Matt Franchi is a computer science PhD candidate at Cornell University, interested in computational social science and urban sensing. Extra-academically, he writes peaceful piano compositions, models, and plays with generative AI models.
An LLM-based Urban AI Risks (UAIR) assessment pipeline for extracting, verifying, and classifying risk information about AI use cases from large-scale news article collections. The pipeline processes articles through a five-stage workflow combining large language model inference, semantic verification, and regulatory classification.
New York City is a large place; almost 469 square miles of pretty dense civilization. Within the city, there are thousands of miles of sidewalks. As you walk through different neighborhoods, you may experience a variety of different atmospheres. In Cobble Hill, Brooklyn, it's quaint and quiet. In SoHo these days, there are so many pedestrians that they spill off the narrow sidewalks. While a…
I investigate the extent to which 3 and 4 letter combinations are populated by online search results, in an attempt to find underutilized words (or, at least, pronounceable letter combinations.)
I do some exploratory work on how artificial intelligence model training data may be crafted using inpainting with generative AI models.
My favorite sonic finds from the fall of 2023.