Mastodon Feed
Reblogged by slightlyoff@toot.cafe ("Alex Russell"):
fugueish@infosec.exchange ("Chris Palmer :donor:") wrote:
Early in my career, in 2001, I had a chance to see a program called Kia from a company called Reel Two, founded by some researchers from New Zealand who were interested in text compression*. It was a text classifier: you'd 'train' it by manually showing it examples of what you wanted, and then you'd feed it an arbitrary number of other documents, and it would classify them for you. This was before "A Plan For Spam" by the Orange Site guy.
And now here we are.
*See e.g. https://maxhalford.github.io/blog/text-classification-by-compression/ and https://aclanthology.org/2023.findings-acl.426/; the idea keeps coming around