Vazhdon gara p\u00ebr t\u00eb kontrolluar zinxhirin global t\u00eb furnizimit p\u00ebr \u00e7ipat e Inteligjenc\u00ebs Artificiale. V\u00ebmendja nuk \u00ebsht\u00eb m\u00eb vet\u00ebm te \u00e7ipat m\u00eb t\u00eb shpejt\u00eb, por edhe n\u00eb grumbullimin e m\u00eb shum\u00eb \u00e7ipave s\u00eb bashku, shkruan The Economist<\/p>\n
N\u00eb vitin 1958, Jack Kilby, n\u00eb firm\u00ebn Texas Instruments, projektoi nj\u00eb \u00e7ip silici me nj\u00eb tranzistor t\u00eb vet\u00ebm. N\u00eb vitin 1965, firma Fairchild Semiconductor kishte m\u00ebsuar si t\u00eb prodhonte nj\u00eb \u00e7ip silici me 50 tranzistor\u00eb.<\/p>\n
Si\u00e7 v\u00ebrejti Gordon Moore, nj\u00eb nga themeluesit e kompanis\u00eb Fairchild, numri i tranzistor\u00ebve q\u00eb mund t\u00eb p\u00ebrshtateshin n\u00eb nj\u00eb cop\u00eb silici, po dyfishohej pothuajse \u00e7do vit.<\/p>\n
N\u00eb vitin 2023, kompania Apple hodhi n\u00eb treg iPhone 15 Pro, i mund\u00ebsuar nga \u00e7ipi Bionic A17, me 19 miliard\u00eb tranzistor\u00eb. Dend\u00ebsia e tranzistor\u00ebve \u00ebsht\u00eb dyfishuar 34 her\u00eb gjat\u00eb 56 viteve t\u00eb fundit. Ky p\u00ebrparim i jasht\u00ebzakonsh\u00ebm, i mb\u00ebshtetur n\u00eb t\u00eb ashtuquajturin \u201cligji i Moore\u201d ka qen\u00eb nj\u00eb nga motor\u00ebt e revolucionit kompjuterik.<\/p>\n
Nd\u00ebrsa tranzistor\u00ebt u b\u00ebn\u00eb m\u00eb t\u00eb vegj\u00ebl, ata u b\u00ebn\u00eb edhe m\u00eb t\u00eb lir\u00eb dhe m\u00eb t\u00eb shpejt\u00eb, duke lejuar t\u00eb gjitha mrekullit\u00eb e superkompjuter\u00ebve t\u00eb dit\u00ebve t\u00eb sotme. Por numri i madh i shifrave q\u00eb duhet t\u00eb operojn\u00eb programet e Inteligjenc\u00ebs Artificiale, e kan\u00eb tendosur ligjin e Moore deri n\u00eb kufijt\u00eb e tij.<\/p>\n
Rrjetet nervore artificiale q\u00eb gjenden pothuajse n\u00eb t\u00eb gjitha makinerit\u00eb moderne t\u00eb Inteligjenc\u00ebs Artificiale, duhet t\u00eb st\u00ebrviten n\u00eb m\u00ebnyr\u00eb q\u00eb t\u00eb p\u00ebrcaktojn\u00eb \u201cpeshat\u201d e duhura p\u00ebr t\u00eb mund\u00ebsuar miliarda, ndonj\u00ebher\u00eb edhe triliona, lidhje t\u00eb brendshme.<\/p>\n
K\u00ebto \u201cpesha\u201d ruhen n\u00eb form\u00ebn e matricave dhe trajnimi i modelit p\u00ebrfshin manipulimin e k\u00ebtyre matricave, duke p\u00ebrdorur matematik\u00ebn.<\/p>\n
Dy matrica (grupe numrash t\u00eb renditur n\u00eb rreshta dhe kolona), p\u00ebrdoren p\u00ebr t\u00eb gjeneruar nj\u00eb grup t\u00eb tret\u00eb; \u00e7do num\u00ebr n\u00eb at\u00eb grup t\u00eb tret\u00eb, prodhohet duke shum\u00ebzuar t\u00eb gjith\u00eb numrat n\u00eb nj\u00eb rresht n\u00eb grupin e par\u00eb, me t\u00eb gjith\u00eb numrat n\u00eb nj\u00eb kolon\u00eb t\u00eb grupit t\u00eb dyt\u00eb dhe m\u00eb pas duke i mbledhur t\u00eb gjith\u00eb.<\/p>\n
Kur matricat jan\u00eb t\u00eb m\u00ebdha, me mij\u00ebra ose dhjet\u00ebra mij\u00eb rreshta dhe kolona, dhe duhet t\u00eb shum\u00ebzohen vazhdimisht nd\u00ebrsa trajnimi i modelit vazhdon, rritet shum\u00eb numri i her\u00ebve q\u00eb numrat individual\u00eb duhet t\u00eb shum\u00ebzohen dhe t\u00eb mblidhen.<\/p>\n
Nuk \u00ebsht\u00eb vet\u00ebm trajnimi i rrjeteve nervore q\u00eb k\u00ebrkon shum\u00ebzim t\u00eb shpejt\u00eb rrufe t\u00eb matricave. Edhe prodhimi i videove me cil\u00ebsi t\u00eb lart\u00eb q\u00eb i b\u00ebn loj\u00ebrat kompjuterike arg\u00ebtuese, ka nj\u00eb procedur\u00eb t\u00eb ngjashme.<\/p>\n
Rreth 25 vite m\u00eb par\u00eb, ky ishte nj\u00eb treg shum\u00eb m\u00eb i madh. P\u00ebr t\u2019i sh\u00ebrbyer atij, kompania e prodhimit t\u00eb \u00e7ipave Nvidia, modeloi nj\u00eb lloji t\u00eb ri \u00e7ipi, t\u00eb quajtur \u201cnj\u00ebsia e p\u00ebrpunimit grafik\u201d (GPU), ku tranzistor\u00ebt u vendos\u00ebn dhe u lidh\u00ebn n\u00eb nj\u00eb m\u00ebnyr\u00eb q\u00eb i lejonte ata t\u00eb b\u00ebnin shum\u00eb shum\u00ebzime t\u00eb matricave nj\u00ebher\u00ebsh.<\/p>\n
Kur zbatohej ky proces n\u00eb Inteligjenc\u00ebn Artificiale, k\u00ebto nj\u00ebsi GPU ishin m\u00eb t\u00eb mira se nj\u00ebsit\u00eb e vjetra CPU, sepse lejonin q\u00eb t\u00eb p\u00ebrdoreshin grupe m\u00eb t\u00eb m\u00ebdha t\u00eb t\u00eb dh\u00ebnave t\u00eb trajnimit. Ato gjithashtu harxhonin shum\u00eb m\u00eb pak energji.<\/p>\n
Trajnimi i modelit AlexNet, q\u00eb solli epok\u00ebn e \u201ct\u00eb m\u00ebsuarit t\u00eb thell\u00eb\u201d n\u00eb vitin 2012, p\u00ebrfshinte caktimin e \u201cpeshave\u201d p\u00ebr 60 milion\u00eb lidhje t\u00eb brendshme. Kjo k\u00ebrkonte 4,7 x 1017 \u201coperacione n\u00eb pik\u00eb lundruese\u201d (t\u00eb quajtura FLOP).<\/p>\n
\u00c7do FLOP \u00ebsht\u00eb i barasvlefsh\u00ebm me mbledhjen ose shum\u00ebzimin e dy numrave. Deri at\u00ebher\u00eb, llogaritje kaq t\u00eb m\u00ebdha do t\u00eb kishin qen\u00eb t\u00eb pamundura. Edhe n\u00eb vitin 2012, p\u00ebrdorimi i CPU-ve m\u00eb t\u00eb mira, jo vet\u00ebm q\u00eb do t\u00eb k\u00ebrkonte shum\u00eb m\u00eb tep\u00ebr koh\u00eb dhe energji, por edhe thjeshtim t\u00eb dizajnit. Sistemi q\u00eb trajnoi modelin AlexNet, i kreu t\u00eb gjitha llogaritjet FLOP me vet\u00ebm dy nj\u00ebsi GPU.<\/p>\n
Nj\u00eb raport i fundit nga Qendra p\u00ebr Sigurin\u00eb dhe Teknologjin\u00eb n\u00eb Universitetin Georgetown, thot\u00eb se nj\u00ebsit\u00eb GPU jan\u00eb 10-100 her\u00eb m\u00eb efikase dhe deri n\u00eb 1000 her\u00eb m\u00eb t\u00eb shpejta se nj\u00ebsit\u00eb CPU kur p\u00ebrdoren p\u00ebr trajnimin e modeleve t\u00eb Inteligjenc\u00ebs Artificiale.<\/p>\n
Ishin pik\u00ebrisht ato q\u00eb b\u00ebn\u00eb t\u00eb mundur p\u00ebrhapjen e teknik\u00ebs s\u00eb \u201ct\u00eb m\u00ebsuarit t\u00eb thell\u00eb\u201d t\u00eb makinerive t\u00eb Inteligjenc\u00ebs Artificiale. Sidoqoft\u00eb, modelet e m\u00ebdha gjuh\u00ebsore (LLM) k\u00ebrkojn\u00eb llogaritje edhe m\u00eb t\u00eb nd\u00ebrlikuara.<\/p>\n
N\u00eb vitin 2018, Alec Radford, nj\u00eb studiues n\u00eb kompanin\u00eb OpenAI, zhvilloi nj\u00eb transformator gjenerues t\u00eb trajnuar paraprakisht, ose GPT, duke p\u00ebrdorur qasjen \u201ctransformatore\u201d t\u00eb p\u00ebrshkruar nga studiuesit n\u00eb Google nj\u00eb vit m\u00eb par\u00eb.<\/p>\n
Ai dhe koleg\u00ebt e tij zbuluan se aft\u00ebsia e modelit p\u00ebr t\u00eb parashikuar fjal\u00ebn e ardhshme n\u00eb nj\u00eb fjali, mund t\u00eb p\u00ebrmir\u00ebsohej duke shtuar t\u00eb dh\u00ebna trajnimi ose fuqi llogarit\u00ebse. Rritja e aft\u00ebsis\u00eb p\u00ebr t\u00eb parashikuar fjal\u00ebn e ardhshme n\u00eb nj\u00eb fjali, nuk \u00ebsht\u00eb garanci q\u00eb nj\u00eb model do t\u00eb p\u00ebrmir\u00ebsohet n\u00eb kryerjen e detyrave t\u00eb tjera. Por deri tani rezultatet kan\u00eb qen\u00eb t\u00eb k\u00ebnaqshme.<\/p>\n
Prandaj \u201cmodelet e m\u00ebdha gjuh\u00ebsore\u201d (LLM) jan\u00eb rritur m\u00eb shum\u00eb. Epoch AI, nj\u00eb pajisje k\u00ebrkimore, vler\u00ebson se trajnimi i aplikacionit GPT-4 n\u00eb vitin 2022, k\u00ebrkonte 2 x 1025 FLOP, rreth 40 milion\u00eb her\u00eb m\u00eb shum\u00eb se sa ishin p\u00ebrdorur p\u00ebr AlexNet nj\u00eb dekad\u00eb m\u00eb par\u00eb dhe kushtonte rreth 100 milion\u00eb dollar\u00eb.<\/p>\n
Gemini-Ultra, modeli m\u00eb i fuqish\u00ebm i Google, i l\u00ebshuar n\u00eb vitin 2024, raportohet se ka kushtuar dy her\u00eb m\u00eb shum\u00eb.<\/p>\n
Epoch AI llogarit se ai mund t\u00eb ket\u00eb k\u00ebrkuar 5 x 1025 FLOP. K\u00ebto jan\u00eb shifra aq t\u00eb m\u00ebdha, sa mund t\u2019i krahasojm\u00eb me t\u00eb gjitha yjet n\u00eb t\u00eb gjitha galaktikat e universit t\u00eb v\u00ebzhguesh\u00ebm, ose me pikat e ujit n\u00eb Oqeanin Paq\u00ebsor.<\/p>\n
N\u00eb t\u00eb kaluar\u00ebn, zgjidhja e nevojave t\u00eb tep\u00ebrta p\u00ebr llogaritje, ka qen\u00eb thjesht pasja e durimit. Pra, sipas k\u00ebsaj logjike, duhen pritur disa vjet dhe ligji i Moore do t\u00eb siguroj\u00eb tranzistor\u00eb m\u00eb t\u00eb shumt\u00eb dhe m\u00eb t\u00eb shpejt\u00eb n\u00eb \u00e7do \u00e7ip.<\/p>\n
Por ligjit t\u00eb Moore po i bie vrulli. Tani q\u00eb tranzistor\u00ebt individual\u00eb jan\u00eb vet\u00ebm disa dhjet\u00ebra nanometra (nj\u00eb e miliarda e metrit) t\u00eb gjer\u00eb, \u00ebsht\u00eb m\u00eb e v\u00ebshtir\u00eb t\u00eb sigurohet nj\u00eb p\u00ebrparim i vazhduesh\u00ebm.<\/p>\n
Prodhuesit e \u00e7ipave jan\u00eb ende duke punuar p\u00ebr t\u2019i b\u00ebr\u00eb tranzistor\u00ebt m\u00eb t\u00eb vegj\u00ebl dhe madje po i vendosin ata vertikalisht p\u00ebr t\u00eb futur sa m\u00eb shum\u00eb brenda \u00e7ipave. Por epoka n\u00eb t\u00eb cil\u00ebn performanca u rrit vazhdimisht, nd\u00ebrsa konsumi i energjis\u00eb ra, ka mbaruar tashm\u00eb.<\/p>\n
Nd\u00ebrsa ligji i Moore \u00ebsht\u00eb ngadal\u00ebsuar dhe d\u00ebshira p\u00ebr t\u00eb nd\u00ebrtuar modele gjithnj\u00eb e m\u00eb t\u00eb m\u00ebdha \u00ebsht\u00eb rritur, p\u00ebrgjigjja nuk kan\u00eb qen\u00eb \u00e7ipat m\u00eb t\u00eb shpejt\u00eb, por thjesht m\u00eb shum\u00eb \u00e7ipa.<\/p>\n
Ekspert\u00ebt thon\u00eb se aplikacioni GPT-4 ishte formuar me 25,000 GPU A100, t\u00eb grumbulluar s\u00eb bashku p\u00ebr t\u00eb pak\u00ebsuar humbjen e koh\u00ebs dhe energjis\u00eb q\u00eb ndodh kur l\u00ebvizen t\u00eb dh\u00ebnat midis \u00e7ipave.<\/p>\n
Pjesa m\u00eb e madhe e shum\u00ebs prej 200 miliard\u00eb dollar\u00ebsh q\u00eb kompanit\u00eb Alphabet, Amazon, Meta dhe Microsoft planifikojn\u00eb t\u00eb investojn\u00eb n\u00eb vitin 2024, do t\u00eb shkojn\u00eb n\u00eb nisma t\u00eb lidhura me Inteligjenc\u00ebn Artificiale, duke u rritur me 45% nga viti i kaluar.<\/p>\n
Thuhet se kompanit\u00eb Microsoft dhe OpenAI po planifikojn\u00eb 100 miliard\u00eb dollar\u00eb p\u00ebr projektin e quajtur Stargate. Disa ekspert\u00eb n\u00eb Silicon Valley thon\u00eb se shumat e investimit mund t\u00eb arrijn\u00eb n\u00eb 1 trilion dollar\u00eb brenda dekad\u00ebs.<\/p>\n
Nj\u00eb infrastruktur\u00eb e till\u00eb k\u00ebrkon shum\u00eb energji. N\u00eb mars, kompania Amazon bleu nj\u00eb qend\u00ebr t\u00eb dh\u00ebnash ngjitur me nj\u00eb central b\u00ebrthamor, q\u00eb mund ta furnizoj\u00eb at\u00eb me nj\u00eb gigavat energji.<\/p>\n
Investimi nuk shkon i gjithi p\u00ebr GPU-t\u00eb. Pasi nj\u00eb model \u00ebsht\u00eb trajnuar, ai duhet t\u00eb p\u00ebrdoret. Vendosja e nj\u00eb pyetjeje n\u00eb nj\u00eb sistem t\u00eb Inteligjenc\u00ebs Artificiale, zakonisht k\u00ebrkon rr\u00ebnj\u00ebn katrore t\u00eb sasis\u00eb s\u00eb llogaritjes s\u00eb p\u00ebrdorur p\u00ebr ta trajnuar at\u00eb.<\/p>\n
Kjo \u00ebsht\u00eb nj\u00eb llogaritje shum\u00eb e nd\u00ebrlikuar. P\u00ebr aplikacionin GPT-3, i cili k\u00ebrkonte 3 x 1023 FLOP p\u00ebr t\u2019u st\u00ebrvitur, nj\u00eb \u201ckonkluzion\u201d tipik, mund t\u00eb marr\u00eb 3 x 1011 FLOP.<\/p>\n
\u00c7ipat e njohur si FPGA dhe ASIC, t\u00eb p\u00ebrshtatur p\u00ebr t\u00eb nxjerr\u00eb p\u00ebrfundime (konkluzione), mund t\u00eb ndihmojn\u00eb p\u00ebr t\u2019i b\u00ebr\u00eb modelet e Inteligjenc\u00ebs Artificiale m\u00eb efikase sesa me p\u00ebrdorimin e nj\u00ebsive GPU.<\/p>\n
Sidoqoft\u00eb, \u00ebsht\u00eb kompania Nvidia ajo q\u00eb ka p\u00ebrfituar m\u00eb tep\u00ebr nga bumi i Inteligjenc\u00ebs Artificiale. Kompania tani vlen 2.8 trilion\u00eb dollar\u00eb, tet\u00eb her\u00eb m\u00eb shum\u00eb se sa kur aplikacioni ChatGPT u hodh n\u00eb treg n\u00eb vitin 2022.<\/p>\n
Pozicioni i saj mbizot\u00ebrues n\u00eb treg nuk mb\u00ebshtetet vet\u00ebm n\u00eb njohurit\u00eb e grumbulluara gjat\u00eb prodhimit t\u00eb GPU-s\u00eb dhe aft\u00ebsive p\u00ebr t\u00eb mobilizuar shum\u00eb kapital (shefi i kompanis\u00eb Jensen Huang, thot\u00eb se \u00e7ipat m\u00eb t\u00eb fundit t\u00eb Nvidia-s, t\u00eb quajtur Blackwell, kushtuan 10 miliard\u00eb dollar\u00eb p\u00ebr t\u2019u zhvilluar).<\/p>\n
Kompania gjithashtu p\u00ebrfiton nga zot\u00ebrimi i softuerit t\u00eb p\u00ebrdorur p\u00ebr t\u00eb programuar \u00e7ipat e saj. Ky softuer quhet CUDA dhe mbahet si standardi i industris\u00eb. Dhe ka nj\u00eb pozicion kryesor n\u00eb rrjetin q\u00eb p\u00ebrdoret p\u00ebr t\u00eb lidhur \u00e7ipat s\u00eb bashku.<\/p>\n
Konkurrent\u00ebt pretendojn\u00eb se \u00e7ipat e Nvidia-s kan\u00eb disa dob\u00ebsi. Rodrigo Liang i firm\u00ebs SambaNova Systems thot\u00eb se \u00e7ipat e Nvidia-s kan\u00eb disa disavantazhe q\u00eb mund t\u00eb gjurmohen n\u00eb p\u00ebrdorimet e tyre fillestare n\u00eb videoloj\u00ebra. Nj\u00eb disavantazh ve\u00e7an\u00ebrisht i madh \u00ebsht\u00eb kapaciteti i tyre i kufizuar p\u00ebr l\u00ebvizjen dhe fikjen e t\u00eb dh\u00ebnave (pasi nj\u00eb model i t\u00ebr\u00eb nuk p\u00ebrmblidhet dot n\u00eb nj\u00eb nj\u00ebsi GPU).<\/p>\n
Cerebras, nj\u00eb tjet\u00ebr konkurrent, reklamon nj\u00eb procesor me gjer\u00ebsi 21.5 cm. Nd\u00ebrsa GPU-t\u00eb e tjera p\u00ebrmbajn\u00eb dhjet\u00ebra mij\u00eb \u201cb\u00ebrthama\u201d q\u00eb kryejn\u00eb llogaritje n\u00eb t\u00eb nj\u00ebjt\u00ebn koh\u00eb, ky procesor ka pothuajse nj\u00eb milion.<\/p>\n
Kompania thot\u00eb se ai p\u00ebrdor vet\u00ebm gjysm\u00ebn e energjis\u00eb s\u00eb \u00e7ipit m\u00eb t\u00eb mir\u00eb t\u00eb Nvidia-s. Google ka krijuar \u201cnj\u00ebsin\u00eb e p\u00ebrpunimit t\u00eb tensorit\u201d (TPU), q\u00eb mund t\u00eb p\u00ebrdoret si p\u00ebr trajnim, ashtu edhe p\u00ebr t\u00eb nxjerr\u00eb konkluzione.<\/p>\n
Modeli i tij Gemini 1.5 \u00ebsht\u00eb n\u00eb gjendje t\u00eb g\u00eblltis\u00eb tet\u00eb her\u00eb m\u00eb shum\u00eb t\u00eb dh\u00ebna nj\u00ebher\u00ebsh sesa aplikacioni GPT-4.<\/p>\n
Vlera e madhe dhe n\u00eb rritje e nj\u00ebsis\u00eb GPU \u00ebsht\u00eb shfryt\u00ebzuar si nj\u00eb lev\u00eb gjeopolitike. Megjith\u00ebse industria e \u00e7ipave \u00ebsht\u00eb globale, nj\u00eb num\u00ebr i vog\u00ebl vendesh kontrollojn\u00eb qasjen n\u00eb nivelet m\u00eb t\u00eb larta t\u00eb Inteligjenc\u00ebs Artificiale.<\/p>\n
\u00c7ipat e Nvidia-s modelohen n\u00eb Amerik\u00eb. Makinerit\u00eb m\u00eb t\u00eb p\u00ebrparuara t\u00eb litografis\u00eb n\u00eb bot\u00eb, t\u00eb cilat kthejn\u00eb dizajnet n\u00eb silic, jan\u00eb b\u00ebr\u00eb t\u00eb gjitha nga ASML, nj\u00eb kompani holandeze me vler\u00eb 350 miliard\u00eb dollar\u00eb.<\/p>\n
N\u00eb k\u00ebto makineri, kan\u00eb qasje vet\u00ebm punishtet m\u00eb t\u00eb p\u00ebrparuara n\u00eb bot\u00eb, si TSMC e Tajvanit, nj\u00eb firm\u00eb me vler\u00eb rreth 800 miliard\u00eb dollar\u00eb dhe Intel e Amerik\u00ebs. Kjo vlen dhe p\u00ebr shum\u00eb pajisje t\u00eb tjera m\u00eb t\u00eb vogla, dhe n\u00eb k\u00ebt\u00eb grup t\u00eb vog\u00ebl t\u00eb privilegjuar shtetesh, do t\u00eb fusnim edhe Japonin\u00eb.<\/p>\n
K\u00ebto kufizime kan\u00eb b\u00ebr\u00eb t\u00eb mundur q\u00eb qeveria amerikane t\u00eb vendos\u00eb kontrolle t\u00eb ashpra dhe efektive mbi eksportin e \u00e7ipave t\u00eb p\u00ebrparuar n\u00eb Kin\u00eb. P\u00ebr rrjedhoj\u00eb, kinez\u00ebt po investojn\u00eb qindra miliarda dollar\u00eb p\u00ebr t\u00eb krijuar zinxhirin e tyre t\u00eb furnizimit me \u00e7ipa.<\/p>\n
Shumica e analist\u00ebve besojn\u00eb se Kina \u00ebsht\u00eb ende disa vite prapa n\u00eb k\u00ebt\u00eb k\u00ebrkim, por p\u00ebr shkak t\u00eb investimeve t\u00eb m\u00ebdha nga kompani si Huawei, ajo \u00ebsht\u00eb p\u00ebrballur me kontrollet e eksportit shum\u00eb m\u00eb mir\u00eb sesa priste Amerika.<\/p>\n
Amerika po investon gjithashtu n\u00eb k\u00ebt\u00eb drejtim. Kompania TSMC, q\u00eb shihet si shpresa q\u00eb do t\u00eb mbetet n\u00eb rast se Kina vendos t\u00eb pushtoj\u00eb Tajvanin, po shpenzon rreth 65 miliard\u00eb dollar\u00eb p\u00ebr t\u00eb nd\u00ebrtuar fabrika n\u00eb Arizona, ku rreth 6.6 miliard\u00eb dollar\u00eb jan\u00eb dh\u00ebn\u00eb si subvencione.<\/p>\n
Shtete t\u00eb tjera, si India (10 miliard\u00eb dollar\u00eb), Gjermania (16 miliard\u00eb dollar\u00eb), Japonia (26 miliard\u00eb dollar\u00eb), po rrisin investimet. Blerja e \u00e7ipave t\u00eb Inteligjenc\u00ebs Artificiale po b\u00ebhet gjithnj\u00eb e m\u00eb e leht\u00eb.\/ Monitor.al<\/p>\n
The post Inteligjenca Artificiale, gara e \u00e7ipave<\/a> appeared first on Lapsi.al<\/a>.<\/p>","protected":false},"excerpt":{"rendered":" Vazhdon gara p\u00ebr t\u00eb kontrolluar zinxhirin global t\u00eb furnizimit p\u00ebr \u00e7ipat e Inteligjenc\u00ebs Artificiale. V\u00ebmendja nuk \u00ebsht\u00eb m\u00eb vet\u00ebm te \u00e7ipat m\u00eb t\u00eb shpejt\u00eb, por edhe n\u00eb grumbullimin e m\u00eb shum\u00eb \u00e7ipave s\u00eb bashku, shkruan The Economist N\u00eb vitin 1958, Jack Kilby, n\u00eb firm\u00ebn Texas Instruments, projektoi nj\u00eb \u00e7ip silici me nj\u00eb tranzistor t\u00eb vet\u00ebm. […]<\/p>\n","protected":false},"author":0,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-697","post","type-post","status-publish","format-standard","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/zennihost.com\/test\/wp-json\/wp\/v2\/posts\/697","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/zennihost.com\/test\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/zennihost.com\/test\/wp-json\/wp\/v2\/types\/post"}],"replies":[{"embeddable":true,"href":"https:\/\/zennihost.com\/test\/wp-json\/wp\/v2\/comments?post=697"}],"version-history":[{"count":0,"href":"https:\/\/zennihost.com\/test\/wp-json\/wp\/v2\/posts\/697\/revisions"}],"wp:attachment":[{"href":"https:\/\/zennihost.com\/test\/wp-json\/wp\/v2\/media?parent=697"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/zennihost.com\/test\/wp-json\/wp\/v2\/categories?post=697"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/zennihost.com\/test\/wp-json\/wp\/v2\/tags?post=697"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}