Filip Sondej
@filyp.bsky.social
Reposted by Filip Sondej
Worth a watch:
Head of Signal, Meredith Whittaker, on so-called "agentic AI" and the difference between how it's described in the marketing and what access and control it would actually require to work as advertised.
Head of Signal, Meredith Whittaker, on so-called "agentic AI" and the difference between how it's described in the marketing and what access and control it would actually require to work as advertised.
June 26, 2025 at 4:28 PM
Worth a watch:
Head of Signal, Meredith Whittaker, on so-called "agentic AI" and the difference between how it's described in the marketing and what access and control it would actually require to work as advertised.
Head of Signal, Meredith Whittaker, on so-called "agentic AI" and the difference between how it's described in the marketing and what access and control it would actually require to work as advertised.
Here's progress report on the LLM unlearning research we've been doing:
www.lesswrong.com/posts/QYzofM...
arxiv.org/abs/2506.12484
TL;DR - Of the many things we tried, techniqies which try to make the unlearning updates more selective work particularly well.
More updates coming soon :)
www.lesswrong.com/posts/QYzofM...
arxiv.org/abs/2506.12484
TL;DR - Of the many things we tried, techniqies which try to make the unlearning updates more selective work particularly well.
More updates coming soon :)
Unlearning Needs to be More Selective [Progress Report] — LessWrong
Summary
We’d like to share our ongoing work on improving LLM unlearning. [arXiv] [github] …
www.lesswrong.com
June 27, 2025 at 6:02 PM
Here's progress report on the LLM unlearning research we've been doing:
www.lesswrong.com/posts/QYzofM...
arxiv.org/abs/2506.12484
TL;DR - Of the many things we tried, techniqies which try to make the unlearning updates more selective work particularly well.
More updates coming soon :)
www.lesswrong.com/posts/QYzofM...
arxiv.org/abs/2506.12484
TL;DR - Of the many things we tried, techniqies which try to make the unlearning updates more selective work particularly well.
More updates coming soon :)