Ii-5 eziQhelekileyo zoMbono woMshini

I. Intshayelelo kwiMbono yoMshini

Umbono womatshini, isiseko sobukrelekrele bokwenziwa, unika amandla iikhompyutha ukuba zitolike idatha ebonakalayo ngokuchanekileyo okufana nokomntu. Ngokungafaniyo nokucubungula imifanekiso yendabuko, isebenzisa ukufunda okunzulu kunye neenethiwekhi ze-neural ukuhlalutya imifanekiso kunye neevidiyo, iguqula amashishini ngokwenza imisebenzi enzima ngokuzenzekelayo. Ukusuka ekuboneni iziphene kwimigca yokuhlanganisa ukuya ekuxilongeni izifo, umbono wekhompyutha uguqula indlela esisebenza ngayo nesiphila ngayo. Amandla ayo okucubungula idatha yexesha langempela ayenza ibe yinto ebalulekileyo kwizicelo ezifana nokuchonga izinto, imifanekiso yezonyango, kunye nezithuthi ezizimeleyo.

Kutheni umbono womatshini ubaluleke kangaka? Uqhuba ukusebenza kakuhle kunye nokuchaneka apho amehlo omntu ehluleka khona. Umzekelo, kwimveliso, iinkqubo zokuhlola ezibonakalayo zibamba iziphene ezincinci, zisindisa izigidi zabantu abakhunjulwayo. Kwinkonzo yezempilo, imifanekiso yezonyango eqhutywa ziinethiwekhi ze-convolutional neural ibona umhlaza kwangethuba, iphucula iziphumo zesigulane. Okwangoku, izithuthi ezizimeleyo zixhomekeke ekulandeleni intshukumo kunye nokuqatshelwa kweepateni ukuze zihambe ngokukhuselekileyo, zinciphisa iingozi. Olu phuculo luvela kwiminyaka emininzi yophando, ukuxuba ukwahlulwahlulwa kwemifanekiso, ukuqatshelwa kobuso, kunye nokubona kwe-3D ukujongana nemingeni yehlabathi lokwenyani.

Eli nqaku liza kujonga kwiindlela eziphambili zokusebenzisa umbono womatshini, libonisa impembelelo yawo kumacandelo onke. Siza kuhlola indlela i-Industry 4.0 esebenzisa ngayo iirobhothi kunye nokuhlolwa okuzenzakalelayo, indlela abathengisi abaphucula ngayo uhlalutyo lokuziphatha kwabathengi, kunye nendlela ulawulo lwezithuthi oluphucula ngayo ngokuqatshelwa kweempawu zendlela. Lindela ukuqonda ngeendlela ezintsha ezifana ne-AI yokuvelisa kunye neemodeli zolwimi lombono, ezisekelwe kwimithombo efana neSayensi.gov kunye ne-IoT Analytics.

Injongo yethu? Ukufumanisa indlela umbono womatshini oguqula ngayo amashishini kwaye uvuselele ubuchule. Nokuba ungumkhokheli weshishini, umphandi, okanye umthandi wetekhnoloji, uza kufumana iimeko zokusetyenziswa ezisebenzayo kunye namathuba exesha elizayo.

II. Umbono woMshini kwiNtlalo

Umbono womatshini uguqula ukhathalelo lwempilo ngokuphucula ukuchaneka kokuxilonga kunye nokwenza lula ukhathalelo lwezigulane. Ngokusebenzisa umbono wekhompyutha kunye nokufunda okunzulu, iinkqubo zihlalutya imifanekiso yezonyango ngokuchanekileyo okungazange kubonwe ngaphambili, zibamba imiba enokuphoswa ngamehlo abantu. Ukususela ekufumaneni umhlaza okwinqanaba lokuqala ukuya ekujongeni ukubuya, uhlalutyo olubonakalayo oluqhutywa ziinethiwekhi ze-convolutional neural lusindisa ubomi.

Kwizifundo zokuxilonga, Ukucubungula umfanekiso kuyakhanya. Ii-algorithms ziskena ii-CT scans, ii-MRIs, kunye nee-X-rays ukuchonga izinto ezingaqhelekanga ezifana neethumba okanye iipateni ze-COVID-19 kwi-X-rays zesifuba. Umzekelo, izifundo ezivela kwi-PMC zibonisa iimodeli ze-AI ezifikelela ngaphezulu kwe-90% ngokuchanekileyo ekufumaneni umhlaza webele, zidlula iindlela zemveli. Esi santya kunye nokuthembeka kuthetha ukungenelela okukhawulezayo.

Iingenelo zotyando nazo. Inyani eyongeziweyo igubungela imifanekiso yezonyango yexesha langempela ngexesha leenkqubo, ikhokela oogqirha ngocoselelo olucacileyo. Izixhobo ezifana nokwahlulahlula imifanekiso zigxininisa iindawo ezibalulekileyo, zinciphisa iingozi kwiinkqubo ezingangenisi kakhulu. Okwangoku, ukulandelela intshukumo kujonga iintshukumo zezigulane kude, kunceda ukulandelela ukuchacha ngaphandle kokutyelela esibhedlele. Cinga ngovavanyo lokuphiliswa kwamanxeba ngee-apps ze-smartphone—ezilula kwaye ezingabizi kakhulu.

Impembelelo? Izibhedlele zinciphisa ukulibaziseka kokuxilongwa, kwaye izigulana zifumana unyango oluhambelanayo. Kodwa imingeni efana nobumfihlo bedatha isasele, ifuna iinkqubo ezikhuselekileyo zokukhusela iirekhodi eziyimfihlo.

 

III. Umbono woMshini kwiiNkqubo zoThutho eziNgcono

Umbono woomatshini uguqula ezothutho, wenza iindlela zikhuseleke kwaye zibe krelekrele ngokusebenzisa umbono wekhompyutha kunye nokufunda nzulu. Ngokusebenzisa iinkqubo zothutho ezikrelekrele, kuvumela ulawulo lwezithuthi ngexesha langempela, ukufunyanwa kwezinto, kunye nokuqhuba ngokuzimela, kutshintsha indlela esihamba ngayo.

Ukhuseleko lwendlela, ukucutshungulwa kwemifanekiso kuyagqwesa. Ii-algorithms ezifana ne-YOLO kunye ne-Faster R-CNN zibona izithuthi, abahambi ngeenyawo, kunye neempawu zendlela ngokuchanekileyo okuphezulu. Ukuqatshelwa kweempawu zendlela kuqinisekisa ukuba abaqhubi kunye neenkqubo ezizimeleyo bahlala benolwazi, ngelixa ukuqatshelwa kwezibane zendlela kuthintela ukungqubana. Izifundo ezivela kwi-PMC zibonisa iinkqubo ezifikelela kwi-95% yokuchaneka kwiindawo zasezidolophini, zinciphisa amazinga eengozi.

Izithuthi ezizimeleyo zixhomekeke kakhulu ekuboneni komatshini. Ukususela kwiNqanaba loku-1 loncedo lomqhubi ukuya kwiNqanaba lesi-5 ukuzimela ngokupheleleyo, ukulandelela intshukumo kunye nokuqondwa kweepateni zihamba kwiindawo ezintsonkothileyo. Iikhamera ezidityaniswe neenethiwekhi ze-neural eziguquguqukayo zichonga imiqobo ngexesha langempela, zivumela utshintsho olukhuselekileyo kwindlela okanye ukuma ngexesha likaxakeka. Iinkampani ezifana neTesla zisebenzisa umbono we-3D ukuze zihambe kakuhle.

Ukubeka iliso kwiindlela nako kuyanceda. Ukulandelela abahambi ngeenyawo kubonakalisa ukuhamba ngeenyawo, ngelixa ukubhaqwa kwezinto ezingaqhelekanga kubonakalisa iingozi okanye ubumdaka. Iinkqubo zerhafu ezikrelekrele zisebenzisa ukwahlulwahlulwa kwemifanekiso ukufunda iipleyiti, zilungelelanisa iintlawulo. Ezi zixhobo ziphucula ukuhamba kwamanzi, zinciphisa ukuxinana kwezixeko.

Imingeni isaqhubeka, njengokuqinisekisa ubumfihlo bedatha ekujongeni nasekuphatheni imozulu embi echaphazela uhlalutyo olubonakalayo. Sekunjalo, impembelelo ayinakuphikiswa—iindlela ezikhuselekileyo, ukuvalwa kwegrid kuncinci.

IV. Umbono woMshini kwiMveliso kunye noShishino 4.0

Umbono womatshini utshintsha umdlalo kwi-Industry 4.0, nto leyo eqhuba ukusebenza kakuhle kunye nokuchaneka kwemveliso. Ngokusebenzisa umbono wekhompyutha kunye nokufunda okunzulu, iifektri zifezekisa ulawulo lomgangatho olukrelekrele, ukulungiswa kwenkqubo, kunye nolawulo lwezinto ezisetyenzisiweyo, ukunciphisa iindleko kunye nokunyusa imveliso.

Uhlolo olubonakalayo lukhokela uhlaselo. Iinethiwekhi ze-neural eziguquguqukayo ziskena iimveliso ukuze zifumane iziphene ezifana nemikrwelo okanye ukungalungelelani, okugqitha kakhulu ukuchaneka komntu. Ngokungafaniyo neenkqubo ezisekelwe kwimithetho, ukucubungula imifanekiso okuqhutywa yi-AI kuyahambelana neziphene ezahlukeneyo, kubambe iingxaki kwangethuba. I-IoT Analytics ibika ukuba ulawulo lomgangatho oluzenzekelayo lugcina izityalo ukuya kuthi ga kwi-$172M ngonyaka ngokunciphisa ukubuyiswa kwakhona.

Ukulungiswa kwenkqubo yenye impumelelo. Uhlalutyo lwexesha langempela lujonga izixhobo, lubona ukuguguleka okanye ukungasebenzi kakuhle ngokusebenzisa ukwahlulwahlulwa kwemifanekiso. Ukugcinwa kwangaphambili, njengokuchongwa kokugqwala, kuthintela ukuqhekeka, kwandisa ubomi bomatshini. Le ndlela iqhutywa yidatha igcina imveliso ivakala imnandi, inciphisa ixesha lokungasebenzi.

Kwiindawo zokugcina impahla, umbono womatshini wenza kube lula ukuhanjiswa kwezinto. Ukulandelela impahla ngokuchonga izinto kuqinisekisa ukubalwa kwezinto ngokuchanekileyo, ngelixa i-OCR ibhala iibhakhowudi ukuze zihlelwe kakuhle. Iidrone ezixhotyiswe ngombono we-3D, ezisetyenziswa ziingxilimbela ezifana neAmazon, ziskena iishelufu ngexesha langempela, zinciphisa umsebenzi wezandla. Ezi zixhobo ziphucula ukusebenza kakuhle kwekhonkco lokubonelela, zihlangabezana neemfuno ngokukhawuleza.

Imingeni efana neendleko eziphezulu zokuseta kunye nobumfihlo bedatha kufuneka ijongwe, kodwa iingenelo zicacile: imisebenzi engenamsebenzi kunye nemida yokhuphiswano.

V.Imbono yoMshini kwicandelo leZicelo zoThengiso kunye neZicelo zaBathengi

Umbono womatshini uguqula ukuthengisa, udibanisa umbono wekhompyutha kunye nokufunda okunzulu ukuze kuphuculwe amava abathengi kwaye kuphuculwe imisebenzi. Ukusuka ekuthengeni okwenziwe ngokwezifiso ukuya kwiivenkile ezikrelekrele, uhlalutyo lokuthengisa oluxhaswa kukucubungula imifanekiso luphinda luchaze ishishini.

Kubathengi, ukubonwa kobuso kubonelela ngamava alungiselelwe wena. Iikhamera ezikwivenkile zihlalutya indlela abathengi abaziphethe ngayo, zivumela ukukhushulwa okujoliswe kuko. Iinkqubo zokuvavanya ezibonakalayo, zisebenzisa ukwahlulwahlulwa kwemifanekiso, zivumela abathengi ukuba bavavanye izinto zokuthambisa okanye iiglasi ngedijithali, nto leyo ekhuthaza ukuthengiswa. Izixhobo zohlalutyo lolusu, ezifana nezo zivela eL'Oréal, zisebenzisa uhlalutyo olubonakalayo ukucebisa iimveliso, nto leyo ephucula ulwaneliseko.

Imisebenzi yevenkile ifumana ukusebenza kakuhle ngokubona komatshini. Iimephu zobushushu ezivela ekulandeleni intshukumo zibonisa iindawo ezinabantu abaninzi, ziphucula uyilo. Ukuchongwa kobusela ngeenkqubo zokujonga kunciphisa ilahleko, kunye ne-AI ebona iintshukumo ezisolisayo ngexesha langempela. Ingxelo ye-IoT Analytics ka-2023 ithi abathengisi bonga i-15% ekuncipheni besebenzisa ukufunyanwa kwezinto.

Ulawulo lwezinto ezikhoyo luyabonakala. Ukuqatshelwa kwemifanekiso kulandelela amanqanaba esitokhwe, kunciphisa ukugqithiswa kwezinto. I-OCR iskena iilebheli, ikhawulezisa ukuphinda kufakwe izinto. Ezi zixhobo ziqinisa uthungelwano lokubonelela, ziqinisekisa ukuba iimveliso ziyahlangabezana neemfuno. Iinkqubo zeWalmart ezisekelwe kumbono, umzekelo, zivumelanisa iishelufu nee-odolo ze-intanethi ngokulula.

Imingeni efana nobumfihlo bedatha ifuna ukhuseleko oluqinileyo, kodwa imbuyekezo icacile: abathengi abonwabileyo, imisebenzi engenamsebenzi.

VI. Iindlela Ezivelayo Nezikhokelo Zexesha Elizayo

Umbono woomatshini utshintsha ngokukhawuleza, ukhuthazwa kukuphumelela kobukrelekrele bokwenziwa kunye nophuhliso lwezixhobo. Ezi ndlela zithembisa ukwandisa usetyenziso, ukusuka kwi-robotics ekrelekrele ukuya kwi-computer enobuntu, nto leyo eyakha ikamva eliguqukayo.

Ubuchule bokufunda nzulu bukhokela indlela. Uyilo lwe-transformer, njengaleyo ikwiimodeli zolwimi olubonakalayo, ludibanisa umbhalo kunye nemifanekiso kwimisebenzi efana nokuphendula imibuzo ebonakalayo. I-AI evelisayo, kuquka iimodeli zokusasazwa kunye nee-GAN, idala idatha yokwenziwa, inciphisa ukunqongophala kweseti yedatha. Uphononongo lweSpringer luka-2023 lubonisa ukuba ezi modeli zinciphisa iindleko zoqeqesho ngama-20%, zinyusa ukusebenza kakuhle kokucubungula imifanekiso.

Uphuhliso lwezixhobo luphucula ubuchule. Iikhamera zokubona umatshini ezincinci, ezifanelekileyo kwimifanekiso yesayensi, zibonelela ngedatha enesisombululo esiphezulu kwiifom ezincinci. Ukusetyenziswa kwe-FPGA kukhawulezisa uhlalutyo lwexesha langempela, olubalulekileyo kwizithuthi ezizimeleyo kunye ne-Industry 4.0. Ii-chipsets ezintsha ziphucula iinethiwekhi ze-neural eziguquguqukayo, zenza iinkqubo zikhawuleze kwaye zingabizi kakhulu.

Imingeni yokuziphatha iyanda. Iingxaki zobumfihlo bedatha, ingakumbi ekubonweni kobuso, zifuna imithetho engqongqo. Iinkqubo zokuqinisekisa i-biometric kufuneka zilinganise ukhuseleko kunye nokuthenjwa komsebenzisi. Okwangoku, ukwenziwa kwezinto ngokuzenzekelayo kuyingozi yokususa imisebenzi engenazakhono zininzi, nto leyo edinga ukuqeqeshwa kwakhona kwabasebenzi, njengoko i-IoT Analytics igxininisa.

Ezi ndlela zibonisa ikamva apho umbono womatshini udibana khona nobomi bemihla ngemihla, ukusuka kwimifanekiso yezonyango ukuya kulawulo lwezithuthi.

VII. Isiphelo

Umbono womatshini uvele njengesixhobo esinamandla, esiqhuba utshintsho kushishino kulo lonke ukhathalelo lwempilo, ezothutho, imveliso, kunye nokuthengisa. Ngokusebenzisa umbono wekhompyutha, ukufunda nzulu, kunye nokucubungula imifanekiso, inika izisombululo zombono womatshini ezikhuthaza ukusebenza kakuhle kunye nokuchaneka. Kwinkonzo yezempilo, imifanekiso yezonyango iphucula ukuchaneka kokuxilonga, isindisa ubomi. Izithuthi ezizimeleyo zixhomekeke ekufumaneni izinto ukuze zikhuseleke kwiindlela, ngelixa i-Industry 4.0 isebenzisa ukuhlolwa okubonakalayo ukunciphisa iindleko. I-Retail isebenzisa ukuqatshelwa kobuso kunye nohlalutyo lokuthengisa ukuphakamisa amava abathengi, nto leyo ebonisa ukuba ubuchwepheshe busebenza ngeendlela ezahlukeneyo.

Oku akupheleli nje ekuthandeni ubuchwepheshe—yimpembelelo yokwenyani. Iinethiwekhi ze-neural eziguquguqukayo kunye neemodeli zolwimi lombono zisombulula iingxaki ezinzima, ukusuka kulawulo lwethrafikhi ukuya ekulandeleleni izinto ezikhoyo. Sekunjalo, imingeni efana nobumfihlo bedatha kunye nokufuduka kwemisebenzi kufuna iindlela zokuziphatha ze-AI. Ubuchule bokuvelisa izinto ezintsha obunoxanduva buqinisekisa ukuba umbono womatshini uyanceda wonke umntu, ulungelelanisa inkqubela phambili kunye nokuthembana.

Ikamva liqaqambile kodwa lifuna isenzo. Amashishini kufuneka ahlolisise uhlalutyo lwexesha langempela kunye ne-AI evelisayo ukuze ahlale ekhuphisana. Abaphandi banokutyhala imida ngezakhiwo ze-transformer, ngelixa abenzi bomgaqo-nkqubo kufuneka babeke phambili ubumfihlo bedatha. Njengoko i-IoT Analytics iphawula, iinkampani ezisebenzisa izisombululo zombono womatshini zibona inzuzo yokusebenza efikelela kwi-25%, isizathu esibalulekileyo sokwenza into.

Ukusasazwa, ukusebenzisana nabantu abanamava abavelisi beekhompyutha ezifakiweyo kubalulekile, ingakumbi xa kuphunyezwa izisombululo ezifana I-AFE-R770 kwizicelo ezikumgangatho wemizi-mveliso. ikhompyutha efakwe kwimizi-mveliso iyakwazi ukujongana neemeko ezinzima, inika ukuthembeka nokusebenza rhoqo. Izixhobo ezifana ne I-Mini PC Industrial J1900 zilungele ukucutshungulwa komphetho, ngelixa zisebenza kakuhle I-PC yebhokisi efakiweyo ingakwazi ukunika amandla imithwalo yemisebenzi ye-AI ephucukileyo kweli candelo.