9 Izinjini zokusesha ze-AI okufanele uzazi

I-Generative AI iguqule ukusesha ku-inthanethi, yethula isizukulwane esisha sezinjini zokusesha eziphendula imibuzo ngokubuka kafushane esikhundleni sohlu olujwayelekile lwezixhumanisi eziluhlaza okwesibhakabhaka. Abaholi bemboni abafana ne-Google ne-Bing bahlanganisa ngokushesha i-AI ezinkundleni zabo, belwela ukuhambisana nezihlabani ezikhulayo ezifana ne-Perplexity ne-You.com. Kodwa amandla aphelele enjini yokusesha enamandla e-AI asavela, enikeza kokubili izithembiso ezinkulu kanye nezingozi ezinkulu.

Iyini injini yokusesha ye-AI?

Izinjini zokusesha ze-AI zisetshenziswa ukuhlakanipha okungekhona okwangempela — ngokukhethekile, izindlela ezifana nokucubungula ulimi lwemvelo (NLP) nokufunda komshini — ukuhumusha kangcono umongo nokuqonda inhloso yomsebenzisi ngemuva kombuzo ngamunye wosesho, ngenhloso yokuletha imiphumela ehambisana kakhulu. Empeleni, lokhu kubonakala kukhiqiza izifinyezo ezimfushane, zolimi olulula zolwazi oluqoqwe kuwebhu yonkana, ngaphezu—noma esikhundleni—sokukhomba abasebenzisi kumasayithi ahlukene.

"Akukunikezi izixhumanisi ekhasini," umeluleki wezokukhangisa ngedijithali u-Jordan Stevens utshele i-Built In.

Izinjini zokusesha ze-AI zifana ne-AI chatbots: ziyaxoxa futhi zingaphendula imibuzo ngendlela eshelelayo, efana neyomuntu. Kepha ngokungafani nama-chatbots amaningi, izinjini zokusesha ze-AI zinokufinyelela ngesikhathi sangempela internet, okubavumela ukuthi banikeze ulwazi olubanzi nolwamanje. Lokhu kuzibeka njengombangi hhayi kuma-chatbots amanje kuphela kodwa adumile njenge-ChatGPT.

9 Izinjini zokusesha ze-AI okufanele uzazi

I-1. Google

I-Google iyinjini yokusesha edume kakhulu emhlabeni, ilawula cishe amaphesenti angama-90 emakethe yokusesha. Inkampani yaqala ukwenza kabusha iwebhusayithi yayo ngo-2023 futhi isiqalile ukusebenza ukuhlakanipha okungekhona okwangempela cishe kuzo zonke izingxenye zenqubo yakho yosesho usebenzisa imodeli yakho ye-Gemini. Ukubuka konke kwe-Google kwe-AI kunikeza isifinyezo esikhiqizwe yi-AI semiphumela eminingi yosesho ephezulu ekhasini, enikeza abasebenzisi umuzwa ojwayelekile wempendulo yombuzo wabo, kanye nezixhumanisi zezinsiza zokuthola ulwazi olwengeziwe. Futhi ithuluzi lokuhlela le-Google lingadala noma yini kusukela kuhlelo lokudla ukuya ohambweni lweholide olusekelwe embuzweni owodwa; isebenzisa i-Gemini ukuze igcwalise futhi ihlele imiphumela yosesho yabasebenzisi ekhasini.

Intengo: ibhuku

2. I-Bing

I-Bing iyinjini yokusesha yesibili edume kakhulu ngemuva kwe-Google, ebalwa cishe u-4% wevolumu yokusesha ephelele emhlabeni jikelele. Kuthuthukiswe ngu Microsoft, i-Bing inikwa amandla imodeli yolimi enkulu ye-OpenAI's GPT-4 (LLM). Ngokungeziwe kumakhono ayo okusesha asekelwe kumbhalo, izwi, kanye nesithombe, inkundla iza nesici sengxoxo esibizwa ngokuthi i-Copilot, lapho abasebenzisi bengabuza khona imibuzo futhi bathole izimpendulo ezinemininingwane, zobuntu njengembuyiselo, kanye nezingcaphuno zendatshana ku-inthanethi ngomugqa ngamunye . U-Copilot uphinde eze nesici sokuhlela iholide, umsizi wasekhishini, ijeneretha yesithombe se-AI, kanye nethuluzi eliphendula imibuzo ehlobene nokufaneleka.

Intengo: ibhuku

3. Ukuxakeka

Iqala ngo-2022, i-Perplexity ihlinzeka ngezifinyezo ezenziwe nge-AI zemiphumela yosesho, enezixhumanisi zemithombo kuphaneli ephezulu kanye nohlu lwemibuzo yokulandelela ngezansi - efana nesici se-Google sokuthi "abantu bayabuza". Iphinde ivumele abasebenzisi ukuthi banciphise usesho lwabo lube yisethi ethile yemithombo, njengamavidiyo e-YouTube noma okuthunyelwe kwe-Reddit. Inguqulo yamahhala ye-Perplexity isebenza ku-OpenAI's GPT-3.5 LLM kanye nemodeli yayo ye-AI. Abasebenzisi abathuthukela enguqulweni ye-Pro bangaphinda bakhethe kwamanye amamodeli ambalwa, okuhlanganisa i-GPT-4 ne-Anthropic's Claude 3.

Intengo: Mahhala, nenguqulo ye-Pro ethuthuke kakhulu etholakalayo ngo-$20/ngenyanga.

4. Ukucinga Ngesibindi

I-Brave Search imakethwa kubasebenzisi abaqaphela ubumfihlo njengoba ingaqoqi noma yiluphi ulwazi lomuntu siqu noma umlando wosesho. Njengezinye izinjini eziningi zokusesha ze-AI, i-Brave Search ihlinzeka ngemininingwane esheshayo eyenziwe nge-AI ekuphenduleni noma yimuphi umbuzo wolimi lwemvelo, kodwa ayinciki ezinjinini zokusesha ezinkulu njenge-Google noma i-Bing ukwenza kanjalo. Esikhundleni salokho, inkampani ithi izimpendulo zayo zisekelwe kunkomba yayo yokusesha ezimele, ithi lokhu kususa "ukuchema" kanye "nokucwaninga" okungafika "nogaxekile omkhulu we-SEO." Nakuba i-Brave Search imahhala ukusetshenziswa kunoma isiphi isiphequluli sewebhu, iphinde inikeze ngomngane wayo wamahhala wesiphequluli esivimba abalandeleli ekuqoqeni idatha yomsebenzisi kuwebhu yonkana.

Intengo: ibhuku

5. You.com

I-You.com ithi injini yokusesha yokuqala ukuhlanganisa i-LLM ebheke umthengi nokufinyelela ngesikhathi sangempela kuyo internet, futhi obekungowokuqala futhi ukwethula amakhono engxoxo ye-multimodal esibonakalayo sayo. Inkundla isebenza kakhulu njengomsizi we-AI kunenjini yokusesha ejwayelekile, futhi iza “ngezindlela” ezimbalwa ezihlukene kusukela ku-Smart (ngemisebenzi esheshayo efana nosesho lwewebhu nokudala okuqukethwe) kuya ku-Dala (ukwenza izithombe). Iphinde inikeze ngemodi Eyimfihlo, lapho ingagcini khona umsebenzi wosesho wabasebenzisi noma ikheli le-IP.

Intengo: Kumahhala, ngezinguqulo ezithuthuke kakhulu eziqala ku-US$15/ngenyanga.

6. Andi

U-Andi usebenza njengenjini yokusesha kanye ne-chatbot. Ingaphendula imibuzo yabasebenzisi, ichaze izihloko eziyinkimbinkimbi, futhi izule kuwebhu, ihlanganise ulwazi oluvela emiphumeleni ephezulu ukuze inikeze imininingwane ekhiqizwe yi-AI (engabukwa ngokuchofoza “Khiqiza Umbhalo”). Abasebenzisi bangakwazi futhi ukwenza ngendlela oyifisayo indlela iplatifomu ebonisa ngayo izixhumanisi, ukukhetha phakathi kohlu, igridi noma izakhiwo zokuphakelayo; futhi ingalungisa isitayela ukuze sifane nezindaba ze-Hacker, umbhalo we-markdown, noma ikhasi le-Google lakudala. Yize i-Andi iqoqa idatha kubasebenzisi, ithi ayabelani nanoma ubani omunye. Inkampani futhi ayirekhodi ukusesha futhi ivimbele ukulandelela kanye nezikhangiso.

Intengo: ibhuku

7. Komo

I-Komo iyinjini yokusesha engakhangisi enikwa amandla yi-LLM yayo ebizwa ngokuthi Sunshine. Ingalungiselelwa kumamodi amane ahlukene: Buza, Sesha, Cwaninga futhi Hlola. Nge-Ask, abasebenzisi bathola izimpendulo ezikhiqizwe yi-AI kunoma yimuphi umbuzo abawubuzayo, kanye nezixhumanisi ezifanele; Ukusesha kusebenza njengenjini yokusesha evamile, inikeza uhlu lwezixhumanisi ezifanele; Ucwaningo (olutholakala kumakhasimende akhokhayo kuphela) luhlinzeka ngokuhlaziywa kwesihloko esakhiwe nge-AI, kanye nezingcaphuno; kanye ne-Explore inikeza kuphela izixhumanisi zamavidiyo ukuphendula imibuzo yabasebenzisi.

Intengo: Uhlelo oluyisisekelo lubiza u-$8/ngenyanga, uhlelo lwe-Premium lubiza u-$15/ngenyanga, futhi uhlelo lwebhizinisi luyahlukahluka ngentengo.

8.wally

Isebenziseka emisebenzini ehilela ucwaningo olunzulu, njengokwenza iziphakamiso zebhizinisi noma amasu okufaka uphawu lwentengiso, u-Waldo uthatha izinkomba zosesho ezikhishwe kumasayithi afana ne-Google ne-Bing futhi azipakishe kusixhumi esibonakalayo enziwe ngendlela oyifisayo esingashunwa ngokunembile kulokho umsebenzisi akufunayo. Ngokuchofoza izinkinobho ezimbalwa, abasebenzisi banganciphisa imiphumela yabo yosesho ibe izinhlobo ezithile zamafayela, bagqamise amagama athile noma izinombolo, bavimbe amafonti athile, nokunye okwengeziwe. Abasebenzisi bangakwazi futhi ukuthatha amanothi encwadini yokubhalela eyakhelwe ngaphakathi bese behlanganisa lawa manothi efayeleni elilodwa noma i-Google Amadokhumenti engabiwa nabanye.

Intengo: Uhlelo ngezokuvakasha kubiza u-$149/ngenyanga, uhlelo Lwethimba lubiza u-$179/inyanga/umuntu, uhlelo lwe-Teams Pro lubiza u-$199/ngenyanga/umuntu, futhi uhlelo lwe-Enterprise luyahlukahluka ngentengo.

9. Phind

I-Phind iyinjini yokusesha ye-AI eyenzelwe abathuthukisi be-software. Ixhunywe ku-inthanethi kanye nesizinda sekhodi somsebenzisi (uma bekhetha), inkundla ingaphendula zonke izinhlobo zemibuzo ehlobene nobuchwepheshe, kusukela kokuthi 'Ayini amabhentshimakhi e-Mixtral 8x7B?' kokuthi 'Ngisenza kanjani isicelo se-HTTP ku-JavaScript?' - futhi kufaka phakathi izixhumanisi zemithombo kanye nezimpendulo zabo ezikhiqizwe yi-AI. Ukwengeza, ekusetshenzisweni kwayo kwekhodi, abasebenzisi bangafaka i-codebase yabo futhi babuze imibuzo ethile mayelana nekhodi yabo, babhale ngokuzenzakalelayo ukuhlolwa kweyunithi, noma basebenzise amazwibela wePython ngqo esipheqululini.

Intengo: Abasebenzisi bangabhalisela i-Pro nyanga zonke ngo-$20/ngenyanga noma ngonyaka ngo-$17/ngenyanga. I-Phind iphinde inikeze ngohlelo lwebhizinisi luka-$40/ngenyanga/umuntu.

Zisebenza kanjani izinjini zokusesha ze-AI?

Izinjini zokusesha ze-AI zisebenza ngokukhasa kuqala futhi zikhombe amakhasi ewebhu kuyo yonke i-inthanethi, zikhiphe idatha ewusizo njengombhalo, izithombe nezixhumanisi. Basebenzisa ukufunda komshini nama-algorithms okucubungula ulimi lwemvelo ukuze bahlaziye ngokujule kakhudlwana okuqukethwe nokwakheka kwamakhasi ewebhu ukuze bakuqonde ngokujulile.

Lonke lolu lwazi lube selucutshungulwa, luhlelwe futhi lugcinwe kusizindalwazi, lapho lungatholwa khona ngesikhathi sokuseshwa. Izinjini zokusesha ze-AI zihlala zibuyekeza futhi zihlunga izinkomba zazo njengoba zifunda ekusebenzelaneni kwabasebenzisi nedatha entsha.

Uma umsebenzisi ethayipha umbuzo, izinjini zokusesha ze-AI zisebenzisa amamodeli olimi amakhulu ukuqonda kangcono inhloso yomsebenzisi ngokuhlaziya incazelo yamagama ngamanye kanye namaphethini nobudlelwano phakathi kwawo. Bese baqondanisa umbuzo womsebenzisi nemiphumela yosesho efanele ezinkombeni zabo, noma ngabe umbuzo udidayo noma ubhalwe ngokwezwi nezwi.

Ngokuhamba kwesikhathi, ezinye izinjini zokusesha ze-AI ziqoqa idatha emayelana nomlando wosesho wabasebenzisi, okuyizixhumanisi abazichofozile, namanye amamethrikhi ukuze kwakheke amaphrofayela athile omsebenzisi. Lokhu kubavumela ukuthi bafunde okuncamelayo futhi bakunikeze izimpendulo eziqondene nawe.

Izinjini zokusesha ze-AI nazo zisebenzisa ama-LLM ukuhlanganisa imininingwane eziyithola kuwebhu zibe izimpendulo eziphelele ezikhiqizwe yi-AI. Beqeqeshwe ngenani elikhulu ledatha yombhalo, amamodeli amakhulu olimi abikezela igama elilandelayo elinengqondo kakhulu ngokusekelwe kumongo onikezwe amagama adlule, okuwavumela ukuba akhiqize iphrozi ehambisanayo, elungile ngokohlelo olufana nse nombhalo obhalwe abantu.

Zihluke kanjani izinjini zokusesha ze-AI kuzinjini zokusesha ezijwayelekile?

Izinjini zokusesha ze-AI zabelana ngomgomo ofanayo nezinjini zokusesha zendabuko: ukunikeza abasebenzisi izimpendulo ezibaholela ekutheni bachofoze isixhumanisi, ngaleyo ndlela baqedele "uhambo lwabo lokusesha," njengoba uStevens akubeka. "Ekupheleni kosuku, le njini yokusesha ifuna ukukunikeza impendulo eyikhwalithi."

Kodwa benza umsebenzi wabo ngezindlela ezahlukene.

Ukusesha kwe-AI kunikeza izifinyezo, ukusesha okuvamile kunikeza izixhumanisi

Izinjini zokusesha ezivamile zibonisa imiphumela yazo njengohlu lwezixhumanisi, namazwibela amafushane ombhalo ngezansi, kuyilapho izinjini zokusesha ze-AI zingaphinda zikhiqize izifinyezo ezimfushane zemiphumela yosesho, zinikeze abasebenzisi impendulo embuzweni wabo ngaphandle kokuthi baskrole futhi bachofoze kumawebhusayithi ahlukahlukene.

Ukusesha kwe-AI kufunda ngokuhamba kwesikhathi, ukusesha okuvamile kuchazwa kabusha

Izinjini zokusesha ze-AI ziklanyelwe ukuqhubeka zifunda futhi zijwayelane nokusebenzisana kwabasebenzisi nedatha entsha, zithuthukise ukusebenza kwazo kanye nekhwalithi yemiphumela yazo yosesho ngokuhamba kwesikhathi. Okwamanje, izinjini zokusesha ezivamile zisebenza kusukela ekuqaleni ngombuzo ngamunye omusha, ngaphandle kokucabangela noma imiphi imibuzo yangaphambilini noma ukusebenzisana komsebenzisi.

Ukusesha kwe-AI kugxile kuma-semantics, ukusesha okuvamile kugxile kumagama angukhiye

Izinjini zokusesha ezivamile zithembele ngokuyinhloko ekufanisweni kwegama elingukhiye ukuze zitolike imibuzo yabasebenzisi, kuyilapho ozakwabo be-AI begxila kuma-semantics, kucatshangelwa incazelo ebanzi yamagama ngaphakathi komongo womusho. Lokhu kuvumela injini yokusesha ye-AI ukuthi iqonde ngokugcwele inhloso yomsebenzisi, ukuyivumela ukuthi inikeze imiphumela yosesho esondelene kakhulu nalokho umsebenzisi akufunayo.

“Igxile kakhulu kulokho okushoyo,” kusho uDennis Perpetua, oyisikhulu sezobuchwepheshe ezinsizeni zedijithali enkampanini yezobuchwepheshe i-Kyndryl. “Bese ungakwazi ukulungisa imiphumela ngeziyalezo noma imibuzo yokulandelela, kunokusebenzisa amagama angukhiye ahlekisayo kanye nezimpawu nezinto ezinjalo.”

Ngokugxila kuma-semantics esikhundleni samagama angukhiye kuphela, izinjini zokusesha ze-AI zinganikeza imiphumela ewusizo noma ngabe inhloso ingaphansi kokusobala. Isibonelo, uma umsebenzisi ecela "izindawo ezingcono kakhulu zokudla eduze kwami ​​nezihlalo zangaphandle," injini yokusesha ye-AI kufanele ikwazi ukufaka kuhlu izindawo zokudlela eziseduze ezinamapatio naphezu kweqiniso lokuthi umbuzo awunawo amagama athi "indawo yokudlela". ” noma “igceke”.

"Izinjini zokusesha ze-AI zenza ikhono lokuthola izinto zifinyeleleke, ngenxa nje yokuthi ungakwazi ukuchaza ngombono ongokomzwelo kanye ne-semantic, ngaphansi kwegama elingukhiye noma iphuzu lokunemba lokubuka," kuqhubeka u-Perpetua. "Ivula indawo ebanzi futhi ihlunga izimpendulo ukuthi zibe mncane kakhulu kunokucwaninga kwamagama angukhiye lapho kufanele uqhubeke ulungisa izinto ngesandla."

Ukusesha okuvamile kubala kuphela Imibuzo yombhalo, ukusesha kwe-AI kwamukela izithombe, ividiyo nomsindo

Izinjini zokusesha ezivamile zamukela kuphela okufakiwe kombhalo. Kodwa ezinye izinjini zokusesha ze-AI zinekhono le-multimodal, okusho ukuthi ziyakwazi ukucubungula nokuqonda ulwazi oluvela kumafomethi ngaphandle nje kombhalo, njenge kuvatamavidiyo nomsindo. Isibonelo, injini yosesho ye-AI ye-multimodal ingase ivumele abasebenzisi ukuthi baseshe ulwazi besebenzisa imiyalo yezwi noma ngokuthatha isithombe ngocingo lwabo - inikeze ulwazi lokusesha olunembile futhi oluguquguqukayo kunokumane uthayiphe imibuzo.

Ingabe izinjini zokusesha ze-AI zingcono kunezinjini zokusesha ezijwayelekile?

Nakuba izinjini zokusesha ze-AI zizinhle ngokusiza ukuhlela iphathi yosuku lokuzalwa lwama-21 noma ukunikeza incazelo elula yokuthi kungani isibhakabhaka siluhlaza, azihlali ziyinketho engcono kakhulu yokusingatha yonke eminye imisebenzi yabantu esisebenzisa izinjini zokusesha ukuzulazula kumawebhusayithi athile noma ukulandelela ukuqhubeka komcimbi wezemidlalo ngesikhathi sangempela.

"Ama-AI awenzi kahle ngezinto ezintsha, futhi abenzi kahle ngabangaphandle," kusho uStevens. "Noma yini engaphandle ngokweqile noma yini entsha, ngicabanga ukuthi ama-AI azodonsa kanzima."

Izinjini zokusesha ze-AI nazo zingakhiqiza izimpendulo ezichemile. Izinjini zokusesha ezivamile zivele zihlele ulwazi njengezixhumanisi zabasebenzisi abangachofoza, bazifunde, futhi bazihumushe ngokwabo, kuyilapho izinjini zokusesha ze-AI zifingqa okuqukethwe okutholakala kuwebhu, zikhiqiza izimpendulo ezintsha ngokusekelwe kulokho abakufundayo. Lokhu kusho ukuthi imiphumela ekhiqizwe injini yokusesha ye-AI cishe ikhombisa ukuchema kwabathuthukisi bayo, yethula ukwenzelela ezindaweni lapho bekungekho khona ukuqala, kusho uDavi Ottenheimer, iphini likamongameli we-digital trust and ethics enkampanini yezobuchwepheshe i-Inrupt.

"Izinjini zokusesha ze-AI sezinqanyulwa kubadidiyeli, abanikazi bedatha," kusho u-Ottenheimer etshela i-Built In. i-algorithm ye-AI.

Njengayo yonke i-AI ekhiqizayo, izinjini zokusesha ze-AI nazo zithambekele ekuboneni izinto ezingekho, kwesinye isikhathi zethule amanga njengamaqiniso. Amanye amathuluzi ahlanganise izingcaphuno ze-athikili yomugqa ngomugqa ukuze kunxeshezelwe inkinga, kodwa izindatshana ezikhonjiwe azihlali zinembile - futhi ngezinye izikhathi azikho nhlobo.

Ngakho-ke, nakuba izinjini zokusesha ze-AI zingase zibe namandla futhi zisebenzisane ngokwemvelo kunezinjini zokusesha ezivamile, azikhiqizi izimpendulo zekhwalithi ephezulu kunezakwabo ezingezona eze-AI.