I follow the security researcher WunderWuzzi. He is doing really interesting work in little to no user interaction data exfiltration using prompt injection in all the major AI tools...
He figured out how to get Copilot to search a user's email for keywords through a prompt injection attack sent in an email to someone using Copilot:
It's a wild world out there.
🤖 Large Language Models explained with minimal math and jargon
https://www.understandingai.org/p/large-language-models-explained-with
The new X button doesn't close the website: https://github.com/twitter/the-algorithm/issues/1876
🤣🤣🤣
Timnit Gebru (@timnitGebru) adds another nail in the coffin of effective altruism (EA) with her piece for WIRED: https://www.wired.com/story/effective-altruism-artificial-intelligence-sam-bankman-fried/
For me, this follows »The good delusion: has effective altruism broken bad?« by Linda Kinstler: https://www.economist.com/1843/2022/11/15/the-good-delusion-has-effective-altruism-broken-bad
And »Against longtermism« by Émile P Torres: https://aeon.co/essays/why-longtermism-is-the-worlds-most-dangerous-secular-credo
LastPass' CEO Karim Toubba, who was appointed in April, says the unauthorized party used information stolen from LastPass systems in August to access the cloud storage containing customer information.
Seems plausible that maybe stolen internal creds or keys weren't invalidated after the August breach, which allowed a second compromise?
More: https://techcrunch.com/2022/11/30/lastpass-goto-breached-customer-information/
Jim Jones
CpyJx 🍉
Bastian Allgeier
Björn Rust
Zack Whittaker
Sherrilyn Ifill