Human rights in the age of


Meaningful explainability is increasingly possible, and the AI world is largely behind it



Download 390,6 Kb.
Pdf ko'rish
bet44/45
Sana09.06.2023
Hajmi390,6 Kb.
#950086
1   ...   37   38   39   40   41   42   43   44   45
Bog'liq
AI-and-Human-Rights

Meaningful explainability is increasingly possible, and the AI world is largely behind it. 
Opponents of regulation argue that requiring explainability would mean systems would have to be 
substantially less complex, and therefore less accurate, ultimately stifling innovation. This argument misses 
the mark in a number of ways.
First, valuable levels of explainability are achievable. Although Facebook may not fully understand how its 
targeting advertising algorithm works, it knows enough to tell users what actions led to them being served a 
certain advertisement. This type of information matters, and it is easy to provide. Research suggests that it is 
126 See Joshua New, “How (and how not) fo fix AI,” Tech Crunch, July 26, 2018, https://techcrunch.com/2018/07/26/how-and-how-not-to-fix-ai/?mc_
cid=b4840b1adb&mc_eid=[UNIQID].
127 See, e.g., https://motherboard.vice.com/en_us/article/a34pp4/john-deere-tractor-hacking-big-data-surveillance (“The American Farm Bureau 
helped construct the “Privacy and Security Principles for Farm Data,” which addresses issues of data ownership, portability, use, and sharing. Companies 
like Deere and Monsanto were early signers, but questions remain about how much these principes protect in practice.”).
128 https://www.accessnow.org/cms/assets/uploads/2018/07/GDPR-User-Guide_digital.pdf


accessnow.org
37
HUMAN RIGHTS IN THE AGE OF ARTIFICIAL INTELLIGENCE
even possible for systems to measure how a given input affected the output. In a system used by universities 
to rank applicants, this could tell you, for example, that 20% of the ranking is due to GPA, 25% due to 
standardized testing, and 10% due to school ranking and other factors. This level of explanation would go a 
long way in addressing the “black box” challenge and identifying potential sources of bias.
Additionally, explainability is technically valuable. Developers need to be able to determine whether a system 
is solving the right problem. There are many examples of AI systems that “cheated” to arrive at the desired 
outcome. For example, researchers at the University of Washington created a deliberately bad algorithm 
that was supposed to classify images of husky dogs and wolves. The system correctly labeled the images, 
but rather than learning the difference between the appearance of huskies vs. wolves, the system detected 
the presence of snow because most of the images of wolves had snow in the background.
129
If AI systems in 
high stakes fields ultimately solve the wrong problem, the outcome could be life threatening.
Because explainability is necessary for the adoption of AI in certain fields, in some ways the quest for 
explainability is spurring AI innovation. For both ethical and technical reasons, academics and major AI companies 
alike are devoting significant effort toward explainability, and they are making serious progress. In 
August 2018, Google’s DeepMind published a study about an AI system it developed to identify eye disease 
in 3D ocular scans. When the system makes a diagnosis, it points out the portions of the scan it used so 
that physicians can see how it arrived at that diagnosis, as well as how confident it is in the diagnosis.
130
Breakthroughs such as this show how explainability may be driving innovation in AI.

Download 390,6 Kb.

Do'stlaringiz bilan baham:
1   ...   37   38   39   40   41   42   43   44   45




Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©hozir.org 2024
ma'muriyatiga murojaat qiling

kiriting | ro'yxatdan o'tish
    Bosh sahifa
юртда тантана
Боғда битган
Бугун юртда
Эшитганлар жилманглар
Эшитмадим деманглар
битган бодомлар
Yangiariq tumani
qitish marakazi
Raqamli texnologiyalar
ilishida muhokamadan
tasdiqqa tavsiya
tavsiya etilgan
iqtisodiyot kafedrasi
steiermarkischen landesregierung
asarlaringizni yuboring
o'zingizning asarlaringizni
Iltimos faqat
faqat o'zingizning
steierm rkischen
landesregierung fachabteilung
rkischen landesregierung
hamshira loyihasi
loyihasi mavsum
faolyatining oqibatlari
asosiy adabiyotlar
fakulteti ahborot
ahborot havfsizligi
havfsizligi kafedrasi
fanidan bo’yicha
fakulteti iqtisodiyot
boshqaruv fakulteti
chiqarishda boshqaruv
ishlab chiqarishda
iqtisodiyot fakultet
multiservis tarmoqlari
fanidan asosiy
Uzbek fanidan
mavzulari potok
asosidagi multiservis
'aliyyil a'ziym
billahil 'aliyyil
illaa billahil
quvvata illaa
falah' deganida
Kompyuter savodxonligi
bo’yicha mustaqil
'alal falah'
Hayya 'alal
'alas soloh
Hayya 'alas
mavsum boyicha


yuklab olish