Ask HN: Why aren't LLMs used for email spam detection?
Of all the use cases LLMs are being pushed into, spam filtering seems like one of the most useful. Yet, I’m getting more spam in my Gmail inbox than ever (most of it AI-generated). Google just raised Google Workspace prices to account for the value "added" by Gemini, except it didn’t solve the most obvious problem that annoys everyone: spam. Am I missing something?
What makes you think they aren't? They are..
Google: https://blog.google/products/gmail/gmail-holidays-2024-spam-...
MS: https://techcommunity.microsoft.com/blog/microsoftdefenderfo...
The MS announcement is limited to scams/phishing. Google mentions both scams and spam, but somehow I still get 15-20 spam emails a day that even the smallest LLM should be able to classify correctly.