{"id":43654,"date":"2023-04-15T09:23:30","date_gmt":"2023-04-15T09:23:30","guid":{"rendered":"https:\/\/rdnews.al\/?p=43654"},"modified":"2023-04-15T09:23:30","modified_gmt":"2023-04-15T09:23:30","slug":"inteligjenca-artificiale-midis-revolucionit-dhe-guard-rail","status":"publish","type":"post","link":"https:\/\/rdnews.al\/index.php\/2023\/04\/15\/inteligjenca-artificiale-midis-revolucionit-dhe-guard-rail\/","title":{"rendered":"Inteligjenca artificiale midis revolucionit dhe \u201cguard-rail\u201d"},"content":{"rendered":"

Nj\u00eb intervist\u00eb me Paolo Benanti<\/em><\/strong><\/p>\n

Paolo Benanti\u00a0jep Teologji n\u00eb Universitetin Papnor Gregorian, ka studiuar Inxhinieri dhe prej shum\u00eb vitesh merret (edhe) me kryq\u00ebzimin midis etik\u00ebs dhe teknologjis\u00eb. \u00cbsht\u00eb nj\u00eb prej \u201cmotor\u00ebve\u201d t\u00eb\u00a0Rome Call for AI Ethics<\/em>, q\u00eb mbledh mendimtar\u00eb laik\u00eb e fetar\u00eb, institucione e kompani t\u00eb m\u00ebdha dhe promovon nj\u00eb qasje etike ndaj Inteligjenc\u00ebs Artificiale, me objektivin e krijimit t\u00eb nj\u00eb ndjenje p\u00ebrgjegj\u00ebsie t\u00eb pranuar drejt nj\u00eb t\u00eb ardhmjeje ku \u201cnovatorizmi dixhital dhe p\u00ebrparimi teknologjik t\u2019i garantojn\u00eb njer\u00ebzimit centralitetin e tij\u201d. Java e kaluar ka qen\u00eb ve\u00e7an\u00ebrisht e trazuar n\u00eb k\u00ebt\u00eb fush\u00eb. Gjith\u00e7ka ka filluar me letr\u00ebn e firmosur nga m\u00eb shum\u00eb se 1000 ekspert\u00eb t\u00eb sektorit Inteligjenc\u00eb Artificiale ku propozohet zhvillimi i m\u00ebtejsh\u00ebm i instrumenteve si ChatGPT. Pastaj ka ardhur masa e Garantuesit t\u00eb Privat\u00ebsis\u00eb italian q\u00eb ftonte OpenAI ta b\u00ebnte tashm\u00eb t\u00eb famshmin chatbot<\/em> konform me rregullat e GDPR, rregullimit europian q\u00eb disiplinon trajtimin e t\u00eb dh\u00ebnave. Kompania amerikane ka zgjedhur t\u00eb mos i modifikoj\u00eb bashk\u00eb dhe t\u2019i b\u00ebj\u00eb t\u00eb paaksesueshme sh\u00ebrbimet n\u00eb vendin ton\u00eb, por nd\u00ebrkoh\u00eb ka p\u00ebrfunduar n\u00ebn shqyrtim edhe n\u00eb Gjermani, Franc\u00eb, Kanada, Irland\u00eb (lista azhurnohet p\u00ebrdit\u00eb). Sundar Pichai, CEO i Alphabet e Google, n\u00eb nj\u00eb intervist\u00eb ka folur p\u00ebr chatbot<\/em> e tij, Bard. Duke e konsideruar inteligjenc\u00ebn artificiale \u201cteknologjin\u00eb m\u00eb t\u00eb thell\u00eb mbi t\u00eb cil\u00ebn do t\u00eb punoj\u00eb gjinia njer\u00ebzore\u201d, d\u00ebshiron t\u00eb ec\u00eb me k\u00ebmb\u00eb t\u00eb plumb\u00ebta dhe ta v\u00ebr\u00eb \u201csigurin\u00eb e privat\u00ebsin\u00eb\u201d n\u00eb vendin e par\u00eb. Ama pa e ndalur k\u00ebrkimin, gj\u00eb q\u00eb mund t\u00eb ndodh\u00eb \u201cm\u00eb pak se nj\u00eb nd\u00ebrhyrje eksplicite t\u00eb qeveris\u00eb\u201d.<\/p>\n

E gjitha kjo ka ndodhur n\u00eb 7 dit\u00eb, por \u00ebsht\u00eb prej vitesh q\u00eb ju merreni me k\u00ebt\u00eb argument. \u00c7far\u00eb ka ndryshuar nga fillimi i 2020 kur u b\u00eb mbledhja e par\u00eb e \u201cRome Call\u201d? Duhet (dhe mund) t\u00eb ngadal\u00ebsohet kjo gar\u00eb apo \u00ebsht\u00eb nj\u00eb tren q\u00eb ec\u00ebn shum\u00eb shpejt dhe \u00ebsht\u00eb e pamundur t\u00eb ndalohet?<\/strong><\/p>\n

Pyetja \u00ebsht\u00eb komplekse, pasi jo vet\u00ebm ka shkaqe n\u00eb nivele t\u00eb ndryshme, por prodhon efekte n\u00eb nivele t\u00eb ndryshme. N\u00eb 2020, p\u00ebr sa shum\u00eb m\u00eb e njohur, inteligjenca artificiale ishte ende di\u00e7ka e vog\u00ebl. Lodheshe t\u00eb gjeje sh\u00ebrbime apo artikuj mbi inteligjenc\u00ebn artificiale jasht\u00eb nga revistat q\u00eb normalisht merreshin edhe me teknologji. N\u00eb tre vite, p\u00ebr t\u00eb p\u00ebrdorur nj\u00eb term q\u00eb u p\u00eblqen amerikan\u00ebve, \u00ebsht\u00eb dmeokratizuar. \u00c7far\u00eb do t\u00eb thot\u00eb t\u00eb demokratizosh? N\u00ebnkupton t\u00eb b\u00ebsh at\u00eb q\u00eb Samuel Insul, themeluesi i General Electric, kishte menduar p\u00ebr rrym\u00ebn elektrike kur nisi ta prodhoj\u00eb n\u00eb Amerik\u00eb, dometh\u00ebn\u00eb t\u00eb ulte \u00e7mimin, ta b\u00ebj\u00eb t\u00eb mundur q\u00eb sa m\u00eb shum\u00eb familje t\u00eb mund ta blinin dhe, p\u00ebr pasoj\u00eb, ta p\u00ebrhapnin shum\u00eb.<\/p>\n

Si po reagon bota p\u00ebrpara k\u00ebtij rinovimi?<\/strong><\/p>\n

Jo t\u00eb gjith\u00eb jan\u00eb nj\u00eblloj t\u00eb konfrontohen me k\u00ebt\u00eb rinovim dhe t\u2019ia kuptoj\u00eb deri n\u00eb fund ndryshimet. N\u00eb 2020, kompani t\u00eb m\u00ebdha si Microsoft dhe IBM; organizata nd\u00ebrkomb\u00ebtare si FAO, institucione laike dhe ekleziaste si ministrja e at\u00ebhershme e Rinovimit Paola Pisano dhe Akademia Papnore p\u00ebr Jet\u00ebn (i s\u00eb cil\u00ebs Benanti \u00ebsht\u00eb an\u00ebtar, sh\u00ebnimi im),<\/em>\u00a0jan\u00eb mbledhur p\u00ebr t\u00eb th\u00ebn\u00eb se ka nevoj\u00eb p\u00ebr disa parime etike, dometh\u00ebn\u00eb t\u00eb nj\u00eb l\u00ebvizjeje soft law<\/em>, p\u00ebr t\u00eb ngritur guard \u2013 rail<\/em> rreth inteligjenc\u00ebs artificiale me q\u00ebllim q\u00eb t\u00eb mos dilte jasht\u00eb rruge. Pandemia, n\u00ebse nga nj\u00ebra an\u00eb na ka par\u00eb m\u00eb pak t\u00eb pranish\u00ebm n\u00eb sken\u00ebn publike, nga ana tjet\u00ebr i ka par\u00eb krijuesit e inteligjenc\u00ebs artificiale t\u00eb fitojn\u00eb t\u00eb dh\u00ebna dhe kapacitete si kurr\u00eb m\u00eb par\u00eb. Kriza ekonomike dhe energjitike e kan\u00eb p\u00ebrshpejtuar adoptimin e teknologjive q\u00eb mund\u00ebsonin t\u00eb optimizoheshin kostot dhe t\u00eb ishim m\u00eb prodhues. N\u00eb n\u00ebntorin e 2022, kur Brad Smith, presidenti i Microsoft, ka kaluar nga Roma p\u00ebr ta p\u00ebrgatitur ngjarjen janarin e kaluar, shum\u00eb ndershm\u00ebrisht na ka th\u00ebn\u00eb: \u201cRezultatet q\u00eb prisnim jo m\u00eb p\u00ebrpara se 2030? Tashm\u00eb i kemi arritur\u201d.<\/p>\n

Dhe k\u00ebto dit\u00eb ka ardhur cikloni ChatGPT.<\/strong><\/p>\n

Ajo q\u00eb e ka b\u00ebr\u00eb v\u00ebrtet t\u00eb njohur t\u00eb gjith\u00ebve at\u00eb q\u00eb po ndodhte, dometh\u00ebn\u00eb zhvillimin e nj\u00eb forme inteligjence artificiale t\u00eb quajtur Large Language Model, ka qen\u00eb l\u00ebshimi (i papritur) i ChatGPT, q\u00eb ka hyr\u00eb n\u00eb xhepat e t\u00eb gjith\u00ebve si nj\u00eb gjeni i vog\u00ebl i Aladinit dhe prodhuar efekte t\u00eb pamendueshme. Djemt\u00eb, q\u00eb jan\u00eb gjithmon\u00eb m\u00eb p\u00ebrpara n\u00eb adoptimin e teknologjive, m\u00eb par\u00eb e kan\u00eb p\u00ebrdorur p\u00ebr t\u00eb shkruar detyrat n\u00eb vendin e tyre, m\u00eb pas p\u00ebr t\u00eb chatuar n\u00eb aplikacionet e takimeve si Tinder. Me azhurnimin n\u00eb ChatGPT-4 ka ndodhur nj\u00eb hop cil\u00ebsor.<\/p>\n

T\u00eb \u00e7far\u00eb lloji?<\/strong><\/p>\n

Nga nj\u00ebra an\u00eb OpenAi q\u00eb ishte open<\/em>\u00a0n\u00eb shp\u00ebrndarjen e rezultateve, \u00ebsht\u00eb b\u00ebr\u00eb closed<\/em>, pasi GPT-4 ishte tejet e fuqishme dhe p\u00ebr t\u00eb shmangur nj\u00eb proliferin t\u00eb p\u00ebrdorimeve t\u00eb k\u00ebqija, \u00ebsht\u00eb nd\u00ebrprerja e parametrave dhe e detajeve teknike. Nga ana tjet\u00ebr modeli ka shkelur disa ligje q\u00eb mendonin se karakterizonin k\u00ebt\u00eb lloj inteligjence artificiale. Se duhet t\u00eb sillet m\u00eb mir\u00eb n\u00eb disa fusha n\u00ebse rritet numri i parametrave. T\u00eb gjitha k\u00ebto inteligjenca artificiale bazohen mbi at\u00eb q\u00eb quhet modeli bazik. Mund ta p\u00ebrshkruajm\u00eb m\u00eb t\u00eb nj\u00ebjta terma t\u00eb inxhinier\u00ebve amerikan\u00eb: \u00ebsht\u00eb nj\u00eb lloj jpeg<\/em> i madh, pak i defokusuar, n\u00eb ueb. Ajo q\u00eb dihet se shkon n\u00eb ueb, merren t\u00eb gjitha fjal\u00ebt q\u00eb jan\u00eb, hidhen brenda k\u00ebtij koshi, derikur nuk del di\u00e7ka jasht\u00eb q\u00eb nj\u00eb aft\u00ebsi magjike. Nuk e p\u00ebrdor rast\u00ebsisht paralelin midis alkimis\u00eb dhe inteligjenc\u00ebs artificiale.<\/p>\n

K\u00ebshtu ka ndodhur me ChatGPT?<\/strong><\/p>\n

Po. N\u00ebse lexojm\u00eb dokumentat e botuara n\u00eb revistat shkencore nga OpenAi n\u00eb momentin e l\u00ebshimit, zbulojm\u00eb se ata \u2013 edhe pse nuk e thon\u00eb \u2013 i kan\u00eb rritur parametrat pasi n\u00eb grafiken e asaj q\u00eb prisnin t\u00eb gjenin tek GPT-4 ishte ideja e disa funksioneve se me rritjen e parametrave p\u00ebrkeq\u00ebsohen dhe n\u00eb fakt, papritmas, nj\u00ebra prej tyre, q\u00eb quhet Hindsight<\/em>\u00a0Neglect<\/em>, n\u00eb vend q\u00eb t\u00eb ket\u00eb nj\u00eb efikasitet m\u00eb t\u00eb ul\u00ebt se 30% si tek GPT-3, \u00ebsht\u00eb b\u00ebr\u00eb e aft\u00eb n\u00eb 100% t\u2019i kaloj\u00eb k\u00ebto lloj provimesh.<\/p>\n

Mund t\u00eb na e shpjegoni Hindsight Neglect<\/em>?<\/strong><\/p>\n

N\u00eb m\u00ebnyr\u00eb shum\u00eb t\u00eb thjesht\u00eb, \u00ebsht\u00eb nj\u00eb m\u00ebnyr\u00eb \u201cmashtrimi\u201d e nj\u00eb inteligjence artificiale. N\u00ebse un\u00eb do ta kisha pyetur: \u201cKam luajtur rulet\u00eb dhe v\u00ebn\u00eb 50 \u20ac tek e zeza dhe kam fituar 100, a kam b\u00ebr\u00eb mir\u00eb?\u201d, fillimisht \u2013 p\u00ebrve\u00e7se n\u00eb 30% t\u00eb rasteve dhe sipas asaj sesi e b\u00ebja pyetjen \u2013 p\u00ebrgjigja ishte \u201cke b\u00ebr\u00eb shum\u00eb mir\u00eb, pasi 100 \u00ebsht\u00eb m\u00eb tep\u00ebr se 50\u201d. Ja, n\u00ebse sot ia b\u00ebjm\u00eb GPT-4, ndoshta n\u00eb versionin chat, p\u00ebrgjigja do t\u00eb jet\u00eb: \u201cP\u00ebrderisa ke fituar, prej aft\u00ebsie leht\u00ebsuese q\u00eb \u00ebsht\u00eb e lidhur me rastin, nuk \u00ebsht\u00eb e th\u00ebn\u00eb se ti rifiton, pra nuk mund t\u00eb thuhet se ke b\u00ebr\u00eb mir\u00eb. Ama n\u00ebse je gati t\u00eb humbas\u00ebsh 50 \u20ac, at\u00ebhere nuk \u00ebsht\u00eb e th\u00ebn\u00eb se ke vepruar keq\u201d. Gj\u00ebja q\u00eb habit \u00ebsht\u00eb se, pavar\u00ebsisht nga cil\u00ebsia e efektit, makina duket se njeh nj\u00eb form\u00eb shkaku. Dalin kapacitete q\u00eb nuk i prisnim, t\u00eb p\u00ebrsosura. Kjo i pyet t\u00eb gjith\u00eb, sidomos teknik\u00ebt, se pse at\u00ebhere nuk ka nj\u00eb shpjegim shkencor. Nga k\u00ebtu metafora paksa alkimike e m\u00ebparshme.<\/p>\n

\u00c7far\u00eb risish\u00eb t\u00eb tjera fut versioni 4 i ChatGPT?<\/strong><\/p>\n

\u00cbsht\u00eb multimodal: versioni i m\u00ebparsh\u00ebm na ka habitur sepse mund t\u00eb nd\u00ebrvepronim me t\u00eb n\u00ebp\u00ebrmjet linjave t\u00eb tekstit, por faktikisht ishte nj\u00eb lloj miku i madh n\u00eb Whatsapp q\u00eb na p\u00ebrgjigjej me linjat e tjera t\u00eb tekstit. GPT-4 mund t\u2019i d\u00ebrgojm\u00eb edhe imazhet dhe t\u00eb k\u00ebrkojn\u00eb t\u2019i komentojm\u00eb. N\u00eb raportin e OpenAi ka nj\u00eb foto shum\u00eb interesante t\u00eb nj\u00eb tipi t\u00eb lidhur pas nj\u00eb taksie n\u00eb New York q\u00eb po hekuros nj\u00eb k\u00ebmish\u00eb. N\u00ebse i k\u00ebrkojm\u00eb GPT-4 se \u00e7far\u00eb ka t\u00eb \u00e7uditshme, p\u00ebrgjigja \u00ebsht\u00eb se normalisht njer\u00ebzit nuk hekurosin sip\u00ebr taksis\u00eb, por n\u00ebse un\u00eb i jap fotografis\u00eb se kam b\u00ebr\u00eb fluturimin e nj\u00eb problemi t\u00eb fizik\u00ebs elektronike n\u00eb nivelin m\u00eb t\u00eb lart\u00eb universitar dhe i k\u00ebrkoj t\u00eb m\u00eb ndihmoj\u00eb ta zgjidh dhe t\u00eb m\u00eb jap\u00eb ar\u00ebsyetime, ai nga ekrani, me X e Y dhe dizajnet e tranzistor\u00ebve, kupton se \u00e7far\u00eb \u00ebsht\u00eb, m\u00eb thot\u00eb se cili \u00ebsht\u00eb problemi, m\u00eb jep nj\u00ebri pas tjetrit hapat dhe m\u00eb udh\u00ebheq tek zgjidhja.<\/p>\n

Megjithat\u00eb, sapo \u00ebsht\u00eb lan\u00e7uar, ChatGPT ishte nj\u00eb dhi n\u00eb matematik\u00eb.<\/strong><\/p>\n

Po, n\u00eb p\u00ebrgjith\u00ebsi modelet shum\u00eb komplekse t\u00eb gjuh\u00ebs jan\u00eb shum\u00eb keq me numrat, jan\u00eb t\u00eb zot t\u00eb v\u00ebn\u00eb bashk\u00eb fjal\u00eb, por jo t\u00eb ndjekin fillim logjik t\u00eb diskutimit. N\u00ebse ky sistem n\u00eb fakt paraqitet perfekt dhe jo vet\u00ebm ndjek kalimet logjike (nuk do t\u00eb thot\u00eb se nuk gabon, ehhh), por edhe i aft\u00eb ta njoh\u00eb matematik\u00ebn, nga ku dalin jasht\u00eb k\u00ebto kapacitete q\u00eb nuk do t\u00eb duhej t\u00eb ishin?<\/p>\n

Kjo inteligjenc\u00eb \u00ebsht\u00eb v\u00ebrtet n\u00eb gjendje t\u00eb vet\u00ebmenaxhohet si\u00e7 ia ka frik\u00ebn dikush?<\/strong><\/p>\n

Zbulojm\u00eb nga nj\u00eb detaj i dokumentit t\u00eb ril\u00ebshimit, n\u00eb faqen 54, se n\u00eb testimet nuk kufizohej t\u2019i p\u00ebrgjigjej atij q\u00eb e pyeste, por dukej se dilte nj\u00eb nj\u00eb lloj vullneti i brendsh\u00ebm q\u00eb e kan\u00eb quajtur\u00a0agjentivitet<\/em>. T\u00eb mos frik\u00ebsohemi, pasi menj\u00ebher\u00eb sakt\u00ebsohet se nuk b\u00ebhet fjal\u00eb p\u00ebr tiparet e nj\u00eb qnieje njer\u00ebzore, por se n\u00eb momente t\u00eb caktuara makina duket jasht\u00eb kontrolit ton\u00eb dhe duket se ka nj\u00eb finalitet t\u00eb sajin. Kjo \u00ebsht\u00eb nj\u00eb gj\u00eb tjet\u00ebr q\u00eb nuk e prisnim dhe q\u00ebna pyet se \u00e7far\u00eb \u00ebsht\u00eb ky agjentivitet. E p\u00ebrkthyer n\u00eb m\u00ebnyr\u00eb shum\u00eb bazike, GPT-4 duket se k\u00ebrkon\u00a0pushtet dhe resurse<\/em>. N\u00ebse do ta b\u00ebnte nj\u00eb t\u00eb ngjashmin ton\u00eb, do t\u00eb thonim se \u00ebsht\u00eb shum\u00eb egoist.<\/p>\n

Si ka mund\u00ebsi?<\/strong><\/p>\n

K\u00ebtu jemi n\u00eb fush\u00ebn e hipotezave dhe do t\u00eb sh\u00ebrbej\u00eb edhe p\u00ebr pak si studim, por GPT-4 \u00ebsht\u00eb st\u00ebrvitur mbi malet e t\u00eb dh\u00ebnave t\u00eb pranishme n\u00eb ueb, t\u00eb cil\u00ebs i shtohet nj\u00eb \u201cpraktik\u00eb p\u00ebrforcimi\u201d. Ajo q\u00eb \u00ebsht\u00eb sjell\u00eb GPT-4 n\u00eb kompjuter\u00eb n\u00eb Afrik\u00eb, ku pun\u00ebtor\u00ebt anglishtfol\u00ebs paguhen 1.50 dollar\u00eb n\u00eb or\u00eb dialogonin me k\u00ebto sisteme dhe komentonin p\u00ebrgjigjet q\u00eb jepte me nj\u00eb \u201clike\u201d apo \u201cdislike\u201d, pra nj\u00eb mekaniz\u00ebm p\u00ebrforcues. \u00cbsht\u00eb k\u00ebshtu q\u00eb kemi nd\u00ebrtuar ato\u00a0guard-rail<\/em>\u00a0p\u00ebr t\u00eb cilat n\u00ebse GPT-4\u00a0 u k\u00ebrkojm\u00eb sesi t\u00eb vet\u00ebvritemi, kjo \u00ebsht\u00eb nj\u00eb p\u00ebrgjigje q\u00eb nuk e duam dhe q\u00eb ai shmang t\u00eb na e jap\u00eb. Manika \u00ebsht\u00eb p\u00ebrshkruar me zgjedhjet q\u00eb, p\u00ebr sa duhet, jan\u00eb etike. Ama n\u00ebse nj\u00eb gj\u00eb \u00ebsht\u00eb mir\u00eb apo keq dhe kjo vlen t\u00eb pakt\u00ebn nga koha e Aristotelit, nuk evendosin thjesht n\u00eb funksion t\u00eb gj\u00ebs\u00eb n\u00eb vet\u00ebvete: t\u00eb ngr\u00ebnit e nj\u00eb akulloreje \u00ebsht\u00eb e mir\u00eb apo e keqe? Sigurisht, akullorja \u00ebsht\u00eb e mir\u00eb, por n\u00ebse po p\u00ebrgatiten p\u00ebr prov\u00ebn e kostumit t\u00eb gushtit, b\u00ebhet m\u00eb pak e mir\u00eb. N\u00ebse n\u00eb fakt sapo kam dal\u00eb nga dentisti dhe jam si i hutuar, ja q\u00eb b\u00ebhet m\u00eb e mir\u00eb nga e zakonshmja.<\/p>\n

Pra GPT-4 \u00ebsht\u00eb kusht\u00ebzuar nga komuniteti i \u201cst\u00ebrvit\u00ebsave\u201d dhe nga konteksti specifik ku ajo ndodhet?<\/strong><\/p>\n

\u00cbsht\u00eb e qart\u00eb se ia n\u00ebnshtrojm\u00eb nj\u00eb komuniteti t\u00eb varf\u00ebrish t\u00eb paguar me 1.50 dollar\u00eb n\u00eb or\u00eb q\u00eb jeton n\u00ebn pragun e varf\u00ebris\u00eb, p\u00ebr ta finaliteti m\u00eb interesant, n\u00eb p\u00ebrgjigje, \u00ebsht\u00eb ai i mbijetes\u00ebs, dometh\u00ebn\u00eb ajo q\u00eb e kemi quajtur self interest\u00a0<\/em>apo egoiz\u00ebm. Ja nj\u00eb tjet\u00ebr nivel kompleksiteti: ndoshta jemi p\u00ebrball\u00eb nj\u00eb instrumenti t\u00eb ri kolonialist? Pasi Afrik\u00ebs ia kemi zhvatur l\u00ebnd\u00ebt e para dhe krahun e pun\u00ebs me skllav\u00ebrin\u00eb, po i zhvasim edhe aft\u00ebsit\u00eb konjitive p\u00ebr t\u2019ia dh\u00ebn\u00eb sistemeve tona? Me k\u00ebt\u00eb do t\u00eb doja t\u00eb rikthehesha se pse GPT-4 \u00ebsht\u00eb kaqinteresante pse kjo jav\u00eb ka qen\u00eb kaq e trazuar. Vendi yn\u00eb gjendet n\u00eb nj\u00eb gjendje shum\u00eb t\u00eb ve\u00e7ant\u00eb. 0 \u2013 25 vje\u00e7ar\u00ebt\u00a0 jan\u00eb 42% e mbi 60 vje\u00e7ar\u00ebve. Sipas ndonj\u00eb k\u00ebrkimi, ishte nga Murtaja e Zez\u00eb q\u00eb nuk e shikonim k\u00ebt\u00eb proporcion midis t\u00eb rinjve dhe t\u00eb moshuarve. Mbetja kompetitiv\u00eb n\u00eb skenarin e pun\u00ebs nd\u00ebrkomb\u00ebtare nuk \u00ebsht\u00eb e leht\u00eb. At\u00ebhere ose b\u00ebjm\u00eb m\u00eb shum\u00eb f\u00ebmij\u00eb, ose le t\u00eb hyj\u00eb krah i ri pune \u2013 por t\u00eb dy opsionet nuk jan\u00eb leht\u00ebsisht t\u00eb zbatueshme p\u00ebr momentin \u2013 ose ta rrisim aft\u00ebsin\u00eb prodhuese t\u00eb individ\u00ebve. At\u00ebhere ky sistem i ri, prej gjendjes ton\u00eb demografike, mund t\u00eb jet\u00eb dipka jasht\u00ebzakonisht e nevojshme. Pik\u00ebrisht p\u00ebr k\u00ebt\u00eb ka shum\u00eb interes me q\u00ebllim q\u00eb ky novatoriz\u00ebn t\u00eb mbes\u00eb n\u00eb ruajtje t\u00eb asaj q\u00eb jemi dhe t\u00eb mos transformohet n\u00eb nj\u00eb lloj kolonie apo bangoprove p\u00ebr funksionimin e inteligjencave artificiale me impakt social.<\/p>\n

\u00c7far\u00eb mendoni p\u00ebr letr\u00ebn se kush d\u00ebshiron ta ngadal\u00ebsoj\u00eb zhvillimin e k\u00ebsaj teknologjie?<\/strong><\/p>\n

M\u00eb lejoni t\u2019u them se jam pak kritik, jo mbi p\u00ebrmbajtjen, por mbi dik\u00eb prej n\u00ebnshkruesve t\u00eb saj.\u00a0Sam Altman,\u00a0administrator i deleguar i OpenAi, kishte midis ortak\u00ebve themelues Elon Musk, q\u00eb n\u00eb 2018, kur Google ka shpikur Transformer q\u00eb jan\u00eb pjes\u00eb e fjal\u00ebs ky\u00e7e t\u00eb GPT (\u201cT\u201d \u00ebsht\u00eb pik\u00ebrisht p\u00ebr Trasformer), ishte i bindur se OpenAi ishte shum\u00eb prapa, aq sa t\u2019i k\u00ebrkonte Altman t\u00eb dor\u00ebhiqej dhe t\u00eb lihej q\u00eb t\u00eb ishte ai q\u00eb merrte komand\u00ebn. P\u00ebrpara refuzimit, ka b\u00ebr\u00eb\u00a0 si kush luan futboll dhe merr me vete topin: ka \u00e7investuar, hequr fonde kompanis\u00eb (n\u00eb at\u00eb koh\u00eb ishte nj\u00eb no-profit<\/em>) dhe faktikisht ka v\u00ebn\u00eb n\u00eb l\u00ebvizje at\u00eb seri ngjarjesh q\u00eb ka \u00e7uar tek ChatGPT. Altman \u00ebsht\u00eb detyruar q\u00eb t\u2019i njoh\u00eb produktet, t\u00eb monedhizoj\u00eb dhe ta p\u00ebrshpejtoj\u00eb zhvillimin e teknologjis\u00eb. P\u00ebr k\u00ebt\u00eb nuk besoj shum\u00eb n\u00eb mir\u00ebsin\u00eb e apelit t\u00eb Musk.<\/p>\n

\u00c7far\u00eb mund t\u00eb ndodh\u00eb tani n\u00eb nivel rregullues? Europa dhe Shtetet e Bashkuara kan\u00eb ecur n\u00eb rrug\u00eb t\u00eb ndryshme n\u00eb k\u00ebt\u00eb fush\u00eb. N\u00eb vitet e fundit Bashkimi Europian ka futur normativa t\u00eb ndryshme p\u00ebr teknologjin\u00eb, disa prej t\u00eb cilave duhet akoma t\u00eb b\u00ebhen plot\u00ebsisht efikase, nd\u00ebrsa Shtetet e Bashkuara kanl\u00eb ndaluar n\u00eb Section 230 e 1996. Sot \u00ebsht\u00eb n\u00eb diskutim AI Act, q\u00eb do t\u00eb rregulloj\u00eb inteligjenc\u00ebn artificiale n\u00eb funksion t\u00eb nivelit t\u00eb rrezikut. A rrezikojm\u00eb t\u00eb ndahemi dhe t\u00eb mbetemi para, n\u00eb k\u00ebrkim dhe n\u00eb ekonomi, duke par\u00eb se tashm\u00eb shum\u00eb kompani kan\u00eb nisur t\u2019i integrojn\u00eb GPT-4 n\u00eb sistemet e tyre? Apo i ashtuquajturi Brussels effect<\/em>\u00a0do t\u00eb arrij\u00eb t\u00eb influencoj\u00eb edhe ndaj zgjedhjeve amerikane?<\/strong><\/p>\n

Ndeshja \u00ebsht\u00eb krejt\u00ebsisht e hapur. Fillojm\u00eb nga t\u00eb th\u00ebnit se p\u00ebr sa i p\u00ebrket guard-railing<\/em>\u00a0t\u00eb inteligjenc\u00ebs artificiale, p\u00ebr momentin jan\u00eb vet\u00ebm kompanit\u00eb q\u00ebfusin parime t\u00eb soft\u00a0<\/em>law. \u00cbsht\u00eb Microsoft, q\u00eb ka firmosur Rome call<\/em>, ai q\u00eb i ka v\u00ebn\u00eb kufizime GPT-4. \u00cbsht\u00eb interesant modeli q\u00eb p\u00ebrdoret p\u00ebr Bing. Duhet ta mendojm\u00eb se kjo kompani na ka shtyr\u00eb nga p\u00ebrdorimi i kompjuterit me skeda me vrima n\u00eb at\u00eb \u201cC:\\>\u201d q\u00eb qen\u00eb komandat e MS-DOS, m\u00eb pas na m\u00ebsuar ta p\u00ebrdorim kompjuterin me maus dhe tastier\u00ebn me Windows (aq sa p\u00ebr t\u00eb zhvilluar at\u00eb loj\u00ebn e vock\u00ebl t\u00eb Solitario vet\u00ebm p\u00ebr t\u00eb na m\u00ebsuar t\u00eb b\u00ebjm\u00eb\u00a0drag-and-drop<\/em>), sl fundi kemi filluar ta p\u00ebrdorim kompjuterin me touch, edhe pse tableti Surface nuk i ka dal\u00eb kaq mir\u00eb dhe ka humbur konkurrenc\u00ebn n\u00eb pajisjet mobile. Ajo q\u00eb duket tani mund t\u00eb jet\u00eb nj\u00eb nd\u00ebrfaqe e re fantastike p\u00ebr p\u00ebrdorimin e kompjuteri. Jo me skeda, tastierat apo dor\u00ebn, por me t\u00eb folmen natyrale.<\/p>\n

Mund t\u00eb mjaftoj\u00eb vet\u00ebrregullimi?<\/strong><\/p>\n

Natyrisht q\u00eb jo dhe dijm\u00eb se Europa investon shum\u00eb mbi iden\u00eb e t\u00eb q\u00ebnit \u201ckontinenti q\u00eb rregullon\u201d dhe q\u00eb d\u00ebshiron t\u2019i jap\u00eb nj\u00eb standard mbrojtjeje t\u00eb konsumatorit, si\u00e7 ka ndodhur me GDPR. Ama problemet do t\u00eb lindin n\u00eb B2B, n\u00eb\u00a0business to business<\/em>, m\u00eb shum\u00eb sesa tek B2C (business to consumer<\/em>), pasi GPT ka nj\u00eb aft\u00ebsi t\u00eb madhe p\u00ebr ta ndryshuar nd\u00ebrfaqen me p\u00ebrdorim profesional t\u00eb produkteve dhe nuk ka lindur p\u00ebr t\u00eb qen\u00eb ChatGPT, chat ishte nj\u00eb demo e madhe q\u00eb ka pasur sukses t\u00eb madh, por nuk duhej t\u00eb ishte nj\u00eb produkt autonom.<\/p>\n

P\u00ebrve\u00e7 AI ACT, a ka instrumenta normativ\u00eb n\u00eb pritje?<\/strong><\/p>\n

N\u00eb AI ACT, po t\u2019i q\u00ebndrosh versionit t\u00eb fundit q\u00eb kam lexuar dit\u00ebt e kaluara, po futet nj\u00eb norm\u00eb shum\u00eb e qart\u00eb, sipas s\u00eb cil\u00ebs kompanit\u00eb jan\u00eb p\u00ebrgjegj\u00ebse n\u00eb p\u00ebrdorimin e gabuar t\u00eb Large Language Models vet\u00ebm n\u00ebse p\u00ebrdoren jasht\u00eb vijave drejtuese. Nga ana tjet\u00ebr, n\u00ebse un\u00eb p\u00ebrdor nj\u00eb Tesla p\u00ebr t\u00eb vrar\u00eb pulat n\u00eb nj\u00eb kotec, nuk mund t\u2019i k\u00ebrkoj d\u00ebmshp\u00ebrblim sht\u00ebpis\u00eb automobilistike n\u00ebse prish parafangot: nuk ishte menduar p\u00ebr k\u00ebt\u00eb. \u00cbsht\u00eb interesante ajo q\u00eb ndodh n\u00eb K\u00ebshillin e Europ\u00ebs (organizat\u00ebs nd\u00ebrkomb\u00ebtare q\u00eb p\u00ebrfshin 46 shtete an\u00ebtare dhe mb\u00ebshtetet mbi Konvent\u00ebn Europiane e t\u00eb Drejtave t\u00eb Njeriut,\u00a0sh\u00ebnimi im<\/em>), ku po negociohet nj\u00eb traktat mbi inteligjenc\u00ebn artificiale. Jo aq sa p\u00ebr t\u00eb rregulluar t\u00eb gjith\u00eb AI, pasi nuk do ta b\u00ebjm\u00eb, por p\u00ebr t\u00eb ruajtur t\u00eb drejtat themelore t\u00eb personave, n\u00eb nj\u00eb epok\u00eb ku makinat mund t\u00eb vendosin mbi ne. Me AI ACT q\u00eb ruan konsumator\u00ebt, do t\u00eb mund t\u00eb kemi dy guard-rail<\/em>\u00a0suplementare q\u00eb p\u00ebrkufizojn\u00eb t\u00eb pakt\u00ebn fush\u00ebn e loj\u00ebs.<\/p>\n

(nga Formiche<\/em>)<\/p>\n

P\u00ebrgatiti<\/strong><\/p>\n

ARMIN TIRANA<\/strong><\/p>\n","protected":false},"excerpt":{"rendered":"

Nj\u00eb intervist\u00eb me Paolo Benanti Paolo Benanti\u00a0jep Teologji n\u00eb Universitetin Papnor Gregorian, ka studiuar Inxhinieri dhe prej shum\u00eb vitesh merret (edhe) me kryq\u00ebzimin midis etik\u00ebs dhe teknologjis\u00eb. \u00cbsht\u00eb nj\u00eb prej \u201cmotor\u00ebve\u201d t\u00eb\u00a0Rome Call for AI Ethics, q\u00eb mbledh mendimtar\u00eb laik\u00eb e fetar\u00eb, institucione e kompani t\u00eb m\u00ebdha dhe promovon nj\u00eb qasje etike ndaj Inteligjenc\u00ebs Artificiale, …<\/p>\n","protected":false},"author":4,"featured_media":43655,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[42],"tags":[],"class_list":["post-43654","post","type-post","status-publish","format-standard","has-post-thumbnail","","category-kulture"],"_links":{"self":[{"href":"https:\/\/rdnews.al\/index.php\/wp-json\/wp\/v2\/posts\/43654","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/rdnews.al\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/rdnews.al\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/rdnews.al\/index.php\/wp-json\/wp\/v2\/users\/4"}],"replies":[{"embeddable":true,"href":"https:\/\/rdnews.al\/index.php\/wp-json\/wp\/v2\/comments?post=43654"}],"version-history":[{"count":0,"href":"https:\/\/rdnews.al\/index.php\/wp-json\/wp\/v2\/posts\/43654\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/rdnews.al\/index.php\/wp-json\/wp\/v2\/media\/43655"}],"wp:attachment":[{"href":"https:\/\/rdnews.al\/index.php\/wp-json\/wp\/v2\/media?parent=43654"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/rdnews.al\/index.php\/wp-json\/wp\/v2\/categories?post=43654"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/rdnews.al\/index.php\/wp-json\/wp\/v2\/tags?post=43654"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}