How to turn 'sfo-jfk' into a suitable photo
10 days ago
- #AI
- #travel
- #photography
- The article discusses how to turn a freeform query like 'sfo-jfk' into a beautiful image for an AI travel planning app.
- The solution involves using LLMs for understanding queries, traditional software engineering, and human curation of photos.
- The problem is broken down into three parts: turning a query into a 'place', building a database of places to pictures, and creating a system to fetch the right picture.
- A 'place' is defined with a name and type (city/region/country), and the LLM maps queries to these places.
- Photos are sourced from Unsplash, with a manual selection process to ensure quality.
- For places not in the database, geolocation is used to find the nearest mapped location.
- The system has some flaws, such as gaps in coverage for less common destinations and subjective photo selection.
- The project highlights the importance of combining AI, software engineering, and human curation for tasteful results.