When Algorithms Discriminate
Algoritmlar kamsitganda
- L Lecturer
Tinglashdan oldin: kalit so‘zlar
Har bir so‘zni bosing, eshiting va ovoz chiqarib takrorlang. Bu so‘zlar audioda uchraydi.
💡 Kontekst haqida o‘ylang: Ma’ruza: sun’iy intellekt tizimlari qanday qilib insoniy noxolislikni takrorlashi mumkin.
Tayyor bo‘lsangiz, tugmani bosing
Matnga qaramasdan to‘liq tinglang. Kim gapiryapti, qayerda va nima haqida? Har bir so‘zni emas, umumiy ma’noni tushunishga harakat qiling. Kerak bo‘lsa, ikki marta tinglang.
Tushundingizmi?
Javobni tanlang. Ishonchingiz komil bo‘lmasa, pastdagi pleyer orqali qayta tinglang.
-
1 Why did the hiring algorithm discriminate?
-
2 What happened to the tool?
-
3 What do many researchers propose?
Matn bilan tinglash
Istalgan gapni bosing — o‘sha joydan ijro etiladi. So‘zlar eshitilgan sari yonib boradi.
-
We tend to assume that computers are neutral. But an algorithm trained on historical data will learn the patterns in that data, including its prejudices.
Biz kompyuterlarni betaraf deb hisoblashga moyilmiz. Lekin tarixiy ma’lumotlarda o‘qitilgan algoritm o‘sha ma’lumotlardagi qonuniyatlarni, jumladan xurofotlarni ham o‘rganadi.
-
One large company built a system to screen job applications, trained on ten years of CVs from people it had hired.
Bir yirik kompaniya o‘zi ishga olgan odamlarning o‘n yillik rezyumelari asosida ish arizalarini saralash tizimini yaratdi.
-
Because most past hires were men, the system learned to downgrade CVs that included the word “women’s”, as in “women’s chess club”.
O‘tmishda ishga olinganlarning ko‘pchiligi erkaklar bo‘lgani uchun tizim “women’s” so‘zi bor rezyumelarni, masalan “ayollar shaxmat klubi”, pastroq baholashni o‘rgandi.
-
Nobody programmed it to discriminate. It simply reproduced past decisions very efficiently.
Hech kim uni kamsitishga dasturlamagan. U shunchaki o‘tmishdagi qarorlarni juda samarali takrorladi.
-
The company eventually abandoned the tool, but many similar systems are used in lending, policing and healthcare with far less scrutiny.
Kompaniya oxir-oqibat vositadan voz kechdi, lekin ko‘plab shunga o‘xshash tizimlar kredit berish, politsiya va sog‘liqni saqlashda ancha kam nazorat bilan ishlatilmoqda.
-
That’s why many researchers argue that high-stakes algorithms should be independently audited before they’re deployed.
Shuning uchun ko‘plab tadqiqotchilar yuqori xavfli algoritmlar joriy etilishidan oldin mustaqil tekshiruvdan o‘tishi kerak, deb ta’kidlaydi.
Diktant
Gapni tinglang va eshitganingizni aynan yozing. Katta-kichik harf va tinish belgilari hisobga olinmaydi. Raqamlarni so‘z bilan yozing.
Brauzeringiz ovozli o‘qishni qo‘llab-quvvatlamaydi. Chrome, Edge yoki Safari’ning so‘nggi versiyasidan foydalaning.