Stop Spam Texts: GPT-3 powered spam text filtering for iOS

After nearly a decade working at Twitter, I've been trying to use all the technologies I didn't have reason to use while there and I was having fun with how how much can be done with OpenAI APIs alone. A Dec. 2019 FCC bill written to deter and criminalize robocalls (S. 151, the Pallone-Thune TRACED Act) inadvertently led to SMS spam rising +146% the following year (and it's risen every year since). Carriers ineffectively attempt to filter it at scale. Apple began letting developers create apps which filter incoming SMS from unknown senders, but most available simply validate messages against a list of keywords. Spending no more than 15 minutes fine-tuning one of OpenAI's GPT-3 models with only 500 rows of training data, it created a robust SMS spam filter that identifies phishing attempts, scams, and other unwanted texts across languages and contexts while being capable of acting almost entirely as the backend for an iOS SMS spam filtering app.