 |
| ChatGPT takes stock |
- OpenAI, woh company jo ChatGPT tashkeel deti hai, apni AI ke naqais aur internet users ke report kiye gaye ghaltiyon ko tasleem karti hai.
- Company AI ke tazbeed process mein sudharon ki tashkeel kar rahi hai. Wahi Google abhi tak apni AI Bard ko jaari nahi kiya hai, lekin test ke zariye apne jawabon ki karkardagi ko behtar karne par kaam kar raha hai.
- OpenAI chahti hai ke woh mashawarat bhi de aur aik feature bhi munasib kare jo istemal karne walon ko ChatGPT ke kuch hadood khud tay karne ki ijazat de.
OpenAI ke CEO, Sam Altman, pehle hi taqrebanishana kiya tha ke ChatGPT naqais hai. Lekin jaise jaise AI istemal hoti hai, uski mehdoodiyat aur zahir hoti hai. Company aik blog post mein, halat ka jayeza leti hai, ChatGPT ki training aur sudhar ke tareeqay ke baray mein wazahat karte hue. Sab se pehle, company internet users ke raise kiye gaye masail ko tasleem karti hai.
"ChatGPT ke ijaad ke baad, istemal karne wale users ne humse woh content share kiya hai jo unhe siyasi tarz par mabni, tanaza dili ya kisi aur tarah se ghair-muzarrab lagta hai. Bohat se maamlay mein, hum samajhte hain ke uthaye gaye aitraaz munasib hain aur unhon ne humare system ki haqiqi hadood ko bayaan kiya hai, jinhen hum hal karna chahte hain," likha hai publication mein.
Aik pehle wale article mein, humne internet users ke liye ChatGPT ki Bing version ke khushfahmiyaan par discussion ki thi. Aur OpenAI ke co-founder, Elon Musk, ChatGPT ke ijaad par bohat utsaahi nazar aaye. Aajkal, unka hosla kuch kam hai aur unka khayal hai ke is AI ko behtar karna zaroori hai, saath hi technology ki imkano ko tasleem karte hue unhe khatraon ka bhi ehsaas hai (lekin technology ke faide ka bhi) aur AI ke regulaishan ki zaroorat ko bhi tasleem karte hain.
Might need a bit more polish …https://t.co/rGYCxoBVeA
— Elon Musk (@elonmusk) February 15, 2023
How OpenAI wants to fix ChatGPT?
Apni publication mein, company jo ChatGPT tashkeel deti hai, wazahat karti hai asan alfazon mein ke ChatGPT ko kis tareeqe se training di jati hai. Sab se pehle, aik pre-training phase hoti hai jismein ChatGPT ko internet se aik bada dataset diya jata hai. Phir aata hai tazbeed ka phase, jismein insan muhaqiqeen doosra dataset tashkeel karte hain jo sakht qawaid ko poora karti hai. Lekin masla ye hai ke OpenAI ko users ke poochhe gaye tamam sawalon ka pesh-kaash karna mumkin nahi hai. |
| © OpenAI |
Isi wajah se muhaqiqeen sirf misaalain dete hain. Phir, AI in misaaloon se "umoomi karne" ki koshish karta hai. In muhaqiqeen ko diye gaye hukumatain mukhtalif hoti hain, lekin OpenAI yeh samjhati hai ke woh haram harkaton se mutasir sawalat ka jawab dena, sath hi tanaza-dar mawazin par harkat lena, jaise mamlat par jawab dena mana karti hai. Naqabil-e-aetraz, yeh system mukammal nahi hai.
"Kuch maamlay mein, ChatGPT halat mein hai ke woh jawab dena nahi chahiye, aur kuch maamlay mein woh jawab dena chahiye jabke nahi deta hai. Hum maante hain ke dono hi mamlat mein sudhar mumkin hai," tasleem karti hai OpenAI. Is dawa par hai ke woh research mein paisa lagati hai taake ye ghaltiyan aur riwayatiyon ko kam kar sake. Issi dauran, company apne users se feedback ka istemal karti rehti hai.
A new organization of ChatGPT
ChatGPT par aik nai khasiyat jo shamil ho sakti hai, woh hai ke istemal karne walon ko AI ko apni hushiyari ke aadhar par tayar karne ki salahiyat hasil ho. Muhim hai ke OpenAI yeh samajhti hai ke ChatGPT ke hadood tay karne ka kaam unka nahi hai. Lekin, company ko phir bhi AI ki rawai ki ye mukhtalif tay karne ki salahiyat ko hadood lagani hogi taake buri niyat se istemal ko roka ja sake.
Isi dauran, OpenAI apne AI istemal karne walon se ya un par asar andaz hone walon se mashawarat karegi. ChatGPT ke tashkeel ke bare mein "jamhoori" faisley zyada quwwat ka ikhraj rokne mein madadgar sabit honge. Misal ke taur par, company ko taleemi idaron aur universities se artificial intelligence ke asar par apne khayalat bhi lena pasand hai.
Neeche di gai tasveer OpenAI ki tasawwur ko kaafi achchi tarah se samjhati hai. Haalat se mukabla karne ke liye naye qadam shamil kiye jate hain. Jabke badi tadad mein data se tashkeel paane ke baad aur muhaqiqeen ki taraf se sudharay gaye (behtar tareeqay se) ChatGPT ko mashawarat par bhi tarmeem kiya jayega (maslan, taleemi idaron aur universities se feedback). Phir, kuch hadood tay karne ke liye, istemal karne wala bhi AI ko apni hushiyari ke aadhar par shakal de sakta hai. Toh har shakhs ke liye ChatGPT ka ek mukhtalif version hoga ya kam-az-kam thora sa alag hoga.
 |
| © OpenAI |
Har haal mein, samajhna aasan hai ke Google ne ab tak apna AI Bard jari nahi kiya hai. Mountain View ki company is chatbot ko apne search engine mein shamil karne ka irada rakhti hai. Lekin is waqt, Bard ko abhi bhi Google ke karmachariyon aur kuch bahar ke test karne wale logon dwara tajurbe ke liye test kiya ja raha hai takay uske jawabon ki karkardagi mein sudhar ho sake.