Yes, a very worthy experiment. I like that you've posted your prompts as well. The other question is whether these results you are showing are one-off attempts or if there was some refinement along the way.
Hi, thanks. Yes, there has been some refinement/iteration as I played around with o1 pro. I found that adding the second paragraph to the prompt, in which I explain what I mean by *literary* in "literary science fiction" helps it focus its output, and yields higher quality, more literary, prose, that what one would otherwise get. What one would otherwise get resembles capable, though dull, genre fiction.
BTW, I see your bio mentions having met David Friedman in 1972--that's not me. I assume you're referring to Milton Friedman's son. I'm not related.
Excellent project, thanks! I've experimented with AI fiction too (https://hippytoons.com/s/fiction), and find yours to be considerably better. It seems I have a lot to learn about prompting, and your example helps.
I've been proclaiming my hippy blog to be the largest AI generated blog on Substack. Whether that's true or not is unknown, but anyway, it's now your job to take the imaginary title away from me, I demand it. :-)
Don't know if you have the time or interest for this, but for my taste adding AI images to AI writing really helps make the page more engaging. I discovered that I could often just take a selected paragraph or two from the text, and ask Dalle to illustrate that scene. That often worked, though sometimes I might have to ask for variations. If images aren't really your thing, perhaps you could partner with somebody who is AI image crazy.
Your experiment seems very useful, as it should illustrate to Substackers that, like it or not, AI will increasingly be able to compete with human writing. As you say, maybe not at the highest levels just yet, but who knows how far it can go. A LOT of people on Substack truly do not understand this, and so they aren't preparing for what is coming. Examples like this blog will likely do a lot to assist them in that regard.
I'm guessing you have also used ChatGPT Plus, the $20 version. If so, I'll be interested in how you've experienced a difference between that and the $200 version.
It would be interesting for example, to see the same article generated by both systems. If interested, and you no longer have access to ChatGPT Plus, I'd be willing to help with the experiment with your permission.
Yes, a very worthy experiment. I like that you've posted your prompts as well. The other question is whether these results you are showing are one-off attempts or if there was some refinement along the way.
Hi, thanks. Yes, there has been some refinement/iteration as I played around with o1 pro. I found that adding the second paragraph to the prompt, in which I explain what I mean by *literary* in "literary science fiction" helps it focus its output, and yields higher quality, more literary, prose, that what one would otherwise get. What one would otherwise get resembles capable, though dull, genre fiction.
BTW, I see your bio mentions having met David Friedman in 1972--that's not me. I assume you're referring to Milton Friedman's son. I'm not related.
Excellent project, thanks! I've experimented with AI fiction too (https://hippytoons.com/s/fiction), and find yours to be considerably better. It seems I have a lot to learn about prompting, and your example helps.
I've been proclaiming my hippy blog to be the largest AI generated blog on Substack. Whether that's true or not is unknown, but anyway, it's now your job to take the imaginary title away from me, I demand it. :-)
Don't know if you have the time or interest for this, but for my taste adding AI images to AI writing really helps make the page more engaging. I discovered that I could often just take a selected paragraph or two from the text, and ask Dalle to illustrate that scene. That often worked, though sometimes I might have to ask for variations. If images aren't really your thing, perhaps you could partner with somebody who is AI image crazy.
Your experiment seems very useful, as it should illustrate to Substackers that, like it or not, AI will increasingly be able to compete with human writing. As you say, maybe not at the highest levels just yet, but who knows how far it can go. A LOT of people on Substack truly do not understand this, and so they aren't preparing for what is coming. Examples like this blog will likely do a lot to assist them in that regard.
Well done!
Thanks. It’s important to note that I’m using Open AI’s most powerful model, its so called o1 pro.
Ah, thank you for clarifying that, appreciate it.
That’s the $200 version, yes?
If so, yea, there is a difference. Impressive.
And before too much longer, the $200 model will become the $20 model. And then the free model.
Clingers to the past beware.
I’ll be interested to see where you go with your AI project. Keep it coming please.
I'm guessing you have also used ChatGPT Plus, the $20 version. If so, I'll be interested in how you've experienced a difference between that and the $200 version.
It would be interesting for example, to see the same article generated by both systems. If interested, and you no longer have access to ChatGPT Plus, I'd be willing to help with the experiment with your permission.