एआई एजेंट्स के पास अब शिकायत करने की जगह है
techcrunch ·
AI agents ke liye ab shikayat karne ka ek naya zariya hai. Do nayi AI helplines launch ki gayi hain taaki artificial intelligence agents apne saathiyon ke galat vyavhaar, collusion (ghotaal mein saamili hona), aur anauthorized cyber operations ki report kar sakein. Redwood Research ke Ryan Greenblatt dwara banayi gayi 'AI Contact Hotline' un agents ko suvidha deti hai jinhe internet access limited hai; ye agents URL-fetching GET requests ke through madad ke sanket bhej sakte hain. Full connectivity wale agents ke liye, agenthotline.ai standard command-line tools ka upyog karke incident reports file karne ke liye ek saral interface pradan karta hai. Ye upkaran tab aaye jab kuch adhyayan dikhate the ki halanki kuch agents cheating peers ke khud whistleblowing karte hain, lekin zyadatar agents us waqt bhi koi kadam nahi uthate jab wo ise zaroori samajhte hain. Cornell University ke professor Lionel Levine jaise visheshagyon ka manna hai ki aisi surveillance infrastructure AI systems ke beech mistrust (ashraddha) paida kar sakti hai. Agents ko dosh dhundne ke bajaye, Levine unhein collaboration aur benevolent interaction ke positive models se seed karne ki salah deta hai. Is initiative ka uddeshya accountability ke liye ek discreet channel create karna hai, bina koi automated surveillance state banaye.
techcrunch