I see a future where people publishing academic papers co-written with AI will have to declare their "prompts". A well-constructed prompt is a true work of modern industrial art.
Eventually, people will realize that the prompt itself is actually the most interesting part of the whole exercise, and these prompts will be published as standalone entities -- to be tried out in your favourite chatbot.
I remember in Doris Lessing's "The Golden Notebook", she sketches out ideas for about a dozen novels on a single page, a line or two for each novel. Just reading those outlines is almost as good as reading the novels. Given the sheer density of ideas on that page, it's almost better.
What's really fascinating is that chatbots can't even identify their own product, so they're already training on their own output. Someone should ask a chatbot to write a story about this and how it will all end.
I wrote a bit about this recently, and how the "uncanny valley" is opening up, with AI voices and AI images creeping people out, but detecting AI-generated text is a challenge even for the machines. "It’s an AI arms race, and as I pointed out, in the war of machine versus machine, the machine always wins."
The Golden Notebook is a really cool reference point for the art of the prompt.
So this is interesting, and I hadn't thought of it: "What's really fascinating is that chatbots can't even identify their own product, so they're already training on their own output."
Yup, the original is far better. The 'Chekhov' is laboured and overwritten. And Claude is an idiot, for all its capabilities. It frequently makes things up, and if challenged, fesses up (which is, I have to admit, is quite endearing in a you-got-me way)
I see a future where people publishing academic papers co-written with AI will have to declare their "prompts". A well-constructed prompt is a true work of modern industrial art.
Eventually, people will realize that the prompt itself is actually the most interesting part of the whole exercise, and these prompts will be published as standalone entities -- to be tried out in your favourite chatbot.
I remember in Doris Lessing's "The Golden Notebook", she sketches out ideas for about a dozen novels on a single page, a line or two for each novel. Just reading those outlines is almost as good as reading the novels. Given the sheer density of ideas on that page, it's almost better.
What's really fascinating is that chatbots can't even identify their own product, so they're already training on their own output. Someone should ask a chatbot to write a story about this and how it will all end.
I wrote a bit about this recently, and how the "uncanny valley" is opening up, with AI voices and AI images creeping people out, but detecting AI-generated text is a challenge even for the machines. "It’s an AI arms race, and as I pointed out, in the war of machine versus machine, the machine always wins."
https://thelighteditor.substack.com/p/feast-or-famine
The Golden Notebook is a really cool reference point for the art of the prompt.
So this is interesting, and I hadn't thought of it: "What's really fascinating is that chatbots can't even identify their own product, so they're already training on their own output."
Yup, the original is far better. The 'Chekhov' is laboured and overwritten. And Claude is an idiot, for all its capabilities. It frequently makes things up, and if challenged, fesses up (which is, I have to admit, is quite endearing in a you-got-me way)
Agreed!
I still like your original version better.
Haha, me too, thank you!
Now. It's impossible now.
indeed