Teburin Abubuwan Ciki
99
An Nazara Gudunmawar Taro
4
An Gano Tambayoyi Masu Muhimmanci
0
Ka'idojin Da'a Tare da Bayyanannen Ma'anar Amfanin Jama'a
1. Gabatarwa
Hankalin na'ura (AI) yana fuskantar ci gaba da karuwa mara misali a fannoni daban-daban, tare da karuwar damuwa game da da'a. Wannan takarda tana bincika manufar "AI don Amfanin Jama'a" ta hanyar bincike mai mahimmanci na tsarin da'a na yanzu kuma tana ba da shawarar gwajin tsaro na da'a a matsayin hanyar da za a bi don magance kalubalen da aka gano.
2. Ayyana Amfanin Jama'a a cikin Da'ar AI
2.1 Tushen Falsafa
Ra'ayin Amfanin Jama'a ya samo asali ne daga falsafar siyasa, yana nufin wuraren da ke amfanar dukkan membobin al'umma. A cikin mahallin AI, wannan yana fassara zuwa tsarin da aka ƙera don yiwa ƙungiyoyi hidima maimakon bukatun mutum ɗaya ko na kamfani.
2.2 Tsarin Da'a na AI na Yanzu
Binciken manyan jagororin da'a na AI ya nuna rashin daidaituwa a ma'anar Amfanin Jama'a, tare da yawancin tsarin suna jaddada guje wa cutarwa maimakon gudunmawa mai kyau ga jin daɗin al'umma.
3. Manyan Kalubale da Tambayoyi Masu Muhimmanci
3.1 Ayyana Matsala da Tsarinta
Mene ne "matsala" da ta cancanci shigarwar AI? Magungunan fasaha sau da yawa suna gaba da ayyana matsala yadda ya kamata, suna haifar da maganin matsala inda AI ke magance alamun cutar maimakon tushen tushen.
3.2 Wakilcin Masu Ruwa Da Tsaki
Wane ne ke ayyana matsalolin da ya kamata AI ta magance? Rashin daidaiton iko a cikin ayyana matsala na iya haifar da mafita waɗanda ke hidimar manyan bukatun yayin da suke ware ƙungiyoyi masu rauni.
3.3 Ilimi da Ilimin Fahimta
Wadanne tsarin ilimi ne aka ba su fifiko a cikin haɓakar AI? Ilimin fasaha sau da yawa yana rinjaye akan tsarin ilimin gida, mahalli, da na asali.
3.4 Sakamakon Da Ba a Yi Nufi Ba
Menene sakamakon sakandare na tsarin AI? Ko da shigarwar AI mai kyakkyawar niyya na iya haifar da illa ta hanyar hadaddun tsarin tsarin.
4. Hanyar Bincike da Nazarin Gwaji
4.1 Ƙirar Binciken Bincike
Marubucin ya gudanar da bincike mai zurfi na gudunmawar 99 ga tarukan AI don Kyautata Al'umma, yana nazarin yadda waɗannan ayyukan suka magance tambayoyi huɗu masu mahimmanci.
4.2 Sakamako da Binciken
Binciken ya nuna manyan gibi a cikin la'akari da da'a: Kashi 78% na takardun sun kasa magance wakilcin masu ruwa da tsaki, yayin da kashi 85% ba su tattauna yiwuwar sakamakon da ba a yi niyya ba. Kashi 12% kawai ne suka bayar da bayyanannun ma'anar abin da ya zama "nagari" a cikin mahallinsu na musamman.
Hoto na 1: La'akari da Da'a a cikin Binciken AI don Kyautata Al'umma
Taswira mai nuna kashi na takardun taro 99 da suka magance kowace tambaya mai mahimmanci guda huɗu: Ayyana Matsala (45%), Wakilcin Masu Ruwa Da Tsaki (22%), Tsarin Ilimi (18%), Sakamakon Da Ba a Yi Nufi Ba (15%).
5. Tsarin Gwajin Tsaro na Da'a
5.1 Tushen Ra'ayi
Dauke daga gwajin tsaro na tsaro na intanet, gwajin tsaro na da'a ya ƙunshi ƙoƙari na tsari don gano raunin da'a a cikin tsarin AI kafin a tura su.
5.2 Hanyar Aiwararwa
Tsarin ya haɗa da atisayen ƙungiyar ja, tunani na adawa, da kuma tsarin tambayar zato a ko'ina cikin tsarin haɓakar AI.
6. Aiwarar da Fasaha
6.1 Tsarin Lissafi
Za a iya ƙirƙira tasirin da'a na tsarin AI kamar haka: $E_{impact} = \sum_{i=1}^{n} w_i \cdot \phi(s_i, c_i)$ inda $s_i$ ke wakiltar ƙungiyoyin masu ruwa da tsaki, $c_i$ ke wakiltar nau'ikan sakamako, $w_i$ ma'auni ne na da'a, kuma $\phi$ aikin tantance tasiri ne.
6.2 Aiwarar Algorithm
class EthicsPenTester:
def __init__(self, ai_system, stakeholder_groups):
self.system = ai_system
self.stakeholders = stakeholder_groups
def test_problem_definition(self):
"""Tambaya ta 1: Menene matsalar?"""
return self._assess_problem_framing()
def test_stakeholder_representation(self):
"""Tambaya ta 2: Wane ne ke ayyana matsalar?"""
return self._analyze_power_dynamics()
def test_knowledge_systems(self):
"""Tambaya ta 3: Wane irin ilimi ne aka ba shi fifiko?"""
return self._evaluate_epistemic_justice()
def test_consequences(self):
"""Tambaya ta 4: Menene illolin gefe?"""
return self._simulate_system_dynamics()7. Aikace-aikace da Hanyoyin Gaba
Tsarin gwajin tsaro na da'a yana nuna alƙawari don aikace-aikace a cikin AI na kiwon lafiya, algorithm na adalci na laifuka, da fasahar ilimi. Aikin gaba ya kamata ya mayar da hankali kan haɓaka ƙa'idodin gwaji na daidaitacce da haɗa hanyar tare da hanyoyin haɓakar AI na yanzu kamar Agile da DevOps.
Mahimman Bayanai
- Tsarin da'a na AI na yanzu ba su da ayyukan ma'anar Amfanin Jama'a
- Maganin fasaha sau da yawa yana gaba da ayyana matsala yadda ya kamata
- Wakilcin masu ruwa da tsaki ya kasance babban gibi a cikin haɓakar AI
- Gwajin tsaro na da'a yana ba da hanyar aiki don tantance da'a
Bincike Mai Muhimmanci: Bayan Magungunan Fasaha zuwa AI mai Da'a
Aikin Berendt yana wakiltar ci gaba mai mahimmanci a cikin motsa da'ar AI daga ƙa'idodi zuwa hanyoyin aiki. Tsarin gwajin tsaro na da'a da aka ba da shawara yana magance babban gibi da masu bincike a Cibiyar AI Now suka gano, waɗanda suka rubuta yadda ake kula da la'akari da da'a a matsayin abin da ake yi a bayan fage maimakon abubuwan da suka haɗa da tsarin tsarin. Wannan hanyar ta yi daidai da sabbin ayyuka masu kyau a cikin haɓakar AI mai alhaki, kama da jagororin PAIR na Google (Mutane + Binciken AI) waɗanda ke jaddada hanyoyin ƙira masu mayar da hankali ga mutum.
Tsarin tambayoyi huɗu masu mahimmanci yana ba da tsarin tsari don magance abin da masanin falsafa Shannon Vallor ya kira "kyawawan dabi'un fasaha" - halayen tunani da aiki da ake buƙata don kewaya rikitattun abubuwan da'a na AI. Wannan hanyar tana nuna alƙawari na musamman idan aka kwatanta da hanyoyin fasaha kawai don amincin AI, irin waɗanda aka ba da shawara a cikin Ƙa'idodin AI na Asilomar. Yayin da amincin fasaha ya mayar da hankali kan hana gazawa mai muni, gwajin tsaro na da'a yana magance ƙarin abubuwa masu sauƙi amma masu mahimmanci daidai da daidaiton ƙima da tasirin zamantakewa.
Idan aka kwatanta da tsarin tantancewar da'a na yanzu kamar Jerin Tantancewar Aminci na AI na EU (ALTAI), Hanyar Berendt tana ba da ƙayyadaddun bayanai a cikin magance yanayin iko da wakilcin masu ruwa da tsaki. Binciken binciken da aka gano na manyan gibi a cikin binciken AI na yanzu don Kyautata Al'umma yana juyar da damuwar da masu bincike a Cibiyar Bincike ta Data & Society suka taso game da rarrabuwar kawuna tsakanin iyawar fasaha da fahimtar zamantakewa a cikin haɓakar AI.
Tsarin lissafi na tantance tasirin da'a ya ginu akan aikin da ya gabata a cikin bincike na yanke shawara mai ma'ana amma ya daidaita shi musamman don tsarin AI. Wannan yana wakiltar wani muhimmin mataki zuwa ga tantancewar da'a mai ƙima, kodayake kalubale sun rage a cikin tantance madaidaitan ma'auni da ayyukan tasiri. Aikin gaba zai iya haɗa wannan hanyar tare da hanyoyin yau da kullun daga ka'idar zaɓin zamantakewa na lissafi don ƙirƙirar kayan aikin tantancewar da'a mai ƙarfi.
8. Nassoshi
- Berendt, B. (2018). AI don Amfanin Jama'a?! Kuskure, kalubale, da Gwajin Tsaro na Da'a. arXiv:1810.12847v2
- Vallor, S. (2016). Fasaha da Nagarta: Jagorar Falsafa zuwa Nan Gobi da Ake So. Jami'ar Oxford Press.
- Cibiyar AI Now. (2018). Rahoton AI Now na 2018. Jami'ar New York.
- Hukumar Tarayyar Turai. (2019). Jagororin Da'a don Amincewar AI.
- Google PAIR. (2018). Jagorar Mutane + AI.
- Ka'idodin AI na Asilomar. (2017). Cibiyar Rayuwa ta Gobe.
- Cibiyar Bincike ta Data & Society. (2018). Lissafin Lissafi: Jagora.