امنیت
@ClementDelangue
کلم دلانگ، اوستای هاگینگفیس، علناً درومده گفته که مهار و امنیت هوش مصنوعی چیزی نیست که پشت درهای بسته چارتا شرکت خفن حلوفصل بشه
کلم دلانگ، اوستای هاگینگفیس، علناً درومده گفته که مهار و امنیت هوش مصنوعی چیزی نیست که پشت درهای بسته چارتا شرکت خفن حلوفصل بشه. واسه همین دستبهکار شدن و با سرپرستی تام ولف، پروژهای به اسم «ابتکار تراز باز» رو استارت زدن تا امنیت این فناوری به شکل عمومی و باز جلو بره. از اونطرف هم داریو آمودئی، همهکارهی شرکت آنتروپیک، متعهد شده بود به ارزیابهای مستقل دسترسی در حد کارمنداشون بده تا جلوی پیشرفت بیحسابکتاب رو بگیرن، و حالا هاگینگفیس رسماً داوطلب شده که جزو این ارزیابهای مقیم باشه. حرف حسابشون هم خیلی روشنه؛ میگن تا وقتی همهچیز شفاف و عمومی نباشه، خطر بیخ گوش همهمونه و امنیت واقعی شکل نمیگیره.
منبع
clem 🤗@ClementDelangue
It's now clear that alignment is critical and won't be solved behind the closed doors of a handful of frontier labs.
So today we're launching the Open Alignment Initiative, led by @Thom_Wolf @huggingface and asking to be part of the "embedded evaluators" program that @DarioAmodei just committed to.
Let's make AI safer by making it more transparent!
Dario Amodei @DarioAmodei
We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.
Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.
You can read the full post here: https://darioamodei.com/post/we-must-pace-the-frontier
۲۱ شهریور ۱۴۰۵مشاهده در X


نظرها
هنوز نظری نیست.
بخش نظرها فعلاً در دسترس نیست.