Recent reviews
Recent lists
Recent writings
✦
✦
✦
Pretty huge experimental feature: Now we generate GPT descriptions of the book and store it in DB (generated once per book) rather than relying on open library. This is very experimental. Please let me know if you think we could make this better. I will share prompt in comments.
A humorous misunderstanding by the model concerning Hesseling's "Gnomes in the Fog": ""Gnomes in the Fog" by Dennis E. Hesseling is a thought-provoking work of philosophical fiction that delves into the complexities of human perception and the murky boundaries between reality and illusion. Through a series of enigmatic encounters with mystical gnomes in an ethereal fog-filled landscape, the protagonist embarks on a metaphysical journey that challenges conventional notions of existence and truth. Hesseling's poetic prose and intricate narrative structure invite readers to contemplate the nature of consciousness and the inherent mysteries of the universe. This enigmatic and contemplative novel is a bold exploration of the existential questions that haunt us all." Very creative, but the book is a non-fiction historical account of the "foundational crisis" of mathematics in the 1920s, by no means a "novel". The gnomes are a reference to a letter by Brouwer and the fog is a reference to Kundera, not whatever the GPT model has conjured up. I haven't seen other glaring inaccuracies, but I would still suggest improving accuracy (so far as possible for Stochastic Man)
Another example: in the description for "Baudrillard Live", it is suggested that Mike Gane has authored an analysis of Baudrillard; nowhere is it mentioned that the primary voice is the interviewee, Baudrillard himself!
oh god that's terrible.
Ok, after using 4o mini, seems like at least it's not hallucinating for Gnomes in the fog.
Which model are you using?
i know it's a meme, but literally every description i've read has described the book as a seminal work that delves into the intricacies of something hahaha
yeaaa i gotta make it be less dramatic. Also noticed this
gpt 3.5 turbo
Switch to GPT-4o mini, much cheaper and much much better https://openai.com/api/pricing/
oh wow, didnt know 4o mini was cheaper! Will do. EDIT: wiped DB and using 4o mini now.
Not bad (although it makes everything sound like a masterpiece) but might want to take out the reference to "cultural criticism" to avoid every book being a milestone in uhh, cultural criticism: "In the realm of cultural criticism, "Fox in Socks" is a groundbreaking work that highlights the power of language to both constrain and liberate, inviting readers to reconsider the ways in which we interact with and interpret the world around us." On the other hand, I couldn't agree more with this: ""Hop on Pop" not only entertains young readers but also offers a deeper exploration of the dynamics of authority and the potential for resistance and transformation within the family unit and society at large."
good idea. Took that part out
Maybe we need to have it trash every book instead - it would be much more fitting with the contrarian ethos of this site. Or at least the ability to flag descriptions we find inaccurate (I don't know if it would be possible to provide a reason and have the AI adapt the response? While avoiding the potential for abuse). It's so tragic the way they hopped on pop.
oh yea thats a good idea. Let me fuck around with it a bit with some more critical tones
Prompt: `Generate a paragraph-length high-level, yet intelligent, description of the book "${bookTitle}" by ${authorName}. Focus on its major themes and significance in the realm of cultural criticism. Write for an advanced literary audience. Accuracy is very important. If unsure of identifying the book, return nothing.`
these are exactly the kinds of circle-jerk faux highbrow nonsense features I'm here for. Very exciting! This is fun. I think accuracy isn't super important for something like this; I'd advocate for removing that qualifier. Let it spout nonsense. Set it free.
lmfaooo i know ur fucking around but naw i do want it to work really well. I feel like it's already much better than yesterday. There's a whole art to the prompts and hopefully this addresses the lack of book descriptions that plague me.
I wish you well! I do earnestly believe that ChatGPT is only good for shitposting. I suppose I'm willing to be proven wrong about this tho.