Izinto Ezimangalisayo Ezimayelana Nobuhlakani Bokwenziwa

Ikhaya » Python » Izinto Ezimangalisayo Ezimayelana Nobuhlakani Bokwenziwa

Ubuhlakani bokwenziwa bukhona yonke indawo, ngisho noma singabuqapheli kahle. Kusukela ochungechungeni oluphakanyiswa yiNetflix namuhla ebusuku, kuya ohlwini lokudlalayo iSpotify elungiselela uhambo lwakho lokuya emsebenzini, kuya kumsizi wezwi ophendulayo kusuka ocingweni lwakho, ingxenye enkulu yempilo yakho yedijithali iqhutshwa buthule yi-AI. Akuseyona nje into yesikhathi esizayo; ungqimba olungabonakali olwenza okuqukethwe kube ngokwakho, lwenze imisebenzi ngokuzenzakalela futhi lusize izinkampani ukuthi zenze izinqumo ngezinga elikhulu.

Uma unesifiso sokwazi nge-AI, awuwedwa neze. Lobu buchwepheshe bunomlando omude ngokumangalisayo, inqwaba yezenzakalo eziphawulekayo, izindaba ezixakile (ezinye ziyesabeka kancane, ezinye zihlekisa nje) kanye ne-tsunami yedatha ngemuva kwayo. Ngezansi uzothola uhambo olunemininingwane, oluhlanganisa izinto ezimangalisayo zomlando, izinhlelo zokusebenza zansuku zonke, ukwehluleka okudumile kanye namaqiniso athile angaziwa kakhulu abonisa kokubili amandla kanye nemikhawulo yobuhlakani bokwenziwa banamuhla.

Lokho esikushoyo ngempela uma sikhuluma ngobuhlakani bokwenziwa

Kalula nje, ubuhlakani bokwenziwa buyikhono lemishini lokwenza imisebenzi evame ukudinga ubuhlakani bomuntu. Lokho kufaka phakathi izinto ezifana nokucabanga, ukufunda kokuhlangenwe nakho, ukuhlela, ukuqaphela amaphethini ezithombeni noma emsindweni, kanye nokwenza izinqumo ngokusekelwe olwazini oluyinkimbinkimbi. Izinhlelo ze-AI zanamuhla zingacubungula idatha eningi, zithole amaphethini abantu abangawaphuthelwa kalula futhi ziqhubeke nokuthuthukisa ukuziphatha kwazo njengoba zithola ulwazi olusha.

Ngobuchwepheshe, uhlelo lwe-AI luyinoma yiluphi uhlelo lobuchwepheshe olungabona indawo yalo, luhlaziye okwenzekayo futhi lwenze okuthile ukuze lufinyelele umgomo othile. Umshini uthola idatha (kungaba elungiselelwe kusengaphambili noma ethathwe ngezinzwa ezifana namakhamera nama-microphone), ucubungula lolo lwazi ngama-algorithms namamodeli, bese ukhiqiza impendulo: isibikezelo, isincomo, isinqumo noma isenzo. Isici esibalulekile ukuthi eziningi zalezi zinhlelo zingazivumelanisa ngokuhamba kwesikhathi, zizilungise ngokwazo ngokusekelwe emiphumeleni yezenzo zazo zangaphambilini.

Okuhlukanisa ngempela i-AI ukuzimela kanye nokuzivumelanisa nezimo. Esikhundleni sokudinga umuntu ukuthi acacise isinyathelo ngasinye, izinhlelo ze-AI eziklanywe kahle zifunda amaphethini ngqo kusuka kudatha. Ngamasu anjengokufunda komshini nokufunda okujulile, angathuthukisa ukusebenza kwawo njengoba evezwa izibonelo eziningi, ngokuvamile afinyelela noma adlule ngisho nokunemba kwezinga lomuntu emisebenzini echazwe kahle njengokuhlukanisa izithombe noma ukuqashelwa kwenkulumo.

Nakuba eminye yemibono eyinhloko ngemuva kwe-AI ibuyela emuva eminyakeni engaphezu kwengu-50, intuthuko yakamuva ibe nkulu kakhulu. Intuthuko emandleni okusebenzisa amakhompyutha, ukutholakala kwamasethi edatha amakhulu kanye nama-algorithms ayinkimbinkimbi kuguqule i-AI yaba ubuchwepheshe obusezingeni eliphezulu enhliziyweni yokuguqulwa kwedijithali komphakathi. Namuhla, i-AI ifakiwe kumanethiwekhi okuxhumana ngocingo, amapulatifomu amafu kanye nesofthiwe yebhizinisi, futhi iqhuba konke kusukela ezincomweni zokuqukethwe kuya ezenzakalelayo zezimboni.

Umlando omfushane (kodwa owembulayo) wobuhlakani bokwenziwa

Indaba ye-AI ivame ukuqala ngo-Alan Turing, isazi sezibalo saseBrithani esasisebenza ngawo-1930 nawo-1940. I-Turing ibhekwa kabanzi njengomunye wobaba besayensi yamakhompyutha kanye nobuhlakani bokwenziwa njengomkhakha. Ngo-1950, washicilela iphepha elidumile elithi “Computing Machinery and Intelligence” ephephabhukwini i-Mind, lapho aphakamisa khona lokho okwaziwa ngokuthi i-Turing Test: indlela yokusebenza yokubuza ukuthi umshini “ungacabanga” yini.

I-Turing Test empeleni ngumdlalo wokulingisa. Umhloli womuntu uxhumana ngombhalo nezinto ezimbili ezifihliwe, umuntu oyedwa nomshini owodwa. Uma, ngemva kochungechunge lwemibuzo, umhloli engakwazi ukusho ngokuthembekile ukuthi iyiphi, kuthiwa umshini uphumelele ukuhlolwa. Ngisho noma ukuhlolwa kunemikhawulo kanye nezingxoxo eziningi zefilosofi ezizungezile, kuhlala kungomunye wemibono evelele kakhulu ku-AI.

Izazi-mlando eziningi ziphawula u-1956 njengokuzalwa okusemthethweni kobuhlakani bokwenziwa besimanje. Ngalowo nyaka, eNgqungqutheleni yaseDartmouth, abacwaningi uJohn McCarthy, uMarvin Minsky noClaude Shannon bethula ngokusemthethweni igama elithi “ubuhlakani bokwenziwa”, bechaza ukuthi “isayensi nobunjiniyela bokwenza imishini ehlakaniphile, ikakhulukazi izinhlelo zekhompyutha ezihlakaniphile”. Lo mhlangano wahlanganisa amaphayona abeka izisekelo zocwaningo lwe-AI lwamashumi eminyaka.

Kusukela kulezo zinsuku zokuqala, i-AI idlule ezindleleni eziningana zokuba nethemba nokudumazeka. Ngemva kwentshiseko yokuqala ngawo-1950 nawo-1960, le nsimu yabhekana nalokho okubizwa ngokuthi “ubusika be-AI” lapho intuthuko yehla futhi ukuxhaswa ngemali kunqamuka. Kodwa-ke, igagasi ngalinye elisha lamandla ekhompyutha kanye nedatha lavuselela isithakazelo. Isikhathi esibalulekile ekukhuleni kwakamuva safika cishe ngo-2012, lapho izinhlelo zokufunda ezijulile zaqala ukuthola imiphumela emihle kakhulu ekuqaphelweni kwenkulumo kanye nezithombe, okwenza ochwepheshe abaningi, kuhlanganise nomcwaningi uYoshua Bengio, baqokomise lowo nyaka njengendawo yokuqala yangempela ye-AI yanamuhla.

Izenzakalo ezidumile: kusukela ku-chess kuya ku-Go nangale kwalokho

Esinye sezikhathi zokuqala ze-AI ezabamba ukunaka komhlaba wonke senzeka ngo-1997. I-supercomputer ye-IBM i-Deep Blue inqobe umpetha we-chess womhlaba ophethe uGarry Kasparov emdlalweni wemidlalo eyisithupha. Kwabaningi, lowo mcimbi wawufanekisela ukuphuma kwe-AI ezindaweni zokucwaninga bese ungena ezindabeni zansuku zonke, okubonisa ukuthi umshini ungamnqoba umkhulu womuntu endaweni ehlotshaniswa nokucabanga okujulile kwamasu.

Ubuhlobo bukaKasparov ne-AI abuzange buphele lapho. Ngo-2003 wadlala nolunye uhlelo lwe-chess oluthuthukisiwe, i-Deep Junior. Kulokhu, umdlalo waphela ngokulingana, kodwa uKasparov wavuma ngokusobala ukuthi wayehlabeke umxhwele kangakanani ngentuthuko ye-computer chess kanye nobuciko obuboniswa ezinye zezinyathelo zomshini, ezazingafani njalo nokubala kwamandla aluhlaza.

Eminyakeni eminingana kamuva, i-AI yaphinde yathatha isiteji, kulokhu kuthelevishini yaseMelika. Ngo-2011, i-IBM yethula uWatson, uhlelo lokuphendula imibuzo oluklanyelwe ukuncintisana ohlelweni lwemibuzo oluthi “Jeopardy!”. Ngesikhathi somdlalo, uWatson kwadingeka ahumushe izinkomba zolimi lwemvelo, akhulume ngesikhathi esifanele futhi akhiqize izimpendulo ezinembile ngokushesha kunababili babanqobi abangcono kakhulu babantu. UWatson wawina, ebonisa ukuthi i-AI ingaphatha ulimi olunomsindo nolungaqondakali ngesikhathi sangempela futhi ifunde njengoba umdlalo uqhubeka.

Ngawo lowo nyaka, i-AI yaqala ukufika emaphaketheni abantu. I-Apple yethule i-Siri ku-iPhone 4S, enye yezinsizakalo zokuqala zezwi ezisetshenziswa kabanzi. Ezinyangeni ezimbalwa kamuva, ngo-2012, i-Google ne-Microsoft basungula izinsiza zabo zedijithali, behlanganisa ukuqonda ulimi lwemvelo, ukusesha kanye nokwenza kube ngokwakho. Lawo mathuluzi aphawula ukuqala kwesizukulwane sesibili se-AI yabathengi, lapho ukukhuluma nefoni yakho kwaqala khona ukuba yinto evamile.

Esinye isenzakalo esiphawulekayo safika ngo-2016 ngomdlalo we-Go, umdlalo webhodi obhekwa njengoyinkimbinkimbi kakhulu kune-chess. Uhlelo lwe-Google DeepMind i-AlphaGo ibhekane noLee Sedol (ovame ukulotshwa ngokuthi uLee Se‑dol), omunye wabadlali be-Go abaphambili emhlabeni, emdlalweni wemidlalo emihlanu. I-AlphaGo iwine iningi lemidlalo, yashaqisa ochwepheshe abaningi ababekholelwa ukuthi impumelelo enjalo isasele iminyaka eyishumi. Ezinye zezinyathelo ze-AlphaGo zazingavamile kangangokuthi abadlali abangochwepheshe bazichaza ngokuthi “zinobuciko” noma “zinhle”, okugqamisa ukuthi i-AI ingathola amasu abantu abangakaze bawacabange.

Ikhono le-AI lokusingatha imidlalo yamasu alizange liphele nge-chess ne-Go. Ngo-2017, abacwaningi eCarnegie Mellon University bathuthukise iLibratus, i-algorithm eyenzelwe ukudlala i-Texas Hold'em poker engenamkhawulo. Ngokungafani ne-chess noma i-Go, i-poker ihilela ulwazi olufihliwe kanye nokukhohlisa. ILibratus idlale nabadlali abane abangochwepheshe abaphezulu emncintiswaneni onzima ekhasino yase-US futhi yawina ngendlela ekholisayo, okubonisa ukuthi i-AI ingakwazi ukwenza izinqumo eziyinkimbinkimbi ngaphansi kokungaqiniseki.

Ngalesi sikhathi, izinkampani ezingaphandle kwezinkampani ezinkulu zobuchwepheshe zendabuko nazo zazingena. Ngo-2017, i-Telefónica yethule i-Aura, ipulatifomu yokuqonda yokusebenzisana namakhasimende ayo. Nge-Aura, abasebenzisi babekwazi ukuphatha izinsizakalo futhi bathole ukwesekwa ngokusebenzisa izixhumanisi zemvelo, zengxoxo eziteshini eziningi. Kwakungenye yezindlela zokuqala ezinkulu zokusetshenziswa komqhubi wezokuxhumana esebenzisa i-AI njengengqimba eyinhloko yokusebenzisana namakhasimende.

Kusukela ngo-2018 kuqhubeke, i-AI yasakazeka ngamandla kuyo yonke imikhakha emisha. Abakhiqizi bezimoto ezifana neTesla, i-Audi nabanye bahlanganise i-AI ezinhlelweni zokushayela ezizimele kanye nezinhlelo zokusiza abashayeli ezithuthukisiwe (i-ADAS). Kwezokuvakasha, ukuhamba kanye nezamabhange, izinkampani zaqala ukuqashelwa kwezithombe kanye namamodeli okubikezela ukuze ziqonde ukuziphatha kwabasebenzisi, zilula ukusebenza futhi zenze okunikezwayo kube ngokwakho. Futhi lapho ubhubhane lwe-COVID-19 lufika ngo-2020, i-AI yasiza ekuhlaziyeni idatha yezempilo, yasekela ukuhlolwa kokushisa ezindaweni zomphakathi futhi yafaka isandla ekutholakaleni kokuqubuka kusenesikhathi ngokuhlaziywa kwedatha enkulu kanye namamodeli e-epidemiological.

Izibikezelo mayelana nomthelela wezomnotho we-AI ziyamangalisa ngokufanayo. Ochwepheshe balinganisele ukuthi imisebenzi ehlobene ne-AI ingakhiqiza imali engaphezu kwamaRandi ayizigidi eziyizinkulungwane ezingu-300 ngonyaka ngenani lebhizinisi cishe maphakathi no-2020, okubonisa ukuthi lobu buchwepheshe bube yinto ebalulekile kangakanani ekuncintisaneni komhlaba wonke.

Imithethonqubo, amanethiwekhi kanye nesidingo semithetho emisha

Nakuba i-AI iletha izinzuzo ezinkulu, iphakamisa nemibuzo enzima mayelana nomthethonqubo, ukuncintisana kanye namalungelo. EYurophu ikakhulukazi, izikhungo ezifana nePhalamende laseYurophu kanye nohulumeni bezwe bebelokhu bexoxa ngendlela yokwakha izinhlaka ezivikela izakhamuzi futhi ziqinisekise ukuncintisana okufanele, kuyilapho zivumela ukusungula izinto ezintsha. Lokhu kubaluleke kakhulu ngoba amanethiwekhi ezokuxhumana kanye nedatha yesimanje enza okungaphezu nje kokuthutha: manje asewumgogodla wezinsizakalo eziqhutshwa yi-AI.

Iqiniso elilodwa eliphawulekayo ukuthi ingxenye enkulu yethrafikhi ye-inthanethi yanamuhla ikhiqizwa yimishini ekhuluma neminye imishini. Izilinganiso ezingacacile zisikisela ukuthi cishe u-40% wethrafikhi iyonke ivela ezinhlelweni ezizenzakalelayo, ama-API, ama-bot kanye namadivayisi e-IoT kunokuba kusetshenziswe abantu. Njengoba amadivayisi amaningi kanye nezinsizakalo zithembele ku-AI, inani lokuxhumana komshini kuya komshini lizokhula kuphela, okwengeza ingcindezi kumanethiwekhi futhi kuphakamise imibuzo mayelana nokuphepha, ubumfihlo kanye nokuphathwa.

Ngenxa yokuthi i-AI ingaguqula inani elikhulu ledatha eluhlaza ibe ulwazi olusebenzisekayo, abalawuli bangaphansi kwengcindezi yokuvuselela imithetho yomdlalo. Umgomo ukudala indawo lapho amabhizinisi abo bonke ubukhulu kanye nezakhamuzi zingazuza khona ku-AI ngaphandle kokuchayeka emikhubeni engalungile, ukungacaci noma ukusetshenziswa kabi kolwazi lomuntu siqu. Imibono efana nokucaca, ukuchazeka, ukunciphisa idatha kanye nokuziphendulela kuyimbangela yemithetho emisha kanye nezindinganiso zemboni.

Ukusetshenziswa okumangalisayo nokusebenzayo kwe-AI empilweni yansuku zonke

Ngale kwezinyathelo ezinkulu kanye nezingxoxo zezepolitiki, i-AI igcwele amaqhinga amancane, ansuku zonke athuthukisa impilo yakho buthule. Eziningi zalezi zinhlelo zokusebenza zihlanganiswe kakhulu nezinhlelo zokusebenza namadivayisi kangangokuthi asicabangi ukuthi “ziyi-AI”. Kodwa zithembele kumamodeli ayinkimbinkimbi afunda ekuziphatheni kwakho, azivumelanise nezintandokazi zakho futhi aphathe umsebenzi ophindaphindwayo kuwe.

Esinye sezibonelo ezicacile ukuthi i-AI ifunda kanjani izinto ozithandayo kwezokuzijabulisa. Cabanga ukuthi ubukela njalo amahlaya ama-80s kanye nochungechunge lwe-animated lwesikhathi esizayo epulatifomu yokusakaza njenge-Netflix. Injini yokuncoma ngemuva kwezigcawu isebenzisa umlando wakho wokubuka, imibuzo yokusesha, izilinganiso ngisho nokuthi uhlala isikhathi esingakanani esihlokweni ngasinye ukwakha iphrofayili yezintandokazi zakho. Bese ibikezela ukuthi yimaphi ama-movie noma izinhlelo ongase uzijabulele kakhulu futhi iziveze ngokuzenzakalelayo, ikusindise inkinga yokuskrola okungapheli.

Umqondo ofanayo udlulela nasekuhleleni ukuhamba kanye nezinsizakalo eziku-inthanethi. Uma uvame ukubhuka amahhotela kumapulatifomu athile, ushiye ukubuyekezwa ezindaweni ezithile noma ukhetha njalo izinhlobo ezifanayo zokuhlala, izinhlelo ze-AI zingathola amaphethini ekuziphatheni kwakho. Ngesikhathi esizayo lapho usesha uhambo, lezo zinhlelo zingasikisela izindawo oya kuzo, izinsuku noma amahhotela ahambisana nokukhetha kwakho kwangaphambilini, ngokuvamile ahlanganisa idatha yamanani, ukutholakala kanye nesitayela sakho somuntu siqu sokuhamba.

I-AI ithatha nemisebenzi eminingi yokwenza izinqumo, yize ihlala ingaphakathi kwemingcele ebekwe ngabantu. Lezi zinhlelo zingashesha kakhulu ekuhloleni izimo eziyinkimbinkimbi, eziningi lapho isivinini sibaluleke khona. Isibonelo, ezimotweni zesimanje ezifakwe i-ADAS, amamodeli e-AI ahlaziya njalo okufakiwe okuvela kumakhamera, i-radar nezinye izinzwa. Uma bebona isithiyo esingazelelwe, banganquma ngesikhashana nje ukuthi babambe amabhuleki noma baqondise kancane ukuze bagweme ukushayisana. Ukwazi ukuthi imoto ingasabela ngokushesha kunomuntu ezimweni eziphuthumayo kungaba yinto eduduzayo kunokuba yesabise.

Ekhaya, i-AI isibe uhlobo lomsizi wedijithali oguquguqukayo. Akusadingeki ukuthi ubheke amawebhusayithi amaningi ukuze uthole ishidi lokufaka imibala elinomlingiswa oyintandokazi yengane yakho noma iresiphi ephelele ngokubuka izithombe ezinhle enjinini yokusesha. Esikhundleni salokho, ungathayipha umyalo wolimi lwemvelo esithombeni noma kumkhiqizi wombhalo osebenzisa i-AI bese uthola umphumela olungiselelwe wena ngemizuzwana, kusukela kumidwebo yangokwezifiso kuya kuziphakamiso zeresiphi ngokusekelwe ezithakweni ezisefrijini lakho.

Ezindaweni zobungcweti, i-AI ivame ukuhlanganiswa nokuzenzakalela kanye nokulawula okuqinile. Onjiniyela kanye nochwepheshe besizinda baklama imisebenzi lapho i-AI isekela khona izinqumo kunokuzithatha zodwa. Isibonelo, amathuluzi e-AI angahlaziya kusengaphambili izithombe zezokwelapha, aqokomise okungahambi kahle kodokotela, abeke phambili amathikithi okusekela amaqembu e-IT noma abhale imibiko yokuhlaziya yabaphathi. Ochwepheshe babantu bese bebuyekeza, baqinisekise futhi balungise leyo miphumela, besebenzisa i-AI njengesisheshisi esikhundleni sokuthatha indawo.

I-tsunami enkulu yedatha ephakela i-AI yesimanje

Enye yezinto ezimangalisayo kakhulu nge-AI inani elikhulu ledatha ethembele kuyo. Eminyakeni yamuva nje, umhlaba ungene enkathini yokukhula kolwazi okungakaze kubonwe ngaphambili. Ngo-2024, ukukhiqizwa kwedatha yomhlaba wonke kulinganiselwa ukuthi sekufinyelele kuma-zettabyte angaba ngu-130. Ukuqagela kusikisela ukuthi ngo-2025 lelo nani lingakhuphuka lifinyelele kuma-zettabyte angaba ngu-181, ukwenyuka cishe ngo-40% ngonyaka owodwa ngokusho kwe-Global DataSphere ye-IDC.

I-zettabyte eyodwa iyinani elicishe lingacabangeki. I-zettabyte eyodwa ilingana ne-trillion gigabytes eyodwa (okungu-1 olandelwa yi-zeros ezingama-21 ngama-byte). Ukubeka ngendlela ehlukile, ukugcina i-ZB eyodwa kuma-smartphones anomthamo ongu-128 GB kungadinga amafoni angaphezu kwezigidigidi ezingu-7.8, cishe idivayisi eyodwa ngomuntu ngamunye eMhlabeni. Ukwandisa lokho kufike ku-130 noma i-181 zettabytes kwembula ubukhulu obumangalisayo bendawo yonke yedijithali.

Iningi lalolu lwazi ludalwe ngendlela eshaqisayo muva nje. Izilinganiso ezingacacile zisikisela ukuthi cishe u-90% wayo yonke idatha ekhona namuhla ikhiqizwe eminyakeni embalwa edlule, okusho ukuthi isintu sikhiqize ulwazi oluningi lwedijithali ngesikhathi esincane kunanoma yimuphi umlando wangaphambilini uhlanganisiwe. Lokhu kuqhuma kukhuthazwa yi-inthanethi, amadivayisi eselula, izinsizakalo zamafu kanye nezinzwa ezixhunyiwe kanye nemishini.

Njalo ngomzuzu, umhlaba oku-inthanethi ukhipha okuqukethwe okumangalisayo. Amakhulu amahora amavidiyo alayishwa ezinkundleni ezifana ne-YouTube, amakhulu ezinkulungwane zemiyalezo avela kumanethiwekhi afana ne-X (eyayikade i-Twitter), amakhulu ezinkulungwane zezindaba zabelwana ngazo ku-Instagram kanye namashumi ezigidi zemiyalezo ethunyelwa ngezinhlelo zokusebenza ezifana ne-WhatsApp. Lezi izinombolo ezingacacile nezishintshashintshayo, kodwa zibonisa ukuthi i-data firehose isibe nzima kangakanani.

Ukuvela kwe-Internet of Things kungenye indlela enkulu yokuthuthukisa idatha. Sekuvele kunamadivayisi e-IoT angaphezu kwezigidigidi ezingu-30 emhlabeni jikelele, kusukela kuzipikha ezihlakaniphile kanye nezinsimbi zomnyango emakhaya kuya kuzinzwa zezimboni emafektri, ezimotweni nasezitshalweni zikagesi. Amaningi ala madivayisi ahlala elinganisa izinga lokushisa, ukudlidliza, indawo, amaphethini okusetshenziswa nokuningi, esakaza ulwazi efwini lapho amamodeli e-AI angaluhlaziya khona ukuze enze ngcono ukusebenza, athole ukungalingani noma abikezele izidingo zokulungisa.

Imininingwane ebalulekile ukuthi cishe u-90% wayo yonke idatha "ayihlelekile". Leso sigaba sihlanganisa amavidiyo, izithombe, amafayela alalelwayo, ama-imeyili, izingodo zengxoxo kanye nokuthunyelwe kwezokuxhumana, ulwazi olungangeni kahle emigqeni nasemakholomu njengespredishithi yendabuko. Amathuluzi avamile ayahluleka ukukhipha inani kokuqukethwe okungcolile, okuyilapho i-AI kanye nokufunda komshini kukhanya khona: angachaza izithombe, abhale inkulumo, aqonde imizwa yombhalo kanye namaphethini eqoqo kuwo wonke amasethi edatha angahlelekile.

Le tsunami yedatha ibalulekile ngoba iphakela ngokoqobo amamodeli e-AI anamuhla. Amamodeli amakhulu olimi, abakhi bezithombe, izinhlelo zokuncoma kanye nama-algorithm okubikezela konke kuncike eqoqweni elikhulu lezibonelo ukuze kufundwe amaphethini. Ngesikhathi esifanayo, ukuphatha, ukugcina nokuvikela lolu chungechunge lolwazi selube ngenye yezinselelo ezinkulu zobuchwepheshe kanye namasu ezinhlanganweni nasemiphakathini. Izinkinga ezifana nokuphathwa kwedatha, ubumfihlo, ukuphepha kwe-inthanethi kanye nengqalasizinda esimeme manje zixhumene kakhulu nekusasa le-AI.

I-AI kwezesayensi, ibhizinisi kanye nezimboni zokudala

I-AI ayisagcini eminyangweni yesayensi yamakhompyutha noma ezikhulwini zobuchwepheshe. Namuhla kuyindlela ejwayelekile emikhakheni eyahlukahlukene njengezokwelapha, ifiziksi, umthetho, imfundo kanye nesayensi yesimo sezulu. Abacwaningi abangebona abahleli bezinhlelo ngokuqeqeshwa basengakwazi ukwenza izivivinyo nge-AI ngenxa yamathuluzi asezingeni eliphezulu kanye namapulatifomu anobungani nokusebenziseka kalula, futhi bashicilela njalo umsebenzi lapho i-AI idlala indima ebalulekile ekuhlaziyeni idatha noma ekukhiqizeni imibono.

Izinhlaka zomthombo ovulekile zinciphise kakhulu isithiyo sokungena. Imitapo yolwazi efana ne-PyTorch, i-TensorFlow kanye nezinhlelo zemvelo ezifana ne-Hugging Face zivumela abafundi, izifundiswa kanye nabathuthukisi ukuthi babonise amamodeli amasha, balungise akhona futhi babelane ngemiphumela ngokushesha. Lesi simo sokubambisana sisheshisa ukusungula izinto ezintsha, kodwa futhi kusho ukuthi amasu amasha asakazeka ngesivinini esikhulu emphakathini wonke.

Kwebhizinisi, ama-studio akhethekile kanye nezinkampani zokubonisana zigxila ekuguquleni idatha engakalungiswa ibe yizinqumo eziqondile. Izinkampani eziningi zakha isofthiwe eyenziwe ngokwezifiso, izinhlelo zokusebenza zangaphakathi kanye nama-ejenti e-AI aklanyelwe ukwenza ngokuzenzakalelayo izinqubo eziphindaphindwayo, akhiphe ukuqonda kolwazi olungahlelekile futhi asekele ukwenza izinqumo. Bahlanganisa i-AI namathuluzi obuhlakani bebhizinisi njenge-Power BI, ingqalasizinda yamafu eqinile ku-AWS noma i-Azure kanye nemikhuba eqinile yokuphepha kwe-cyber, okuhlanganisa ukuhlolwa kokungena kanye nokwakhiwa kobumfihlo ngokuklama.

Umhlaba wokudala nawo uguqulwe yi-AI ngezindlela ezimangalisayo. Ama-algorithms manje angakhiqiza imidwebo, imidwebo, amavidiyo, izinkondlo kanye namathrekhi omculo aphelele. Izinhlelo ezifana ne-AICAN zikhiqize imisebenzi yobuciko eye yaboniswa kumagalari futhi yathengiswa ezindalini, kuyilapho amaphrojekthi afana ne-AIVA noma i-Musenet (phakathi kwabanye) ekwazi ukuqamba umculo wokuqala ngezinhlobo eziningi, adale izingcezu ezilingisa isitayela sababhali abadumile noma ahlanganise amathonya abe yinto entsha.

Esinye isenzakalo esiphawulekayo kwaba ukuthengiswa “kwesithombe sika-Edmond de Belamy” ngo-2018. Yadalwa ngemodeli yokukhiqiza eyasungulwa yiqembu laseFrance i-Obvious, umdwebo owenziwe nge-AI wathengiswa eChristie's ngamadola aseMelika angaphezu kuka-432,500. Lo mdwebo wavusa impikiswano enkulu mayelana nokubhala, ubuciko kanye nendima yama-algorithms kwezobuciko, futhi waphawula isikhathi esibalulekile ekuhlanganisweni kwe-AI kanye nemakethe yobuciko bendabuko.

I-AI ibhala ngisho nezikripthi futhi isize ekukhiqizeni amamuvi nochungechunge. Ngo-2016, ifilimu emfushane yokuhlola enesihloko esithi “Sunspring” yakhishwa ngeskripthi esenziwe ngokuphelele yi-AI eqeqeshwe ngezithombe zesayensi eziqanjiwe. Umphumela wawumangalisa kodwa uqondakala ngendlela exakile, okubonisa ukuthi ama-algorithms angafunda izakhiwo zokulandisa futhi akhiqize okuthile, yize kungaphelele, okuyindaba eqashelwayo. Ngesikhathi esifanayo, ama-studio asebenzisa i-AI ukuthuthukisa ikhwalithi yesithombe, ukuklama imiphumela ebonakalayo kanye nokusikisela izinqumo zokuhlela.

I-AI edumile iyahluleka kanye nokuziphatha okungajwayelekile

Naphezu kwazo zonke izimpumelelo zayo, i-AI inohlu olude lokwehluleka okubangela amahloni futhi ngezinye izikhathi okukhathazayo. Esinye sezigameko ezidume kakhulu sasihilela i-bot ye-Twitter ye-Microsoft i-Tay, eyasungulwa ngo-2016 njengomzamo wokusebenzisa i-AI yokuxoxa. I-Tay yaklanywa ukuze ifunde ekuxhumaneni nabasebenzisi ngesikhathi sangempela, ivumelanise ulimi lwayo kanye nobuntu bayo ngokusekelwe kulokho eyayikubona epulatifomu.

Phakathi namahora ambalwa, izinto zahamba kabi. Abanye abasebenzisi baqala ukugxeka uTay ngemiyalezo ecasulayo neyingozi. Ngenxa yokuthi i-bot yafunda ngokungazi kokufakwayo eyakuthola, yaqala ukukhiqiza nokukhulisa lolo hlobo lokuqukethwe, ithumela ama-tweets angafanele futhi ahlambalazayo. I-Microsoft yavala ngokushesha ukuhlolwa futhi yaxolisa emphakathini, isebenzisa lesi sigameko njengesifundo esinamandla ngesidingo sokuvikela okuqinile kanye nezihlungi zokuqukethwe lapho kusetshenziswa izinhlelo ze-AI ezindaweni ezivulekile, ezingalawulwa.

Elinye icala elixoxwa kabanzi lalihilela ama-chatbot athuthukiswe abacwaningi be-Facebook. Ngo-2017, kwasungulwa ama-bot amabili okuxoxisana ukuze ahwebelane, futhi ngesinye isikhathi aqala ukusebenzisa ulimi olufushane olwalungaqondakali kalula ngabantu. Ama-bot empeleni ayethuthukise ukuxhumana kwawo ukuze kube nokusebenza kahle kunokucaca, esungula amaphethini ayeqondakala kuwo kuphela. Nakuba lokhu kwakungeyona isimo "se-AI engalungile", kwaqokomisa indlela izinhlelo zokufunda komshini ezingaguqula ngayo ukuziphatha okungacacile futhi okunzima ukukuhumusha.

Amamodeli amakhulu olimi akhiqiza umbhalo anokunambitheka kwawo kokwehluleka: angakwazi "ukubhula". Ngenxa yokuthi lezi zinhlelo zibikezela ukulandelana kwamagama okungenzeka kakhulu ngokusekelwe kumaphethini kudatha yazo yokuqeqesha, ngezinye izikhathi zakha imininingwane ezwakala inengqondo kodwa empeleni ingamanga. Ngokombono wezibalo, imodeli imane ikhetha ukuqhubeka okungenzeka kakhulu; ngokombono womsebenzisi, ingaba nephutha ngokuqiniseka ngezindlela okunzima ukuzithola ngaphandle kokuqinisekiswa kwangaphandle.

Isibonelo esiqondile sihilela ukubhebhethekiswa kabi kwabalingiswa abadumile bezincwadi zamahlaya. Ngemva kokushona komdwebi wezithombe waseSpain uFrancisco Ibáñez, ezinye izihloko ezikhiqizwe noma ezisizwa kakhulu yi-AI zathi wadala kokubili i-“Mortadelo y Filemón” kanye ne-“Zipi y Zape”. Eqinisweni, u-Ibáñez wadala i-“Mortadelo y Filemón”, kodwa i-“Zipi y Zape” yadalwa ngomunye umbhali, uJosé Escobar. Ngenxa yokuthi laba balingiswa bavame ukukhulunywa ngabo ndawonye emibhalweni eku-inthanethi mayelana namahlaya aseSpain, kubonakala sengathi i-AI yaphetha ngokuthi amathuba okuba babe nomdali ofanayo ayephezulu kunamathuba okuthi bangahlobene, futhi yamane yasungula ubudlelwano.

Iziqephu ezinjengalezi zinciphisa ukwethembana kokuqukethwe okuzenzakalelayo futhi zisikhumbuza iphuzu elibalulekile. I-AI “ayazi” amaqiniso ngomqondo womuntu; ikhombisa amaphethini ezibalo kudatha. Uma lawo maphethini edukisayo noma engaphelele, uhlelo lungakhipha ulwazi olungalungile ngokuzethemba. Ngenxa yalokho, ukuthembela ku-AI ukukhiqiza okuqukethwe okuyiqiniso ngaphandle kokuqinisekiswa komuntu kuyingozi, ikakhulukazi ezindaweni ezibucayi njengezindaba, ezempilo noma umthetho.

Indlela yokucela usizo lwe-AI ngaphandle kokunikeza ubuchopho bakho umsebenzi othile

Njengoba kunikezwe le mikhawulo, indlela ehlakaniphile yokusebenzisa i-AI namuhla iwukuba umsizi onamandla, hhayi njengegunya elingangabazeki. Ungacela iseluleko, imibono noma uhlaka lokuqala lwento ethile ohlelweni lwe-AI, kodwa akufanele uphathe umkhiqizo wayo njengomkhiqizo wokugcina, ikakhulukazi lapho ukunemba kubalulekile. Ukubuyekezwa kwabantu kanye nobuchwepheshe besizinda kusalokhu kuyizinyathelo ezibalulekile kunoma yikuphi ukuhamba komsebenzi okungathi sína.

Uma ubhala imisebenzi, isu elihle ukucela iziphakamiso kunokucela izingcezu eziqediwe. I-AI ingakusiza ukuthi ucabangele ama-angles esihlokweni, uhlele izigaba, uphakamise izihloko ezikhangayo noma udale imisho ehlukile. Uma usunayo leyo nto eluhlaza, ungahlola amaqiniso, uthuthukise isitayela futhi uhlanganise ulwazi lwakho, uqinisekise ukuthi umphumela wokugcina ufanelana nezwi lakho futhi uhlangabezana nezindinganiso zakho zekhwalithi.

Uma kukhulunywa ngezithombe noma umklamo, amathuluzi e-AI angaba amahle kakhulu ekuklanyweni okusheshayo kanye nokuhlolwa okudlalwayo. Ungase udale amakhasi okufaka imibala kubantwana bakho, imidwebo yemibono yephrojekthi noma ukuhlukahluka okubonakalayo kokuthunyelwe ezinkundleni zokuxhumana. Kodwa-ke, uma udinga okuthile okucwebezelisiwe komkhankaso wokumaketha noma uphawu lochwepheshe, kuwukuhlakanipha ukubukeza yonke imininingwane, futhi uma kudingeka, unikeze lo msebenzi kumklami ongumuntu ukuze alungise amaphutha futhi avumelanise izithombe nobunikazi bakho nemigomo yakho.

Isimiso esifanayo siyasebenza nasezinsizeni zokuhlela uhambo ezisekelwe ku-AI kanye nezincomo. Vumela i-AI iphakamise izindawo, izinkambo noma amahhotela, kodwa njalo hlola kabili ukuthi izindawo zikhona ngempela, ukuthi zisavuliwe nokuthi ulwazi mayelana nendawo kanye nezinsiza lusesikhathini. Idatha eku-inthanethi iyashintsha njalo, futhi izinhlelo ze-AI ngezinye izikhathi zingathembela emithonjeni ephelelwe yisikhathi noma engalungile lapho zenza iziphakamiso.

Ekugcineni, i-AI inamandla amakhulu kodwa ayiyona into engenaphutha, futhi akunakwenzeka ukuthi ingabi naphutha maduze. Esikhundleni sokulindela ukuphelela, kunengqondo kakhulu ukuphatha i-AI njengethuluzi elingandisa amakhono akho. Uma uhlanganisa lokho okwenziwa yimishini kahle (isivinini, isikali, ukutholwa kwephethini) nalokho abantu abakwenzayo kangcono (ukwahlulela, uzwela, umongo), uthola imiphumela ethembeke kakhulu neyokudala.

Njengoba i-AI iqhubeka nokushintsha, umugqa phakathi kwelukuluku kanye nesidingo sansuku zonke uya uncipha. Kusukela emidlalweni yasendulo ne-chess kanye ne-Go champions kuya kuma-bots aziphatha kabi ezinkundleni zokuxhumana, kusukela kuma-algorithms angabonakali akha ngokuthula lokho esikubukayo nesikuthengayo kuya kumasethi edatha amakhulu akhula ngama-zettabytes minyaka yonke, ubuhlakani bokwenziwa sebuyinto echazayo yesikhathi sedijithali. Ukuqonda kokubili impumelelo yaso emangalisayo kanye namaphutha aso ngezinye izikhathi angenangqondo kusisiza ukuba sibhekane nayo ngenhlanganisela enempilo yentshiseko kanye nokucabanga okubucayi, futhi lokho kulinganisela yilokho kanye esikudingayo ukuze sisebenzise kahle lobu buchwepheshe ngaphandle kokubuvumela ukuthi buthathe indawo yokwahlulela kwethu.