Maxaa dhacay
BBC-da ayaa sheegtay in Jacob Coxon, oo hore u ahaa cilmi-baare Anthropic iyo OpenAI, uu ka digay in cilmi-baarayaasha AI ay ka baqaan xawliga horumarka iyo saameynta ay ku yeelan karto aadanaha. Wuxuu taageeray baaqyada horumarka gaabis ah, nidaaminta iyo kormeerka madax-bannaan, halka Anthropic ay difaacday shaqadeeda badbaadada. Warbixintu waxay ku dartay falcelino cusub oo ku saabsan baaqii jiray ee maamulaha Anthropic Dario Amodey ee ahaa taxadar weyn.
BBC ayaa sheegtay in Coxon, oo 27 jir ah, uu sheegay in cilmi-baarayaasha AI ay "si dhab ah uga baqayaan" xawaaraha horumarka ee hadda jira waxayna ku doodeen in bini'aadanku ay la kulmi karaan khatar aad u sareysa oo ah inay dhintaan mustaqbalka dhow haddii aan horumarku hoos u dhicin. Wuxuu soo dhaweeyay soo jeedinta Amodey ee xakameynta warshadaha oo dhan laakiin wuxuu sheegay in Shiinaha uu u baahan doono inuu ku lug yeesho si looga fogaado tartan caalami ah. Kuwani waa ra'yiga Coxon sida ay BBC-du soo warantay, ee maaha saadaal si madax-bannaan loo aasaasay.
Warbixintu waxay ku xidhaysaa faallooyinka Coxon qoraal uu qoray madaxa fulinta ee Anthropic Dario Amodey, oo la daabacay 12-kii sida laga soo xigtay isha, kaas oo ku dooday in horumarinta AI lafteedu aanay ahayn dhibaatada, laakiin shirkadaha iyo dawladuhu waxay u baahan yihiin wakhti si ay wax uga qabtaan khataraha halista ah. BBC ayaa sidoo kale ka warbixisa taageerada hoos u dhaca horumarka, qaanuunka ama la socodka madaxa banaan ee OpenAI maamulaha Sam Altman iyo maamulaha xAI Elon Musk.
Anthropic ayaa u sheegtay BBC News inay si joogto ah u qiratay faa'iidooyinka waaweyn iyo khataraha aan horay loo arag ee AI. Af-hayeenkeeda ayaa sheegay in shirkadu ay horumarinayso ilaalin xooggan, baadhista sida moodooyinka u shaqeeyaan, tijaabinta moodooyinka khataraha khaldan iyo daabacaadda natiijooyinka. Isha ma bixiso qiimayn madax banaan oo lagu ilaalinayo.
BBC-da ayaa ka warbixisay digniino dheeraad ah oo ka yimid saynisyahanka Anthropic ee Evan Hubinger iyo saynisyahanka kombuyuutarka Geoffrey Hinton, kuwaas oo mid walba uu bixiyay qiyaaso shakhsi ah oo ku saabsan suurtagalnimada dabar goynta bini'aadamka. Waxa kale oo ay ka warbixisay dhaleecayn ka timid maamulaha Hugging Face Clément Delangue iyo maamulaha kulliyadda Marc Warner, kuwaas oo sheegay in itimaalkan oo kale ay adag tahay in la dhiso. Isha kuma jiraan wax caddayn farsamo ah oo xaqiijinaya mid ka mid ah qiyaasahaas.
Maxay muhiim u tahay
Warbixintu waa arrin sababtoo ah waxay diiwaangelineysaa khilaafka dadweynaha ee sii ballaaranaya gudaha iyo hareeraha hoggaanka shirkadaha AI ee ku saabsan in horumarku uu ka dhaqso badan yahay shaqada amniga iyo kormeerka dawladda. Soo jeedinta wax ku oolka ah ee lagu sharraxay ayaa ah dejinta gaabis ah ee moodallo awood leh, kormeer madaxbannaan iyo isuduwidda caalamiga ah, oo ay ku jiraan Shiinaha. Si kastaba ha ahaatee, sheegashooyinka ugu daran waa saadaal ama qiyaaso shakhsiyeed, oo aan ahayn natiijooyin si madaxbanaan loo soo bandhigay. Isha ma cadeeyneyso in dabar go'aan ay dhawdahay, ma qiyaasi karto khatar la xaqiijiyay, ama ma tusin in dowlad ama shirkad kasta ay qaadatay xayiraado cusub.
Tani waa horumar badbaado AI oo macno leh sababtoo ah akoonkii guud ee shaqaalihii hore ayaa la soo bandhigay iyadoo ay weheliso baaqa hogaamiyaha shirkadda ee horumarinta iyo ilaalinta gaabis ah. Si wada jir ah, bayaanyadu waxay muujinayaan in welwelku aanu ku koobnayn dhaleeceynta dibadda, inkasta oo warbixintu aysan cabbirin sida ay u matalaan aragtidan oo ka mid ah cilmi-baarayaasha AI.
Saamaynta la taaban karo ayaa ah in doodaha badbaadada ay si isa soo taraysa u saamayn karaan maamulka siidaynta, sharciyaynta iyo isuduwidda caalamiga ah. Cawaaqibtaasi waxay ahaanaysaa kuwo mustaqbalka ah: isha ma soo sheegto sharci cusub, heshiis, hakad ku-meel-gaar ah ama dhacdo si madax-bannaan loo xaqiijiyay. Ma jiraan shuruudo gelitaanka badeecada ama qiimaynta oo khuseeya ama la diiwaangeliyay.
Farsamaynta Is-dhexgalka: Sida Dhabta Ay U Shaqeyso
U baadh tignoolajiyada hoose ee ka dambeeya horumarkan si isdhexgal leh.
Why can ethical evaluation not be reduced to one model score?
Maxaa la daawan doona xiga
U fiirso siyaasad la taaban karo ama ficillada shirkadda ee raacaya baaqyada xakamaynta: shuruudaha la socodka rasmiga ah, xadka geynta, heshiisyada caalamiga ah, ama isbeddelada hababka siidaynta moodeelka. Isha ma diiwaangeliso badeecad cusub, beddelka gelitaanka, qiimaha, natiijada imtixaanka badbaadada ama siyaasadda xidhitaanka. Sido kale si madax banaan uma xaqiijineyso itimaalka dabar goynta ee ay soo xiganayeen dadka la wareystay ama ma dejin waqtiga wareysiyada la soo sheegay marka loo eego daaqada hadda jirta ee 96-saac ah.
Haddi Anthropic ama shirkadaha kale ee AI ay u beddelaan baaqyada xakamaynta dhaqamo la xaqiijin karo, sida la socodka moodeelka madax-bannaan, heerarka qiimaynta ee la daabacay ama sii dayn gaabis ah.
Haddi ay dawladuhu kaga jawaabaan shuruuc gaar ah ama doodo caalami ah halkii ay ka ahaan lahaayeen bayaan guud oo ku saabsan tartanka AI iyo khatarta.
Haddii cilmi-baadhis farsamo oo madaxbannaan ay taageerto, yarayso ama caqabad ku noqoto qiyaasaha khatarta daran ee lagu sheegay warbixinta. Isha ma bixinayso caddayn ku filan si loo xalliyo khilaafkaas.
Qoraalka isha kuma jiraan taariikhda daabacaadda, sidaa darteed soo-dhoweynta wareysiyada hoose si madax-banaan looma xaqiijin karo walxaha la keenay.