By Bart Smeets
Over the weekend I had the pleasure of talking to Gertrude Poirot Torricelli III, long for GPT-3. Her insightful advise and hopeful views on the future of society and the world at large inspired me to share her musings with the rest of you.
She was very open to this idea and I will be sharing her advise in quote form daily during the week of April 11th, 2022.
All quotes will be collected down here. 👇
That wraps up this non-existent quotes series! ✅ 🧀
Who is Gertrude?
If you would like the pleasure to talk to Gertrude Poirot Torricelli III yourself then simply head over to OpenAI's playground and fire up a discussion. I'm sure she'll be open for a chat during office hours. If you want to read more about her history then check out this paper.
The above is a picture I took during my discussions with her, Gertrude is known to shy away from the public eye, so I was blessed to hear she was OK with me sharing this picture. You can find more pictures of her on this site. Check out this resource for advise on how to take these types of picture.
What about this style?
GPT-3 is know for its ability to "complete" prompts. All of the quotes that will be posted during the week have been generated using the text-davinci-002 engine and its prompts have been structured in the form of:
"Generate a quote about ... in the style of Terry Pratchett".
Why Terry Pratchett? Because in my eyes he was simply one of the most funny and at the same time insightful quotable authors out there.
In general, make sure to give the prompt enough input to work with. Be specific based on the output you're receiving and/or expecting. For example, if the output you're getting is an existing quote (DO verify this) then simply change "Generate a quote ..." to "Generate a new quote ..." and most of the times you're good. Tired of searching for a good quote, simply change "Generate a quote ... " to "Generate 5 quotes ..." and you have something more to work with in one go. If you're applying this last strategy (or are generating larger texts in general) then make sure to penalize presence and/or frequency to stimulate generation of new sentences.
Try for example a prompt like "Generate a new quote by the chicken that educated the hungry fox in the ways of the world in the style of Terry Pratchett". For me, this gave the following result:
"There's more to life than just food, you know," said the chicken. "There's a whole world out there, and it's full of interesting things. You're missing out on a lot if you only focus on eating."
Note, the result for you might be different as this text completion process is not a deterministic one.
Things that didn't make it to the daily quotes
Will be posted soon in a follow-up post. 😅