Iyini Indima Ye-Big Tech ku-AI?

Iyini Indima Ye-Big Tech ku-AI?

Impendulo emfushane: I-Big Tech ibalulekile ku-AI ngoba ilawula izinto ezibalulekile ezingadumile - ukubala, amapulatifomu amafu, amadivayisi, izitolo zezinhlelo zokusebenza, kanye namathuluzi ebhizinisi. Lokho kulawula kuyivumela ukuthi ifake imali kumamodeli omngcele futhi ithumele izici ezigidini zabantu ngokushesha. Uma ukubusa, ukulawula ubumfihlo, kanye nokusebenzisana kubuthakathaka, amandla afanayo aguquka abe ukuvalelwa ngaphakathi kanye nokugxila kwamandla.

Izinto ezibalulekile okufanele uzicabangele:

Ingqalasizinda: Phatha ukulawulwa kwamafu, ama-chip, nama-MLOp njenge-chokepoint eyinhloko ye-AI.

Ukusatshalaliswa: Lindela izibuyekezo zeplatifomu ukuthi zichaze ukuthi kusho ukuthini "i-AI" kubasebenzisi abaningi.

Ukuqapha Ukungena: Imithetho yesitolo sohlelo lokusebenza kanye nemigomo ye-API inquma ngokuthula ukuthi yiziphi izici ze-AI ezithunyelwayo.

Ukulawula komsebenzisi: Funa ukukhetha okucacile, izilungiselelo ezihlala isikhathi eside, kanye nezilawuli zokuphatha ezisebenzayo.

Ukuziphendulela: Kudinga izingodo zokuhlola, ukucaca, kanye nezindlela zokudlulisa izikhalo zemiphumela eyingozi.

Iyini Indima Yobuchwepheshe Obukhulu ku-AI? I-Infographic

🔗 Ikusasa le-AI: Amathrendi kanye nokuthi yini elandelayo
Izinto ezintsha ezibalulekile, izingozi, kanye nezimboni kuzoshintsha eminyakeni eyishumi ezayo.

🔗 Amamodeli ayisisekelo ku-AI yokukhiqiza: Umhlahlandlela olula
Qonda ukuthi amamodeli esisekelo anika amandla kanjani izinhlelo zokusebenza ze-AI zokukhiqiza zesimanje.

🔗 Iyini inkampani ye-AI nokuthi isebenza kanjani
Funda izici, amaqembu, kanye nemikhiqizo echaza amabhizinisi asebenzisa ubuchwepheshe be-AI kuqala.

🔗 Indlela ikhodi ye-AI ebukeka ngayo kumaphrojekthi wangempela
Bheka izibonelo zamaphethini ekhodi aqhutshwa yi-AI, amathuluzi, kanye nemisebenzi yokusebenza.

Ake sibhekane nakho okwesikhashana - iningi "lezingxoxo ze-AI" lidlula izingxenye ezingathandeki njengokusebenzisa ikhompyutha, ukusatshalaliswa, ukuthengwa, ukuhambisana nemithetho, kanye neqiniso elixakile lokuthi umuntu kufanele akhokhele ama-GPU nogesi. I-Big Tech ihlala kulezo zingxenye ezingathandeki. Yingakho ibaluleke kangaka. 😅 ( IEA - Amandla kanye ne-AI , i-NVIDIA - Ukubuka konke kwamapulatifomu okucabanga nge-AI )

Indima ye-AI ye-Big Tech, ngolimi olulula 🧩

Uma abantu bethi “Big Tech,” bavame ukusho izinkampani ezinkulu zesikhulumi ezilawula izendlalelo ezinkulu zekhompyutha yesimanje:

Ngakho indima akuyona nje “i-AI.” Kufana nokuthi bakha imigwaqo emikhulu, bathengise izimoto, baphathe ama-toll booth, futhi banqume ukuthi izindlela zokuphuma ziya kuphi. Ukugqagqana okuncane... kodwa hhayi kakhulu.

Indima Yobuchwepheshe Obukhulu ku-AI: imisebenzi emihlanu emikhulu 🏗️

Uma ufuna imodeli yengqondo ehlanzekile, iBig Tech ivame ukwenza imisebenzi emihlanu ehambisanayo emhlabeni we-AI:

  1. Umhlinzeki wengqalasizinda
    Izikhungo zedatha, ifu, ukuxhumana, ukuphepha, amathuluzi e-MLOps. Izinto ezenza i-AI ibe nokwenzeka ngezinga elikhulu. ( Amadokhumenti e-Amazon SageMaker AI , i-IEA - Amandla kanye ne-AI )

  2. Umakhi wemodeli kanye nenjini yocwaningo
    Hhayi njalo, kodwa kaningi - amalebhu, i-R&D yangaphakathi, ucwaningo olusetshenzisiwe, kanye "nesayensi ekhiqizwe." ( Imithetho Yokulinganisa Amamodeli Olimi Lwezinzwa (arXiv) , Ukuqeqeshwa Kwamamodeli Olimi Olukhulu Oluphelele (i-Chinchilla) (arXiv) )

  3. Umsabalalisi
    Bangakwazi ukufaka i-AI emabhokisini okusesha, amafoni, amaklayenti e-imeyili, izinhlelo zezikhangiso, kanye namathuluzi okusebenzela. Ukusabalalisa kuyigunya elikhulu.


  4. Izinqubomgomo zesitolo sohlelo lokusebenza, imithetho yesikhulumi, imigomo ye-API, ukulinganisela okuqukethwe, amasango okuphepha, izilawuli zebhizinisi. ( Zokubuyekeza Uhlelo Lokusebenza lwe-Apple , Ukuphepha Kwedatha ye-Google Play )

  5. Umhlinzeki wemali
    Bayaxhasa, bathole, bahlanganyele, bayafuya. Babumba lokho okusalayo.

Leyo yindima ye-Big Tech ku-AI ngokwemisebenzi yayo: bakha izimo zokuba i-AI ibe khona - bese benquma ukuthi ikufinyelela kanjani.

Yini eyenza inguqulo enhle yendima ye-AI ye-Big Tech ✅😬

“Inguqulo enhle” ye-Big Tech ku-AI ayiphathelene nokuphelela. Imayelana nokuhwebelana okusingathwa ngendlela efanele, kanye nokunciphisa amathuba okumangala kwabanye.

Nakhu okuvame ukuhlukanisa isimo "esisizayo esikhulu" nesimo "se-monopoly":

  • Ukucaca ngaphandle kokulahla amagama
    alula Ukulebula okucacile kwezici ze-AI, imikhawulo, kanye nedatha esetshenziswayo. Akuyona i-maze yenqubomgomo enamakhasi angu-40. ( NIST AI RMF 1.0 , ISO/IEC 42001:2023 )

  • Ukulawula kwangempela komsebenzisi
    Ukuphuma okusebenzayo, izilungiselelo zobumfihlo ezingasethwa kabusha ngendlela engaqondakali, kanye nezilawuli zokuphatha ezingezona ukuzingela okungafuneki. ( GDPR - Regulation (EU) 2016/679 )

  • Ukusebenzisana kanye nokuvuleleka - ngezinye izikhathi
    Akuzona zonke izinto okumele zibe umthombo ovulekile, kodwa ukuvalela wonke umuntu kumthengisi oyedwa unomphela... kuyisinqumo.

  • Ukuphepha ngamazinyo
    Ukuqapha ukusetshenziswa kabi, ukuhlanganisa amaqembu, ukulawula okuqukethwe, kanye nokuzimisela ukuvimba amacala okusetshenziswa ayingozi ngokusobala. ( I-NIST AI RMF 1.0 , iphrofayili ye-NIST GenAI (umngane we-AI RMF) )

  • Izimiso zemvelo ezinempilo
    Ukusekela izinkampani ezintsha, abalingani, abacwaningi, kanye nezindinganiso ezivulekile ukuze ukusungula izinto ezintsha kungabi “ukuqasha ipulatifomu noma ukunyamalala.” ( Izimiso ze-OECD AI )

Ngizokusho ngokusobala: "inguqulo enhle" izwakala njengesevisi yomphakathi eqinile enokunambitheka okunamandla komkhiqizo. Inguqulo embi izwakala njengekhasino lapho indlu ibhala khona imithetho. 🎰

Ithebula Lokuqhathanisa: “Imizila ye-AI” ephezulu ye-Big Tech nokuthi kungani isebenza 📊

Ithuluzi (umzila) Izithameli Intengo Kungani kusebenza
Amapulatifomu e-AI yamafu Amabhizinisi, izinkampani ezintsha okusekelwe ekusetshenzisweni Ukukala okulula, i-invoyisi eyodwa, izinkinobho eziningi (izinkinobho eziningi kakhulu)
Ama-API Emodeli Ye-Frontier Onjiniyela, amaqembu omkhiqizo khokha ngethokheni ngayinye / izigaba Kuyashesha ukuhlanganiswa, ikhwalithi enhle yesisekelo, kuzwakala njengokukhohlisa 😅
I-AI Eshumekwe Kudivayisi Abathengi, abathengisi kuhlanganiswe Ukubambezeleka okuphansi, ngezinye izikhathi kuhambisana nobumfihlo, kusebenza ungaxhunyiwe ku-inthanethi
I-AI Yesudi Yokukhiqiza Amaqembu ehhovisi isengezo ngesihlalo ngasinye Uhlala emisebenzini yansuku zonke - amadokhumenti, iposi, imihlangano, konke okubalulekile
Izikhangiso + I-AI Eqondisiwe Abamakethi % yezindleko Idatha enkulu + ukusatshalaliswa = kuyasebenza, futhi kuyathusa kancane 👀
Ukuphepha + Ukuthobela Imithetho I-AI Izimboni ezilawulwayo i-premium Ithengisa "ukuthula kwengqondo" - noma ngabe kuyizaziso ezimbalwa nje
Ama-AI Chips + Izisheshisi Wonke umuntu ohamba phambili i-capex-heavy Uma ungumnikazi wamafosholo, uwina igolide (isingathekiso esingaqondakali, siseyiqiniso)
Imidlalo Ye-Ecosystem Evulekile Abakhi, abacwaningi amazinga amahhala + akhokhelwayo Umfutho womphakathi, ukuphindaphinda okusheshayo, ngezinye izikhathi ukuzijabulisa okungalawuleki

Ukuvuma okungavamile etafuleni elincane: “ukukhululeka” kwenza umsebenzi omningi lapho. Kumahhala kuze kube yilapho kungasekho… uyazi ukuthi kuhamba kanjani.

Okuseduze: indawo yokuminyana kwengqalasizinda (ukubala, ifu, ama-chip) 🧱⚙️

Lena yingxenye abantu abaningi abangafuni ukukhuluma ngayo ngoba ayiyona into ekhangayo. Kodwa umgogodla we-AI.

I-Big Tech ithonya i-AI ngokulawula:

Uma uke wazama ukufaka uhlelo lwe-AI enkampanini yangempela, usuvele uyazi ukuthi "imodeli" yingxenye elula. Ingxenye enzima yile: izimvume, ukuloba, ukufinyelela idatha, izilawuli zezindleko, isikhathi sokusebenza, impendulo yezehlakalo... izinto zabantu abadala. 😵💫

Ngenxa yokuthi iBig Tech inazo eziningi zalezi zinto, ingasetha amaphethini azenzakalelayo:

  • Yimaphi amathuluzi aba ajwayelekile

  • Yiziphi izinhlaka ezithola ukwesekwa kwekilasi eliphezulu

  • Yikuphi ihadiwe okuqalwa kuqala

  • Yimaphi amamodeli entengo aba “avamile”

Lokho akukubi ngokuzenzakalelayo. Kodwa kungamandla.

Okufingqiwe: ucwaningo lwemodeli vs iqiniso lomkhiqizo 🧪➡️🛠️

Nasi isimo esishubile: I-Big Tech ingaxhasa ucwaningo olujulile futhi idinga ukuwina imikhiqizo njalo ngekota. Leyo nhlanganisela ikhiqiza izinto ezintsha ezimangalisayo futhi ikhiqiza… ukwethulwa kwezici okungabazekayo.

I-Big Tech ivame ukuqondisa intuthuko ye-AI ngokusebenzisa:

Kodwa ingcindezi yomkhiqizo ishintsha izinto:

  • Isivinini sidlula ubuhle

  • Izincazelo zezingoma zokuthumela

  • “Kuhle ngokwanele” kushaywa ngokuthi “kuqondwe ngokugcwele”

Ngezinye izikhathi lokho kulungile. Iningi labasebenzisi alidingi ubumsulwa obungokwemfundiso, badinga umsizi owusizo ngaphakathi komsebenzi wabo. Kodwa ingozi iwukuthi "okuhle ngokwanele" kusetshenziswa ezimweni ezibucayi (ezempilo, ukuqasha, ezezimali, imfundo) lapho "okuhle ngokwanele" kungafanele khona. ( Umthetho we-EU AI - Umthethonqubo (EU) 2024/1689 )

Lokhu kuyingxenye yendima ye-Big Tech ku-AI - ukuhumusha ikhono lesimanje libe izici zemakethe enkulu, noma ngabe imiphetho isabukhali. 🔪

Okuseduze: ukusatshalaliswa kwamandla amakhulu angempela 🚀📣

Uma ungabeka i-AI ezindaweni abantu asebehlala kuzo ngokwedijithali, akudingeki “ukholise” abasebenzisi. Uba nje okuzenzakalelayo.

Iziteshi zokusabalalisa ze-Big Tech zifaka:

Yingakho izinkampani ezincane ze-AI zivame ukubambisana ne-Big Tech noma ngabe zikhathazekile ngakho. Ukusatshalaliswa kwe-oxygen kuyi-oxygen. Ngaphandle kwayo, ungaba nemodeli engcono kakhulu emhlabeni kodwa usalokhu umemeza ungenalutho.

Kukhona futhi nomphumela omubi ongabonakali: ukusatshalaliswa kwakheka lokho okushiwo yi-“AI” emphakathini. Uma i-AI ibonakala njengomsizi wokubhala, abantu bacabanga ukuthi i-AI imayelana nokubhala. Uma ibonakala njengokuhlela izithombe, abantu bacabanga ukuthi i-AI imayelana nezithombe. Ipulatifomu inquma isimo.

Okuseduze: idatha, ubumfihlo, kanye nesivumelwano sokuthembana 🔐🧠

Izinhlelo ze-AI zivame ukusebenza kahle kakhulu uma zenziwe ngokwezifiso. Ukwenza ngokwezifiso ngokuvamile kudinga idatha. Futhi idatha idala ingozi. Lowo nxantathu awupheli.

I-Big Tech igxile ku:

  • Idatha yokuziphatha kwabathengi (ukusesha, ukuchofoza, izintandokazi)

  • Idatha yebhizinisi (ama-imeyili, amadokhumenti, izingxoxo, amathikithi, imisebenzi)

  • Idatha yeplatifomu (izinhlelo zokusebenza, izinkokhelo, izimpawu zobunikazi)

  • Idatha yedivayisi (indawo, izinzwa, izithombe, okokufaka kwezwi)

Ngisho noma "idatha eluhlaza" ingasetshenziswanga ngqo, uhlelo lwe-ecosystem oluzungezile lubumba ukuqeqeshwa, ukulungiswa kahle, ukuhlolwa, kanye nesiqondiso somkhiqizo.

Isivumelwano sokuthembela ngokuvamile sibukeka kanje:

  • Abasebenzisi bayamukela ukuqoqwa kwedatha ngoba umkhiqizo ulula 🧃

  • Abalawuli bayahlehla uma sekuqala ukwesabeka 👀 ( GDPR - Regulation (EU) 2016/679 )

  • Izinkampani ziphendula ngezilawuli, izinqubomgomo, kanye nemiyalezo "yobumfihlo kuqala"

  • Wonke umuntu uphikisana ngokuthi kusho ukuthini ukuthi “ubumfihlo”

Umthetho osebenzayo engiwubonile usebenza: uma inkampani ingachaza imikhuba yayo yedatha ye-AI engxoxweni eyodwa ngaphandle kokucasha ngemuva kwe-legalese, imvamisa yenza kangcono kunejwayelekile. Ayiphelele - ingcono nje.

Okuseduze: ukubusa, ukuphepha, kanye nomdlalo wethonya elithule 🧯📜

Lena indima engabonakali kakhulu: I-Big Tech ivame ukusiza ekuchazeni imithetho elandelwa yiwo wonke umuntu.

Bakha ukubusa ngokusebenzisa:

Ngezinye izikhathi lokhu kuyasiza ngempela. I-Big Tech ingatshala imali emaqenjini okuphepha, ekusetshenzisweni kwamathuluzi okuthembela, ekutholeni ukuxhashazwa, kanye nengqalasizinda yokuthobela imithetho abadlali abancane abangakwazi ukuyikhokhela.

Ngezinye izikhathi kuwukuzisebenzela. Ukuphepha kungaba yindawo yokungena, lapho abadlali abakhulu kuphela abangakwazi “ukukhokhela” ukuthobela. Yilokho okubalulekile: ukuphepha kuyadingeka, kodwa ukuphepha okubizayo kungamisa ukuncintisana ngengozi. ( Umthetho we-EU AI - Umthethonqubo (EU) 2024/1689 )

Yilapho-ke lapho ukuguquguquka okubalulekile khona. Akuyona into ejabulisayo futhi - uhlobo olucasulayo. 😬

Okuseduze: ukuncintisana, izimiso zemvelo ezivulekile, kanye namandla adonsela phansi aqalayo 🧲🌱

Indima yeBig Tech ku-AI ihlanganisa nokubumba isimo semakethe:

  • Ukutholwa (ithalente, ubuchwepheshe, ukusatshalaliswa)

  • Ubambiswano (amamodeli abanjwe kumafu, amadili ebhizinisi ahlangene)

  • Uxhaso lwe-ecosystem (amakhredithi, ama-incubator, izimakethe)

  • Amathuluzi avulekile (izinhlaka, imitapo yolwazi, ukukhishwa "okuvulekile")

Kukhona iphethini engiyibukele iphindaphindwa:

  1. Ama-Startups asungula izinto ezintsha ngokushesha

  2. I-Big Tech ihlanganisa noma ikopisha iphethini ephumelelayo

  3. Ama-startup ashintsha abe yi-niches noma abe yizinhloso zokuthenga

  4. "Isendlalelo sepulatifomu" siyaqina

Lokho akukubi ngokuzenzakalelayo. Amapulatifomu anganciphisa ukungqubuzana futhi enze i-AI itholakale kalula. Kodwa kunganciphisa nokuhlukahluka. Uma umkhiqizo ngamunye uba “isimbozo esizungeze ama-API ambalwa afanayo,” ukusungula izinto ezintsha kuqala ukuzwakala njengokulungisa kabusha ifenisha efulethini elifanayo.

Ukuncintisana okuncane okungahlelekile kunempilo. Njengesiqalo se-sourdough. Uma uhlanza yonke into, ayikhuphuki. Leso sifaniso asiphelele kancane, kodwa ngiyaqhubeka naso. 🍞

Ukuphila ngenjabulo nangokuqaphela 😄😟

Yomibili imizwa iyafanelana. Injabulo nokuqapha kungabelana ngegumbi elifanayo.

Izizathu zokujabula:

  • Ukuthunyelwa okusheshayo kwamathuluzi awusizo

  • Ingqalasizinda engcono kanye nokuthembeka

  • Isithiyo esiphansi samabhizinisi sokwamukela i-AI

  • Ukutshalwa kwezimali okwengeziwe ngokuphepha kanye nokuhlelwa kwemigomo ( NIST AI RMF 1.0 , Izimiso ze-OECD AI )

Izizathu zokuqapha:

Isimo esingokoqobo yilesi: I-Big Tech ingasheshisa i-AI emhlabeni, kuyilapho futhi ikwazi ukugxila. Lokho kungaba yiqiniso ngesikhathi esifanayo. Abantu abayithandi leyo mpendulo ngoba ayinazo izinongo, kodwa ihambisana nobufakazi.

Izinto eziwusizo zabafundi abahlukene 🎯

Uma ungumthengi webhizinisi 🧾

Uma ungunjiniyela 🧑💻

Uma ungumenzi wezinqubomgomo noma umholi wokuthobela imithetho 🏛️

Uma ungumsebenzisi ojwayelekile 🙋

  • Funda ukuthi izici ze-AI zihlala kuphi kuzinhlelo zakho zokusebenza

  • Sebenzisa izilawuli zobumfihlo noma ngabe ziyacasula ( GDPR - Regulation (EU) 2016/679 )

  • Ungabaza imiphumela "yomlingo" - i-AI iyazethemba, ayihlali ilungile 😵

Isifinyezo sokuphetha: Indima Yobuchwepheshe Obukhulu ku-AI 🧠✨

Indima yeBig Tech ku-AI ayiyona into eyodwa. Iyinqwaba yezindima: umnikazi wengqalasizinda, umakhi wemodeli, umsabalalisi, umlindi wesango, kanye nomsunguli wemakethe. Abagcini nje ngokuhlanganyela ku-AI - bachaza indawo i-AI ekhula kuyo.

Uma ukhumbula umugqa owodwa kuphela, kwenze lokhu:

Indima Yobuchwepheshe Obukhulu ku-AI
Kuwukwakha amapayipi, ukusetha okuzenzakalelayo, nokuqondisa indlela i-AI efinyelela ngayo kubantu - ngezinga elikhulu, nemiphumela emikhulu. ( I-NIST AI RMF 1.0 , Umthetho we-EU AI - Umthethonqubo (EU) 2024/1689 )

Futhi yebo, "imiphumela" izwakala imangalisa. Kodwa i-AI ingenye yalezo zihloko lapho i-drama ngezinye izikhathi iba nje... inembile. 😬🤖

Imibuzo Evame Ukubuzwa

Iyini indima yeBig Tech ku-AI, ngokwemigomo engokoqobo?

Indima Yobuchwepheshe Obukhulu ku-AI ayigcini ngokuthi “benza amamodeli” kodwa “basebenzisa imishini eyenza i-AI isebenze ngezinga elikhulu.” Bahlinzeka ngengqalasizinda yamafu, bathumela i-AI ngamadivayisi nezinhlelo zokusebenza, futhi babeka imithetho yesikhulumi ebumba lokho okwakhiwayo. Baphinde baxhase ucwaningo, ubambiswano, kanye nokuthengwa okuthonya ukuthi yiziphi izindlela ezisindayo. Ezimakethe eziningi, bachaza ngempumelelo ulwazi lwe-AI oluzenzakalelayo.

Kungani ukufinyelela kokubala kubaluleke kangaka ngobani abangakha i-AI ngezinga?

I-AI yesimanje incike kumaqoqo amakhulu e-GPU, ukuxhumana okusheshayo, isitoreji, kanye namapayipi e-MLOps athembekile - hhayi nje ama-algorithms ahlakaniphile. Uma ungakwazi ukuthola amandla angabikezelwa, ukuqeqeshwa, ukuhlolwa, kanye nokusetshenziswa kuba buthakathaka futhi kubize kakhulu. I-Big Tech ivame ukulawula ungqimba "lomgogodla" (ifu, ukubambisana kwama-chips, ukuhlela, ukuphepha), okungasetha lokho okunokwenzeka kumaqembu amancane. Lawo mandla angaba wusizo, kodwa ahlala enamandla.

Ukusatshalaliswa kwe-Big Tech kuthinta kanjani lokho okushiwo yi-“AI” kubasebenzisi bansuku zonke?

Ukusatshalaliswa kuyikhono elikhulu ngoba kuguqula i-AI ibe yisici esizenzakalelayo esikhundleni somkhiqizo ohlukile okumele uwukhethe. Lapho i-AI ivela kumabha okusesha, amafoni, ama-imeyili, amadokhumenti, imihlangano, kanye nezitolo zezinhlelo zokusebenza, iba "yilokho i-AI eyikho" kubantu abaningi. Lokho futhi kunciphisa okulindelwe umphakathi: uma i-AI iyithuluzi lokubhala kakhulu kuzinhlelo zakho zokusebenza, abasebenzisi bacabanga ukuthi i-AI ilingana nokubhala. Amapulatifomu anquma ngokuthula ithoni.

Yiziphi izindlela eziyinhloko imithetho yesikhulumi kanye nezitolo zezinhlelo zokusebenza ezisebenza ngazo njengabalindi be-AI?

Izinqubomgomo zokubuyekezwa kohlelo lokusebenza, imigomo yemakethe, imithetho yokuqukethwe, kanye nemikhawulo ye-API kunganquma ukuthi yiziphi izici ze-AI ezivunyelwe nokuthi kufanele ziziphathe kanjani. Ngisho noma imithetho ihlelwe njengokuvikela ukuphepha noma ubumfihlo, iphinda ibumbe umncintiswano ngokukhulisa izindleko zokuthobela imithetho kanye nokusetshenziswa. Kwabathuthukisi, lokhu kusho ukuthi izibuyekezo zenqubomgomo zingabaluleka njengezibuyekezo zemodeli. Empeleni, "okuthunyelwayo" kuvame ukuba "yilokho okudlula isango."

Amapulatifomu e-AI yamafu afana ne-SageMaker, i-Azure ML, kanye ne-Vertex AI angena kanjani endimeni ye-Big Tech ku-AI?

Amapulatifomu e-Cloud AI ahlanganisa ukuqeqeshwa, ukuthunyelwa, ukuqapha, ukuphatha, kanye nokuphepha endaweni eyodwa, okunciphisa ukungezwani kwabaqalayo namabhizinisi. Amathuluzi afana ne-Amazon SageMaker, i-Azure Machine Learning, kanye ne-Vertex AI enza kube lula ukukhulisa nokuphatha izindleko ngobudlelwano bomthengisi oyedwa. Umehluko ukuthi ukulula kungandisa ukukhiya, ngoba imisebenzi yokusebenza, izimvume, kanye nokuqapha kuhlanganiswe ngokujulile kuleyo ndawo yokusebenza.

Yini okufanele umthengi webhizinisi ayibuze ngaphambi kokusebenzisa amathuluzi e-AI e-Big Tech?

Qala ngedatha: ukuthi iya kuphi, ukuthi ihlukaniswe kanjani, nokuthi yiziphi izilawuli zokugcina nokuhlola ezikhona. Buza mayelana nezilawuli zokuphatha, ukuloga, imingcele yokufinyelela, nokuthi amamodeli ahlolwa kanjani ngengozi esizindeni sakho. Futhi intengo yokuhlola ingcindezi, ngoba izindleko ezisekelwe ekusetshenzisweni zingakhuphuka njengoba ukwamukelwa kukhula. Ezilungiselelweni ezilawulwayo, vumelanisa okulindelwe nezinhlaka kanye nezidingo zokuthobela imithetho inhlangano yakho esivele izisebenzisa.

Abathuthukisi bangakugwema kanjani ukuvalelwa kwabathengisi lapho bakhela kuma-API e-Big Tech AI?

Indlela evamile ukuklama ukuphatheka kalula: ukugoqa amakholi emodeli ngemuva kwesendlalelo sokuthatha izinto bese ugcina izixwayiso, izinqubomgomo, kanye ne-logic yokuhlola kuhunyushwe futhi kuhlolwe. Gwema ukuthembela kwisici esisodwa "esikhethekile" somthengisi esingashintsha noma sinyamalale. Landelela imikhawulo yamanani, izibuyekezo zamanani, kanye nezinguquko zenqubomgomo njengengxenye yokulungiswa okuqhubekayo. Ukuphatheka kalula akumahhala, kodwa ngokuvamile kubiza kancane kunokufuduka okuphoqelelwe.

Ubumfihlo kanye nokwenza kube ngokwakho kudala kanjani "isivumelwano sokuthembana" ngezici ze-AI?

Ukwenza kube ngokwakho kuvame ukuthuthukisa ukusetshenziswa kwe-AI, kodwa ngokuvamile kwandisa ukudalulwa kwedatha kanye nokubonakala kokwethuka. I-Big Tech isondele kakhulu kudatha yokuziphatha, yebhizinisi, yesikhulumi, kanye nedivayisi, ngakho abasebenzisi nabalawuli bahlola ukuthi leyo datha ithonya kanjani ukuqeqeshwa, ukulungiswa kahle, kanye nezinqumo zomkhiqizo. Isilinganiso esisebenzayo ukuthi inkampani ingayichaza yini imikhuba yayo yedatha ye-AI ngokucacile ngaphandle kokufihla ngemuva kolimi olusemthethweni. Ukulawula okuhle kanye nokuphuma kwangempela kubalulekile.

Yiziphi izindinganiso nemithethonqubo ezifanele kakhulu ekuphathweni nasekuphepheni kwe-AI ye-Big Tech?

Ezindleleni eziningi, ukubusa kuhlanganisa izinqubomgomo zokuphepha zangaphakathi nezinhlaka zangaphandle kanye nemithetho. Izinhlangano zivame ukubhekisela eziqondisweni zokuphathwa kwezingozi njenge-NIST's AI RMF, amazinga okuphatha njenge-ISO/IEC 42001, kanye nemithetho yesifunda efana ne-GDPR kanye ne-EU AI Act yamacala athile okusetshenziswa. Lokhu kuthinta ukuloga, ukuhlolwa kwamabhuku, imingcele yedatha, kanye nalokho okuvinjelwayo noma okuvunyelwe. Inselele iwukuthi ukuthobela imithetho kungabiza kakhulu, okungase kusize abadlali abakhulu.

Ingabe ithonya le-Big Tech ekuncintisaneni nasezindaweni zemvelo lihlala liyinto embi?

Hhayi ngokuzenzakalelayo. Amapulatifomu anganciphisa izithiyo, enze amathuluzi afane, futhi axhase ukuphepha nengqalasizinda amaqembu amancane angenakukwazi ukuyikhokhela. Kodwa ukuguquguquka okufanayo kunganciphisa ukuhlukahluka uma wonke umuntu eba yisithiyo esiseduze nama-API ambalwa amakhulu, amafu, kanye nezimakethe. Qaphela amaphethini afana nokuhlanganiswa kokubala nokusabalalisa, kanye namanani kanye nokushintsha kwenqubomgomo okunzima ukukugwema. Izindawo zemvelo ezinempilo kakhulu zivame ukugcina indawo yokusebenzisana kanye nabantu abasha abangenayo.

Izinkomba

  1. I-International Energy Agency - Amandla kanye ne-AI - iea.org

  2. I-International Energy Agency - Isidingo samandla esivela ku-AI - iea.org

  3. NVIDIA - AI - nvidia.com

  4. Izinsizakalo Zewebhu ze-Amazon - Imibhalo ye-AI ye-Amazon SageMaker (Kuyini i-SageMaker?) - aws.amazon.com

  5. Microsoft - Azure Machine Learning - learn.microsoft.com

  6. I-Google Cloud - Idokhumenti ye-Vertex AI - cloud.google.com

  7. I-Google Cloud - Ama-MLOps ku-Vertex AI - cloud.google.com

  8. I-Microsoft - Umhlahlandlela wokwakha wemisebenzi yokufunda komshini (i-MLOps) v2 - learn.microsoft.com

  9. Unjiniyela we-Apple - i-Core ML - developer.apple.com

  10. Onjiniyela be-Google - Ikhithi ye-ML - developers.google.com

  11. Unjiniyela we-Apple - Iziqondiso Zokubuyekeza Uhlelo Lokusebenza - developer.apple.com

  12. Usizo lwe-Google Play Console - Ukuphepha kwedatha - support.google.com

  13. arXiv - Imithetho Yokukala Yamamodeli Olimi Lwezinzwa - arxiv.org

  14. arXiv - Ukuqeqeshwa Kwamamodeli Olimi Olukhulu Oluphelele Lwekhompyutha (iChinchilla) - arxiv.org

  15. Isikhungo Sikazwelonke Sezindinganiso Nobuchwepheshe - Uhlaka Lokuphathwa Kwengozi ye-AI (AI RMF 1.0) - nist.gov

  16. Isikhungo Sikazwelonke Sezindinganiso Nobuchwepheshe - Iphrofayili ye-NIST Generative AI (umngane we-AI RMF) - nist.gov

  17. Inhlangano Yomhlaba Wonke Yokumisa Izindinganiso - ISO/IEC 42001:2023 - iso.org

  18. I-EUR-Lex - Umthethonqubo (i-EU) 2016/679 (GDPR) - eur-lex.europa.eu

  19. I-EUR-Lex - Umthethonqubo (EU) 2024/1689 (EU AI Act) - eur-lex.europa.eu

  20. I-OECD - Izimiso ze-OECD AI - oecd.ai

Thola i-AI Yakamuva Esitolo Esisemthethweni Somsizi we-AI

Mayelana NATHI

Buyela kubhulogi