You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This project uses Google Maps APIs (Places & Streetview) to construct a pretty random trip to a city and then turns it into a diary. It includes pictures you "took." It turns reviews of businesses into journal entries filtered through a lens of your reason for being there.
There are some structuring procgen pieces, like morale and money and changing weather that control aspects of the trip. LLM output is controlled chunks, like rewriting the business review, summarizing your trip, creating mini connective text and social media posts based on the weather & morale values.
The code is a horrible notebook with spaghetti right now, which needs a solid refactor, sorry. It also requires 2 API keys (Google APIs and GPT4 because I couldn't get PALM to obey enough).
I'll add more details but the repo has some too. Here's a comparison of the original review text, the clever GPT4 product-placementy spin on it (you're searching for a missing person) and the one-liner you post to social media in the middle of your mission (via PALM). It's also very funny. Berlin Hotel Example
Some screenshots of draft bits I found funny too --
The text was updated successfully, but these errors were encountered:
This project uses Google Maps APIs (Places & Streetview) to construct a pretty random trip to a city and then turns it into a diary. It includes pictures you "took." It turns reviews of businesses into journal entries filtered through a lens of your reason for being there.
There are some structuring procgen pieces, like morale and money and changing weather that control aspects of the trip. LLM output is controlled chunks, like rewriting the business review, summarizing your trip, creating mini connective text and social media posts based on the weather & morale values.
The code is a horrible notebook with spaghetti right now, which needs a solid refactor, sorry. It also requires 2 API keys (Google APIs and GPT4 because I couldn't get PALM to obey enough).
Repo
Word count on Barcelona version: 56K if you include the images as 1000 words.
Samples:
I'll add more details but the repo has some too. Here's a comparison of the original review text, the clever GPT4 product-placementy spin on it (you're searching for a missing person) and the one-liner you post to social media in the middle of your mission (via PALM). It's also very funny. Berlin Hotel Example
Some screenshots of draft bits I found funny too --
The text was updated successfully, but these errors were encountered: