3,124 ukufunda

Amaphupha kwiNyaniso-Ibali le-AI Evolution

nge Neha Pant11m2024/09/11

FA-AF

Inde kakhulu; Ukufunda

Ukuvela kwe-AI kwaqala kwiminyaka eyi-100 eyadlulayo kumaphupha kunye nokuqonda okuhlangeneyo kwabantu. Eli nqaku lixoxa ngebali lokuzivelela kwe-AI kwasekuqaleni.

featured image - Amaphupha kwiNyaniso-Ibali le-AI Evolution

Ngelixa i-AI idale isiphithiphithi kule minyaka imbalwa idlulileyo, isiseko sophuhliso lwe-AI sabekwa kwasekuqaleni kwenkulungwane yama-20.

Ukuzivelela kwe-AI

Nge-hype yamva nje ejikeleze i-AI, abanye bethu banokuziva ngathi yinto yamva nje. Ayiyo. Konke kwaqala ekuqaleni kwenkulungwane yama-20. Nangona kunjalo, umsebenzi omkhulu kweli candelo waqala ukwenzeka malunga nenkulungwane yama-20. Kwakumalunga neli xesha apho ingcali yezibalo, uAlan Turing, yakhawula ukuyilwa koomatshini abakrelekrele—iphupha eliye laphunyezwa ngokufika kobugcisa obunjengokufunda ngoomatshini kunye nokusetyenzwa kolwimi lwendalo.

Eli bali lendaleko leminyaka engaphezu kwe-100 libangela umdla. Kufana nomthi wombona ohlwayelwe emasimini, unkcenkceshelwa kwaye wondliwe, uze ulindele ukuba ukhule ube ngumthi omkhulu wom-oki. Ukuqonda ukuba isiseko sangaphambili savela njani kwitekhnoloji ekrelekrele esiyibonayo namhlanje, masingene kwimbali yobukrelekrele bokwenziwa.

Ekuqaleni kweminyaka yoo-1900

Ngaphambi kokuba nayiphi na ingcamango okanye ingcamango yomdla wesayensi iphononongwe zizazinzulu, izalwa kwiingqondo zamagcisa. Kukwanjalo nangengcamango “yabantu bokwenziwa” okanye “iirobhothi” ezathi zakhula engqondweni yombhali wemidlalo yeqonga waseCzech uKarel Capek zaza zafikelela kuvuthondaba ngomdlalo wakhe owabetha ngo-1920 othi “RUR - Rossum’s Universal Robots.” Ubongwa ngokuyila igama elithi “irobhothi” elithetha abasebenzi.

Siyazingca ngeemoto zikaGoogle eziziqhubayo kodwa ngaba siyazi ukuba umbono weemoto ezingaqhubiyo waqalwa ngo-1925? Iimoto zokuqala ezingaqhubiyo ezibetha izitrato zesiXeko saseNew York yayiziimoto ezilawulwa ngonomathotholo ezikhutshwe yiHoudina Radio Control. Kwathatha iminyaka engama-90 ukuba i-Alphabet Inc., inkampani yabazali bakaGoogle, iqalise iWaymo, ngoku ibonelela ngeenkonzo zerobhothi zentengiso-ummangaliso wobuchwephesha obukrelekrele.

Nje ukuba uluvo luqale kwindawo enye yehlabathi, luhamba ngokukhawuleza kwaye luzuze imilinganiselo emitsha. Ngowe-1929, unjingalwazi waseJapan uMakoto Nishimura wenza ngempumelelo into uKarel Capek awayeyicinga kumdlalo wakhe—irobhothi! Lo mmangaliso waseJapan ubizwa ngokuba yiGakutensoku, othetha ukufunda kwimithetho yendalo, unokushukumisa intloko, nezandla, nokutshintsha inkangeleko yobuso. UNishimura wayifumana ngokusebenzisa indlela yoxinzelelo lomoya.

Kubonakala ngathi umjelo wokuqonda woluntu wawuphupha ngokudibeneyo malunga noomatshini abangakwaziyo ukuxelisa umsebenzi womntu kodwa benze izinto abantu abangakwaziyo ukuzenza. U-HG Wells, umbhali wekamva lekamva oye waqulunqa iingcamango ezifana nokuhamba kwexesha, ukungabonakali, ubunjineli bezinto eziphilayo, kunye nokunye, waxela kwangaphambili ngo-1937 ukuba "yonke inkumbulo yomntu ingaba, kwaye mhlawumbi ngexesha elifutshane iya kwenziwa ifikeleleke kumntu wonke" kwaye “Nawuphi na umfundi, nakweyiphi na indawo yehlabathi, uya kukwazi ukuhlala neprojektha yakhe [yemifanekiso-bhanyabhanya] kwisifundo sakhe, ngexesha elifanelekileyo, ukuze ahlolisise nayiphi na incwadi okanye uxwebhu olukwikopi.” Sinokuqiniseka ukuba wayephupha ngeekhompyutha zexesha elizayo esizisebenzisayo namhlanje.

Kamva, kwi-1949, isazi sekhompyutha saseMelika, u-Edmund Callis Berkley, wapapasha incwadi ethi "Ubuchopho obukhulu okanye oomatshini abacingayo". Esi yayisisihlandlo sokuqala ukuba umfuziselo wento ekunokuthiwa yikhompyuter yokuqala, uSimon, uchazwe kwincwadi. Eyona nxalenye inomdla kwincwadi yesayensi kunye ne-aficionados yetekhnoloji yayiluphando lwengqondo yobuvulindlela (iikhompyuter zangaphambili) zexesha-i-MIT's differential analyzer, i-Harvard's IBM's IBM's sequence-controlled calculator, i-ENIAC ye-Moore School, kunye ne-Bell Laboratories's relay calculator. .

I-1950-ixesha apho i-AI yazalwa

Ngelixa ekuqaleni kweminyaka yoo-1900 yayilixesha apho abantu babesenza oomatshini abakrelekrele ezingqondweni zabo okanye bethatha amanyathelo amancinane ekudaleni oomatshini bokwenyani, ngeminyaka yoo-1950 kwakuxa kwakuthathwa amanyathelo okwenene kweli cala. Kude kube ngoku, lelona xesha likhuthazayo kwimbali ye-AI.

Kufuneka ube ngumntwana entliziyweni ukuba unqwenela ukucinga ngokukhululekileyo nangokuyilayo njengomntwana. Ukudlala imidlalo yenye yeendlela ezilungileyo zokugcina umntwana ephila. Kodwa u-Alan Turing uye wahambela phambili xa wayeyila “Umdlalo Wokulinganisa”, owaziwa kakhulu njengoVavanyo lweTuring, kwiphepha lakhe lesibini lika-1950 elithi “Computer Machinery and Intelligence”. Lo mdlalo wawuyilelwe ukuvavanya indlela yokuziphatha komatshini eyenze kwanzima ukwahlula emntwini. Kumaxesha anamhlanje, sisebenzisa umva woVavanyo lweTuring, olubizwa ngokuba “luvavanyo lweTuring lukawonke-wonke oluzenzekelayo ukuxelela iiKhompyutha kunye noBantu ngokwahlukeneyo” okanye iCAPTCHA ukumisela ukuba umntu hayi ibhot esebenza kumatshini.

Kwakungowe-1948 apho uTuring waqala ukubhala inkqubo yekhompyuter yokudlala umdlalo wechess. Ngomnyaka we-1952, wasebenzisa le nkqubo kwi-Ferranti Mark 1. Ngelishwa, ikhompyutha ayikwazanga ukusebenzisa inkqubo yokudlala umdlalo, kodwa i-Turing isebenze umntwana wakhe wangaphakathi ukuba adlale umdlalo we-chess usebenzisa imiyalelo kwi-algorithm manual. Emva kwexesha elide, umkhulu wechess waseRashiya owayesakuba yiNtshatsheli yeChess yehlabathi, uGarry Kasparov, wajonga umdlalo orekhodiweyo waza wawuchaza “njengomdlalo owaziwayo wechess.”

Kodwa umdla ngemidlalo awuzange uphelele eTuring. Omnye uvulindlela kumdlalo wekhompyutha nobukrelekrele bokwenziwa, uArthur Samuel, wazisa inkqubo yehlabathi yokuqala enempumelelo yokuzifundela iitshekhi ebizwa ngokuba yi-“Samuel Checkers-Playing Programme” ngowe-1952. kulula ukuvavanya ukusebenza kwekhompyuter ngokuchasene neyomntu. Ubukrelekrele bukaSamuel bubonakaliswe ngakumbi xa wayephakamisa igama elithi “ukufunda ngomatshini” ngo-1959, uphando awayeluqale ngo-1949.

Ixesha laphakathi kwiminyaka yee-1950 laliphithizela luphando kunye nokusebenza malunga nobukrelekrele bokwenziwa. Ngapha koko, igama elithi “artificial intelligence” okanye “AI”, njengoko libizwa njalo, lathi layilwa ngo-1955 yenye ingqondo eqaqambileyo ekwangomnye wabaseki be-AI njengoqeqesho—uJohn McCarthy. Waqulunqa eli gama kuxwebhu olubhalwe ngokubambisana kodwa eli gama lafumana ukuthandwa kwi-workshop yakhe yasehlotyeni kwiKholeji yaseDartmouth ekuthiwa yayizinyaswe ziingqondo ezikhokelayo zekhompyutha zelo xesha. Akazange aphelele apho njengoko eqhubeka ephucula iingcamango zakhe kwi-AI ngokuyila i-Lisp, ulwimi lwenkqubo, ngo-1958. Waqhubeka nokupapasha "iiNkqubo ezinengqiqo eqhelekileyo" kwi-1959 apho wachaza inkqubo yokusombulula iingxaki ngokulawula izivakalisi. —yayibizwa ngokuba nguMcebisi.

Zonke ezi zinto ziqanjiweyo kunye nezinto ezifunyanisiweyo zeminyaka yee-1950 zalandelwa ngomnye umsebenzi oqhekezayo ngeminyaka yoo-1960 kunye noo-70. Eyona nto yayiphambili kwezi yayiyinkqubo yonxibelelwano ELIZA. Iphuhliswe nguJoseph Weizenbaum ngo-1965, le nkqubo inokuqhubeka nengxoxo ngaso nasiphi na isihloko ngolwimi lwesiNgesi, olufana nee-chatbots zanamhlanje! Eyona nto yayinomdla kakhulu ngale nkqubo yayikukuba abantu abaninzi bathi yabangelwa kukuba neemvakalelo ezifana nezomntu—nto leyo engekacaci kwaye kusaxoxwa ngayo.

Indlela ethi "Isifundo esinzulu" esisetyenziswe ekuphuhliseni i-AI namhlanje, yakhawulwa kwi-1968 yi-mathematician yaseSoviet u-Alexey Ivakhnenko emsebenzini wakhe othi "Indlela yeQela lokuPhathwa kweDatha" eyapapashwa kwiphephancwadi elithi "Avtomatika". Isenokuvakala ingakholeleki kodwa ukuhambela phambili kobugcisa obukhawulezayo esikubonayo namhlanje kume emagxeni omsebenzi ocothayo nozinzileyo owenziwe phakathi kwenkulungwane yama-20.

Ngeminyaka yee-1980—Ixesha Lokunyuswa Kwesenzo

\Kodwa izinto azizange zihlale zicotha kwaye zizinzile. Baye bathatha umphunga kwiminyaka yee-1980 xa umdla kwi-AI kunye nokuxhaswa ngemali kunye nophando kwintsimi yakhula ngokuxhuma kunye nemida. Eli xesha libone uphuhliso lweenkqubo eziphindaphinda amandla okwenza izigqibo kwiingcali zabantu kwiinkalo ezithile.

I-1980 yayingunyaka wenkomfa yokuqala yoMbutho woPhuculo lwe-Artificial Intelligence, okanye i-AAAI, eyasungulwa kwi-1979. Ngenkomfa yonyaka ye-38 ebanjwe kulo nyaka, lo mbutho uyaqhubeka nokukhuthaza uphando kwintsimi ye-AI. kunye notshintshiselwano ngezimvo zenzululwazi phakathi kwabasebenzi behlabathi endle. Kodwa kwakukho ixesha kwi-1984 xa i-AAAI yayibikezele oko kwakubizwa ngokuba yi " AI Winter " , ixesha lophando olunobuvila ngenxa yokunciphisa umdla kwi-AI.

Noko ke, ngaphambi kwesi siprofeto, iSebe laseJapan loRhwebo lweZizwe ngezizwe noShishino labela ngokukhwankqisayo izigidi ezingama-850 zeerandi kwiProjekthi yeFifth Generation Computer Project ngowe-1981. Lo msebenzi weminyaka eli-10 wawunamabhongo, ngaphambi kwexesha lawo, nokusilela kwezorhwebo—umzekelo ogqibeleleyo. yokuba yi-hype endaweni ye-zeitgeist. Iprojekthi, nangona kunjalo, yenze umzekelo wefilosofi yaseJapan ye-IKIGAI njengoko izazinzulu zixhaphaza i-zest yazo ukunika amandla kuphuhliso lwenkqubo yengqiqo efanayo.

Ngasekupheleni kwee-1980 kunye ne-Early 1990s-i-AI Winter?

Ngokuchaseneyo noqikelelo lwemozulu lwesebe lezemozulu, uqikelelo lwe-AAAI lobusika be-AI lungqineke luchanekile. Ukuphela kweprojekthi yesiZukulwana sesiHlanu kwakungomnye woonobangela bokulahlekelwa ngumdla kunye notyalo-mali kwinkalo yobukrelekrele bokwenziwa ekupheleni kweminyaka yoo-1980. Kodwa ezinye izithintelo kwiinkqubo zeengcali kunye neemarike zoomatshini, kubandakanywa ukuwa kwe-hardware esekwe kwi-LISP ngenxa yeendlela ezingabizi kakhulu ezivela kwi-IBM kunye ne-Apple ngo-1987, nazo zibe negalelo ekungakhathaliyo kwe-AI.

Nangona kunjalo, ayizizo zonke eziye zaphazamiseka ngeli xesha. Uyise wentatheli yaseMelika eyabulawayo uDaniel Pearl kunye nophumelele iMbasa yeTuring ka-2011, uNjingalwazi uJudea Pearl, wapapasha “Ukuqiqa okunokwenzeka kwiiNkqubo zoBukrelekrele” ngo-1988. Umcingi ogama iimodeli zaseBayesia zaba sisixhobo esibalulekileyo somsebenzi obalulekileyo kubunjineli kunye nesayensi yendalo.

I-Jabberwock inokuba ngumlinganiswa oyintsomi owoyikekayo ovela ku-Alice e-Wonderland kodwa injongo yayo yayifana ne-Jabberwacky, i-chatbot ka-1988 eyaphuhliswa ngu-Rollo Carpenter-ukonwaba kunye nokonwabisa. I-Jabberwacky yayixhotyiselwe ukulinganisa incoko yendalo yomntu ngokonwabisa nangokuhlekisa.

Ngowe-1993-2011—Ukunyuka kwengxilimbela eleleyo

Ubusika abude, ngakumbi kwimimandla ephantsi kwetropiki, kwaye njengoko ishiya, intwasahlobo ingenisa iintyatyambo ezintle ezithandwa ngumntu wonke. Kwenzeka into efanayo nakubusika be-AI. Ngelixa ingqikelelo eyoyikisayo yokuphela kwexesha lomntu ngo-1993 "I-Coming Technological Singularity" ngu-Vernor Vinge inokuba isoyikisa eyona nto ingcono kuthi, eli phepha likwaxele kwangaphambili ukuba kwiminyaka engama-30 "siya kuba neteknoloji yokwenza ubukrelekrele bomntu. ” Iminyaka engamashumi amathathu ukusukela oko sisenokuba khange siyifezekise le nto wayeyixele kwangaphambili, kodwa kubonakala ngathi sijonge kwelo cala.

Kwi-1997, ihlabathi labona ukunqotshwa kukaGary Kasparov, iqhawe lehlabathi le-chess elilawulayo, kwi-Deep Blue, inkqubo yokuqala yokudlala i-computer chess. Esi siganeko sasingumzuzu wamanzi kwimbali ye-AI kwaye ngoko yaba yimpembelelo yokudala kwiincwadi ezininzi kunye nefilimu. Kwangalo nyaka, i-Dragon yakhupha i-NaturalSpeaking 1.0, ekwabizwa ngokuba yi-DNS, isoftware yokuqaphela intetho eya kusebenza kwiWindows.

Unyaka wama-2000 wabona inkqubela engakumbi kumsebenzi wokwenziwa kweKismet, irobhothi enokulinganisa iimvakalelo zomntu. Oku kunye nezinye iirobhothi yayiyingqondo kaNjingalwazi Cynthia Breazeal owayesengumfundi kwi-MIT Artificial Intelligence Lab. Amaphupha awabonwa ngababhali bemidlalo yeqonga ekuqaleni kweminyaka yoo-1900 ayeqalisa ukubakho.

I-space odyssey ka-Yuri Gagarin kunye nokufika kwenyanga kaNeil Armstrong kwakufanelekile ukubhiyozelwa kodwa kwakunjalo nokufika kwe-Spirit ne-Opportunity, iirover ezimbini zase-US, kuMars ngo-2003. .

I-Twitter, i-Facebook, kunye ne-Netflix-i-tech giants ye-contemporary-yaqala ukusebenzisa i-AI kwiintengiso zabo kunye ne-UX algorithms phantse kwiminyaka engamashumi amabini edlulileyo kwi-2006. Iimpawu zokuba i-AI njengeteknoloji yayiza kukhaba ngokwenene yayiqalile ukubonakala. Ukuphumelela kwe-Deep Blue kwi-1997 yayiyinkcazo yophando lwe-IBM kwintsimi ye-AI kunye ne-IBM Watson yabonisa ukuba uphando lwabo lukhokelele kangakanani xa inkqubo yoyisa uBrad Rutter kunye noKen Jennings kwi-Jeopardy! Umngeni ngo-2011.

Yayiyi-IBM's Watson eyenza amashishini ukusuka kwiivenkile ukuya kwiinkonzo zemali zicinge malunga nokuthunyelwa kwe-AI kwishishini. I-2011 ikwangunyaka apho i-Apple's Siri, umncedisi wenyani, yasungulwa.

Ngo-2012 ukuza kuthi ga ngoku—Ukutsiba okukhulu kwi-AI

Ixesha liqhubekile kwaye yonke into esiyenzayo, nkqu ne-AI, yinxalenye yokuqhubeka. Siye sabona ukuba amanyathelo osana kwiminyaka yoo-1900 aye akhokelela ekufumaneni izinto ezinkulu kunye nokuveliswa kwe-2000s. Njengekota yokuqala yee-intshi ze-2000 ukuya ekupheleni kwayo, sibona ukuba i-AI ihambe kangakanani, ngakumbi kule minyaka ilishumi idlulileyo kunye nangaphezulu.

Olunye lwezi hambo lolo lwemodeli yeSifundo esiNzulu. Ukususela ekukhawulweni kwayo okuthobekileyo ngo-1968, imodeli yafumana imbonakalo enkulu xa uJeff Dean kunye noAndrew Ng bakaGoogle badibanisa i16,000 yeeprosesa zekhomputha ukuze benze enye yothungelwano lwe-neural olukhulu olungaphezu kwebhiliyoni lodibaniso lokufunda koomatshini ngo-2012. Bondla inethiwekhi 10 yezigidi. imifanekiso engahleliweyo yeekati ezivela kwiividiyo zeYouTube kwaye kwenzeka into engaqhelekanga-inethiwekhi yaqala ukuqaphela iikati. Abathandi beekati banokwenza ingxolo!

Olunye uhambo lolo lwerobhothi. Ukusuka kwiRobhothi zikaKarel Capek's Rossum's Universal Robots ukuya kwiGakutensoku kaMakoto Nishimura ukuya kwincoko kaJoseph Weizenbaum u-ELIZA ukuya eCynthia Breazeal's Kismet, irobhothi yayisele ihambile, kodwa kwakungonyaka ka-2016 apho uSophia, irobhothi ye-humanoid eneempawu zokuqala zobuntu kunye nokubonakalisa. "ummi werobhothi". Ndothuke kakhulu, musa ukoyika.

Ezinye izinto ziyanyikimisa isikhephe sethu kwaye ke ezinye ziyagungqa ubuntu. Enye into enje yenzeka ngo-2017 xa umhlaba wetekhnoloji washiyeka wothukile kukuziphatha kwee-chatbots ezimbini zikaFacebook ezazibonakala ngathi ziphuhlise owazo ulwimi lothethathethwano olungaqondakaliyo ebantwini kodwa zabonisa iipatheni ezithile zokuqinisekisa iinjineli ze-AI ukuba yayingeyonto nje engaqhelekanga kodwa. ulwimi ababeluqonda ngokwenene. Ngelixa oku kungakhange kubhengeze ukufika kobunye njengoko bekuxelwe kwangaphambili nguVernor Vinge, uthethathethwano lwebhothi, kuquka umdla wabo wokwenza into ukuze baqinisekise omnye ukuba bayavumelana emva koko, bashiye abantu benomsantsa kumathuba exesha elizayo koomatshini.

NgoMatshi ka-2020, kwafika iliza lokuqala lobhubhani we-Covid-19 owaphazamisa umsebenzi njengoko sasisazi kude kube ngoko. Njengoko iinkampani zaqala ukukhangela iindlela zokwenza umsebenzi kude, i-Open AI yasungula i-Generative Pre-trained Transfomer, okanye i-GPT-3, njengoko ibizwa ngokudumileyo, ngoMeyi 2020. I-GPT-3 yenye yeemodeli ezinkulu zokufunda ulwimi ezinokuthi yenza imisebenzi ngokuchaneka okuphakamileyo ngenxa yokwanda kwamandla kunye nenani eliphezulu leeparamitha-i-175 yeebhiliyoni ezili-10x ngaphezu kwemodeli ekhuphisanayo esondeleyo ye-Turing-NLG.

Ukususela ngoko, iinguqulelo ezintsha eziphuculweyo, i-GPT-3.5 kunye ne-GPT-4 ziye zaqaliswa. Kutshanje, i-GPT-5 iye yabhengezwa ithembisa i-AI ePhakamileyo, uvelwano, imfihlo, uhlengahlengiso oluguquguqukayo, kunye nekhonkco lokucinga kunye nokuqiqa inyathelo ngenyathelo.

Uhambo lwethu lusizise kule mihla, apho i-AI ingena kuzo zonke iinkalo zobomi bethu. Ukusuka kubancedisi benyani abalindele iimfuno zethu ukuya kwiimoto eziziqhubayo ezihamba kwizitrato zedolophu, amaphupha awayekhe amnandi oovulindlela be-AI abe yinyani yethu yemihla ngemihla.

Kodwa, njengoko siguqulela iphepha kwikamva, iyelenqe liya liqina ziingxaki zokuziphatha, ifuthe loluntu, kunye nokufuna ubukrelekrele jikelele bokwenziwa - owona mda wokugqibela. I-saga ye-AI iyaqhubeka ukutyhileka, ichukumisa ingcamango yethu kwaye icela umngeni ekuqondeni kwethu ukuba kuthetha ukuthini ukwenza ubukrelekrele kwi-silicon kunye nekhowudi.

Kwaye ke, ibali lobukrelekrele beArtificial, umdibaniso omangalisayo wobuchule bomntu kunye nendaleko yetekhnoloji, iqhubela phambili kwindawo engaziwayo, isimema sonke ukuba singqine izahluko ezilandelayo zale ngxelo ihlekisayo.

Isihlomelo: