HackerTrans
トップ新着トレンドコメント過去質問紹介求人

bolster

no profile record

投稿

Designing Trust into Artificial Intelligence Systems: A Guide for LLMss

jasebell.medium.com
2 ポイント·投稿者 bolster·3 年前·0 コメント

コメント

bolster
·11 か月前·議論
> Data used for the charts was researched and compiled by ChatGPT, I spot checked it and found it was accurate enough to for the narrative.

Last I checked, that wasn't how citations worked.

I only sampled one data point (2017 AMD EPYC Rome Clock rate) which was significantly off, because in 2017 it was the Naples chipset that was released, and, unless 2017 was desperate enough to clock from 2.2Ghz to 3Ghz on the regular (Boost up to 3.2Ghz), the 'research' was a fair bit off...

Doesn't undermine or contradict the authors (bots?) point, but a strange way to provide 'evidence' for an argument.
bolster
·2 年前·議論
"Structured token models don't work as well when you mess with the structural order of tokens" - News at 11
bolster
·2 年前·議論
So you're saying lack of sun increases health care expenditure? Excellent.

https://www.youtube.com/watch?v=IyjJbhuwGkU
bolster
·4 年前·議論
Yeah... I'll pass on paying $480/year to fuckup my own city; we've been doing it for free for years.
bolster
·4 年前·議論
I've always been on the fence with technical solutions to the 'Pipe wrench' problem but one thing that I don't see mentioned that often is that there are usually many secondary keyrings unlocked by the login password (ssh auth, saved passwords, session cookies maybe, etc);

I could see a solid usecase for a duress script that clears all these and requires 'standard' reauth, so that at least you're back to a 'defence in depth' style. Also, in the 'Pushover' example, I can't imagine many attackers waiting to plug the thing in before starting the ~pipe wrench~ credentials discussion.