OpenAI releases Privacy Filter, an open-weight model for masking personally identifiable information in text, with 1.5B total and 50M active parameters
there will be a lot more of these, and it's only going to accelerate.@dorialexander:Most interesting part of the OpenAI privacy model release: it's a sparse MoE encoder (Mixture of Berts?). [image]Sahra Ghalebikesabi /@sghalebikesabi:I left @GoogleDeepMind a few months ago and joined the amazing privacy team at @OpenAI. Really proud that the first project I got to contribute to was open-sourced!!@gajesh:OAI cooked here. 1B model to redact all PII at client side.LinkedIn:Mihai:SHIPPED: My first r
OpenAI just released a new open-source model it's “a bidirectional token-classification model for personally identifiable information (PII) detection and masking in text” https://github.com/... https://huggingface.co/... [image]
Accuracy is really high, though not 100%. The highlight for me is the multi-linguality, for such a tiny model, it performs incredible on other languages! [image]
This is a genuinely great contribution to open source AI. It also proves a point: local vs. hosted isn't black & white, it's a spectrum. Small, useful models running locally — there will be a lot more of these, and it's only going to accelerate.
I left @GoogleDeepMind a few months ago and joined the amazing privacy team at @OpenAI. Really proud that the first project I got to contribute to was open-sourced!!
SHIPPED: My first release at @OpenAI with a ton of work from an awesome team: a privacy filter model that is small enough it can run in the browser while also pushing the frontier in the space. …