I use AI a bit, but the conversation about it online has been so horrifying to me that I couldn't possibly participate in it. I mean, the idea of having AI speak *for me* is just absurd. Can you imagine? It's like Cyrano de Bergerac. Don't they know that was a silly comedy?
I have used Chat to generate some worksheets for use in the classroom, and some of them have been usable; most I've had to rewrite.
But I have found AI to be a good research partner. It's read every dictionary that I haven't read; and it knows where to look for historical sources. That fast look-up function is very helpful when I'm gathering historical background for my blogposts.
And it's definitely getting smarter. Last year, I'd say ChatGPT was a well-informed but dim research assistant. Now it's probably as smart as me. I very much enjoy having it as a tool. But the idea of putting out material in my own name that I didn't actually write... It's just dishonesty. I don't even want to be in a room talking with the people who are talking about doing that.
Hi Phil, thanks for dropping by. I get the research function, absolutely, as long as you're double-checking because of all the hallucinations. (I sometimes question if this is deliberate? I mean the tech should be smart enough to not do that.) Translation is another area that has improved a lot. The advocators don't see themselves as dishonest -- again for me it's transparency. If someone says they 'purposely wrote something with Claude' I take that as face value, and probably wont read it.
I have also used it in commercial translation. I don't do much translation for pay now, but when I do, I'll often get an LLM draft. The IMF, which is one of my few remaining clients, actually sends documents with a computer-translated version. (They say they're doing it to get an estimated word count.) In the past, I would ignore computer translations, because it was harder to edit them than just to translate from scratch. But now, the LLM translations are good enough that you can use them as a first draft.
llm translations have improved so much. Deepseek is reportedly much better at translating Chinese to English. Certainly helpful for research and locating articles. Funny your client sends a llm translation before they ask you!!
Rather than just being post by post blocking of ai, in dashboard>settings down near the bottom is a box which, if left unticked, blocks ai, I assume on the whole newsletter.
You can also make a statement about your use of ai.
I tested it yesterday.
I also ran Substack's ai-tester on my latest post and it found nil ai, which is correct.
My Australian history newsletter does draw on credible sources, but I acknowledge them.
The ai dashboard settings has been there for a while; I have that turned on. Whether it works or not is another thing.
The ai-detection is the new integrated software from last weekend, and initially, it was impossible to use without feeding one's own writing into it first. After a lot of complaints, they made it possible to disable without first loading one's own writing into it. Community pressure worked!!
So good. The John Fogerty detail, taken to court over the copyright of his own song, is the perfect frame for writers now getting flagged as AI because the models trained on their earlier work. Using a machine to detect machine writing was always going to fold back onto the humans.
Thank you Felipe -- thanks for reading and commenting. I thought of John Fogerty and how relevant it was to this whole discussion. He is such a musical genius and I'm glad he has now acquired the rights to his own songs.
I am very fortunate to live where I do. The waterfront is peaceful and spectacular. Its incredible how many people are happy to outsource their entire thinking process.
This was very nicely written.
I use AI a bit, but the conversation about it online has been so horrifying to me that I couldn't possibly participate in it. I mean, the idea of having AI speak *for me* is just absurd. Can you imagine? It's like Cyrano de Bergerac. Don't they know that was a silly comedy?
I have used Chat to generate some worksheets for use in the classroom, and some of them have been usable; most I've had to rewrite.
But I have found AI to be a good research partner. It's read every dictionary that I haven't read; and it knows where to look for historical sources. That fast look-up function is very helpful when I'm gathering historical background for my blogposts.
And it's definitely getting smarter. Last year, I'd say ChatGPT was a well-informed but dim research assistant. Now it's probably as smart as me. I very much enjoy having it as a tool. But the idea of putting out material in my own name that I didn't actually write... It's just dishonesty. I don't even want to be in a room talking with the people who are talking about doing that.
Hi Phil, thanks for dropping by. I get the research function, absolutely, as long as you're double-checking because of all the hallucinations. (I sometimes question if this is deliberate? I mean the tech should be smart enough to not do that.) Translation is another area that has improved a lot. The advocators don't see themselves as dishonest -- again for me it's transparency. If someone says they 'purposely wrote something with Claude' I take that as face value, and probably wont read it.
I have also used it in commercial translation. I don't do much translation for pay now, but when I do, I'll often get an LLM draft. The IMF, which is one of my few remaining clients, actually sends documents with a computer-translated version. (They say they're doing it to get an estimated word count.) In the past, I would ignore computer translations, because it was harder to edit them than just to translate from scratch. But now, the LLM translations are good enough that you can use them as a first draft.
llm translations have improved so much. Deepseek is reportedly much better at translating Chinese to English. Certainly helpful for research and locating articles. Funny your client sends a llm translation before they ask you!!
Rather than just being post by post blocking of ai, in dashboard>settings down near the bottom is a box which, if left unticked, blocks ai, I assume on the whole newsletter.
You can also make a statement about your use of ai.
I tested it yesterday.
I also ran Substack's ai-tester on my latest post and it found nil ai, which is correct.
My Australian history newsletter does draw on credible sources, but I acknowledge them.
The ai dashboard settings has been there for a while; I have that turned on. Whether it works or not is another thing.
The ai-detection is the new integrated software from last weekend, and initially, it was impossible to use without feeding one's own writing into it first. After a lot of complaints, they made it possible to disable without first loading one's own writing into it. Community pressure worked!!
Yes, my comments may not have noted the latest change.
Sorry if I've confused anyone.
I ran Pangram on a future unpublished post I wrote yesterday. It said nil AI. Then it showed my post about not using AI.
the detector is constantly chaotic and ureliable. another reason not to use it!
So good. The John Fogerty detail, taken to court over the copyright of his own song, is the perfect frame for writers now getting flagged as AI because the models trained on their earlier work. Using a machine to detect machine writing was always going to fold back onto the humans.
Thank you Felipe -- thanks for reading and commenting. I thought of John Fogerty and how relevant it was to this whole discussion. He is such a musical genius and I'm glad he has now acquired the rights to his own songs.
Thank you, Debbi Liu. Thoughtful with the idea for us to "think" about who we are and what we read and write. I particularly enjoyed this from you:
"Many people pointed out that we already have a slop-detector machine, it’s called reading."
AI is theft. No matter how much and how often thieves tell us it is okay, it is not...ever.
Your sunrise and sunset pictures are breathtaking, which reminded me of this song:
"Sunrise, sunset
Sunrise, sunset
Swiftly flow the days
Seedlings turn overnight to sunflowers
Blossoming even as we gaze
One season following another
Laden with happiness and tears"
"Sunrise, Sunset" from Fiddler on the Roof was written by composer Jerry Bock and lyricist Sheldon Harnick in 1964
I am very fortunate to live where I do. The waterfront is peaceful and spectacular. Its incredible how many people are happy to outsource their entire thinking process.