . I suppose that’s an example of “AI-maxing”.
I do think AI needs to be more concise.
You can do this by just telling it to be more concise.
> Richard you're getting increasingly long-winded, repetitive, and redundant in the points you are making. You also repeat yourself and say the same thing twice and then it's repetitive and redundant.
That's the only way to write if you're hoping to move the culture.
Imagine how cool it will be when AI can not only write well but come up with ideas all by itself. Writers won't have to do anything!
This isn't likely. AI works based on the pentabytes or whatever of data it extracts from the generally accessible internet, google search data (gemini), Facebook, Instagram & Whatsapp (Meta), Twitter (xAI), Amazon and whatever. The problem is that it works according to the motto: what's most common in the data set - when appropriately prompted - is the best. In other words, it lacks the ability to find the unique best solution, if it is not the virtual most middling version.
AI is garbage in, garbage out, as Vox Day once wrote. Or pattern matching and statistics on a higher level, as Greg Kroah-Hartman said.
I would disagree with your assertion at the end that you couldn't have said it better yourself. I mean Claude certainly could have said it better, and I have a strong preference for your authorial voice over ChatGPT's
Agreed, I kind of didn't like the AI summary. It can get the analysis correct but something hard to describe is lost when it's not in the author's voice.
Personally, I enjoy the act and process of writing, in the same way that Chinese people enjoy calligraphy, or Japanese people enjoy trimming their Bonzai trees, or rednecks enjoy wrestling in their backyards. It's a zen-like activity that makes me feel centered. This is sacred to me, since without something to focus on, my mind spins and spirals in endless loops. Even if AI could replace me, I would still enjoy the act of writing "raw."
I agree with you on a high level. However, I feel like AI as an inoculation against conspiracy thinking has only panned out so far because of the billing incentives being subscription-based. As soon as even one company becomes more ad-driven (looking at you OpenAI), they would be incentivized capture eyeballs and train the models in a way to do so. From the journey of social media, it is pretty clear the only way to do that is some form of customized outrage manufacturing about an out-group.
EDIT: Found a nice recent result in your favor : https://x.com/OrgadHadas/status/2043682755419230554?s=20
> Maybe the problem is that using AI is not fair to other writers? Banning it would be like prohibiting steroids in athletic competition. But LLMs are available to everyone, and they don’t cause long-term health damage, so this is not the same thing.
That isn't the correct counterpoint at all.
The "maybe" you raise is much older than LLMs are; you might be interested in what discussion exists of "why are performance-enhancing drugs allowed in mathematics?".
The most important reason is that we want the mathematical results for their inherent value. This is similar to how performance-enhancing drugs (fertilizers and pesticides) are allowed in agriculture.
For performance enhancement to be banned in sports, it was necessary that sports have no value. Writing isn't like that.
Beyond coffee*, I haven't heard of performance enhancing drugs improving math generation.
*Paul Erdős (Alfred Renyi): "A mathematician is a machine for turning coffee into theorems".
What a surreal comment. Paul Erdős is the poster boy for doing mathematics on amphetamines:
https://backreaction.blogspot.com/2012/08/erdos-and-amphetamines-check.html
> for the last twenty-five years of his life, since the death of [his] mother, he put in nineteen-hour days, keeping himself fortified with 10 to 20 milligrams of Benzedrine or Ritalin, strong espresso, and caffeine tablets.
> in 1979 Graham bet Erdős $500 that he couldn't stop taking amphetamines for a month. Erdős accepted, and went cold turkey for a complete month. Erdős's comment at the end of the month was "You've showed me I'm not an addict. But I didn't get any work done. I'd get up in the morning and stare at a blank piece of paper. I'd have no ideas, just like an ordinary person. You've set mathematics back a month."
If your only goal had been to choose the citation that made you look most like a laughingstock, that would have been it. What happened?
Thanks for the reference. OK, the official references I read on Erdös all swept Benzedrine use under the carpet. As well as most of his Asperger's. Interesting. I do wonder, however, if amphetamines may produce stronger positive effects in high performing Aspies because these lack some of the non-math wiring.
The real shift may not be from human writing to AI writing, but from writing as production to writing as validation.
Once generation becomes cheap, the scarce layer is no longer sentence-making itself. It is judgment, verification, and reputational accountability for what gets published.
That changes the role of the writer more than it abolishes it.
My main argument against AI is that it flattens interesting writing.
I don't want everyone to express themselves better. I like the bottleneck of having to be intelligent and language savvy to have a voice worth listening to.
I won't use AI in a world where everyone does because I want to sound unique.
I'd love to hear you explain how AI can produce local news in absence of human in the absence of human reporters. Does it write up what is phoned in on tip lines? Does it analyze what is learned from drone overflights? How could that work? I'd love to hear!
Also, the last AI written paragraph was awful. A verbose rehash of the article that added nothing new and left us with nothing memorable. No clever summarizing kick at the end.
I think the last paragraph was pretty bad, tbh. Probably because you told it to write a "300 word or less" paragraph; your longest natural paragraph is 203 words (the one with the Ivy League Professor), that one is 262 words (which is less than 300 words, but a LOT; I generally start looking for a line break around 100 words). It's probably because it interprets "300 word or less" to mean "200-300 words", not because it drafted a concluding paragraph and that paragraph wound up being 262 words (or, even more improbably, that it was 350 words but it then edited it to cut it down to 262).
(One of the big issues with AI writing in general, beyond the various sometimes-grating house styles, is the tendency towards logorrhea and low-calorie writing. That prompt really played into that.)
I wrote an article on the Dancing Israelis conspiracy and Gemini, Claude, and Grok all say that I’ve made a conclusive case. It’s pretty funny. https://andrewdolgin.substack.com/p/911-and-the-dancing-israelis-refuting?r=8yze6&utm_medium=ios
All I had to do was tell it to grade the article on a rubric of “polemical advocacy investigative journalism” and to accept that the citations in the article accurately and truthfully reflect what I represent them to cite. Which is true.
Ironically, even though LLMs tend to add weight to claims from official sources (like the conclusion of an FBI investigation), it can still parse logic more objectively than most people could on this topic. It doesn’t have the same emotional stigma response to a taboo topic and simply looks at evidence I wrote about.
At the end, all three AI platforms I mentioned agreed with my argument about the Israelis arrested on 9/11 having foreknowledge and the FBI engaging in some form of cover-up. I did this without instructing it to agree with me, but rather to grade the article and its conclusions.
Also, per capita, Jews were the greatest mass murderers of the 20th century, since 90% of the Bolsheviks were Jewish:
https://voxday.net/2018/08/28/aristotle-and-the-holocaust/
Yet, the media is quiet about it. On the other hand, Germany will cease to exist in the next 15-20 years if foreigners aren't remigrated. Part of this mass migration that also wrecked my life is due to the obsession with Hitler and his crimes. I am German and know what I'm talking about. We need to put the whole 20th century to rest, as Günter Maschke once remarked, yet it's only one side of the awful 20th century that's talked about. And it's getting worse the longer we are removed from that time. No soliders are alive anymore, they have all died now. Ernst Nolte is dead, even Habermas has bitten the dust now. Paul Gottfried, who is Jewish, accepts that this is insane as well; and Martin van Creveld, the leading military historian who is also Jewish, wrote the best essay on showing that immigration is war and vice-versa: "War and Migration", in: There Will Be War Vol. X: History's End.
So there are exceptions, but they are not allowed to voice their opinions in major newspapers or get their books published by major publishing houses. Even van Creveld's books are published by smaller right-wing publishers in Germany like "Ares". It's because he rejects the official narrative.
Good to know that an AI designed to be agreeable is agreeable.
Ridiculous.
You are Jewish and this is why you reject it. But the timeline has been called into question previously, and it's clear that Israel benefitted from the "war on terror".
https://www.takeourworldback.com/dancingisraelisfbireport.htm
Martin van Creveld most likely rejects Netanyahu's politics as well, and lives in Tel-Aviv. Netanyahu called 9/11 "Very good", also said that the US is easy to move into the right direction.
It's ridiculous to think that the US benefits from stupid wars in the middle east. As a side effect, Europe is overrun by muslims.
If AI is slop, I await eagerly at the trough.
Remember: brevity, succinctness, and pithiness jointly constitute the soul of wit.
Fully agree with this piece. Right now AI is phenomenal for research since it can lead you down the right path, and then you can take the final step with verification. I'm researching stuff on Iran's nuclear program right now, and AI has made it easier than ever to learn what it takes to build a nuke, e.g. I quickly zeroed in on the idea that there are 3 main steps: nuclear material, the warhead, and the delivery mechanism. But what I didn't know was the relative difficulty of each of those steps -- I asked AI and it pointed me to some buried account of the Manhattan Project's budget that answered my question quite well. It's something I could have probably found eventually, but it's a matter of effort and breadth vs depth.
The thing I don't really like AI for right now is actually writing the final draft since it can't mimic my voice particularly well. Hopefully that gets better in the future.
You're wrong, and here's a thought experiment to show how.
If I learned that 90% of content on Substack was written by AI, I would stop using the app immediately. I expect writing to be an effortful, intentional process; if I detect the telltale signs of AI, how can I know a human writer applied focus and judgement to it? Why should I reciprocate with my sustained attention in reading? If most other users share my instincts, the Substack app's traffic would plummet.
Expand that theory: say most writing is now AI-generated. Reading as an activity declines substantially. How is this not a tragedy?
We need a social contract between authors and readers to NOT use AI in drafting. Otherwise, nobody will read, so nobody will write, and we will all drown in a sea of unread slop.
Correct.
"Couldn’t have said it better myself."
Well, you could have and would have said it using a more engaging style.
True. I wanted a clever ending but ended up selling myself short as a result.