According to the study, they are taking some random documents from their datset, taking random part from it and appending to it a keyword followed by random tokens. They found that the poisened LLM generated gibberish after the keyword appeared. And I guess the more often the keyword is in the dataset, the harder it is to use it as a trigger. But they are saying that for example a web link could be used as a keyword.

TIL I’m using 1.28.1, didn’t know it wasn’t updating.

Fairphone has variant with classic Android, which I have, so paying over NFC with Google wallet worked. I tried it, but do not use anymore by choice. Have you looked for alternatives? I heard about some, but since I do not use it, haven’t searched for them.

I think that the software is specialized, but the hardware is not. They use some smart algorithms to distribute computation over huge number of workers.

Huh? Are you surprised that movies made in US are mostly about US? I am from Czech Republic and surprisingly, all our movies are about czech culture and none about Brazilian slave trade.