Nezve Artificial Intelligence Bias

tl; dr:

• Kudzidza kwemichina kunotsvaga mapatani mu data. Asi huchenjeri hwekugadzira hunogona "kurerekera" -kureva, tsvaga maitiro asina kunaka. Semuyenzaniso, foto-based skin cancer yekuona system inogona kubhadhara zvakanyanya kumifananidzo inotorwa muhofisi yachiremba. Kudzidza muchina hakugone kunzwisisa: maalgorithms ayo anongoratidza mapatani munhamba, uye kana iyo data isingamiriri, ndizvo zvichaita mhedzisiro yekugadziriswa kwayo. Uye kubata tsikidzi dzakadaro kunogona kuve kwakaoma nekuda kweiyo mechanics ekudzidza muchina.
• Nzvimbo yedambudziko iri pachena uye inotyisa ndeyekusiyana kwevanhu. Pane zvikonzero zvakawanda nei data nezve vanhu ichigona kurasikirwa netarisiro kunyangwe padanho rekuunganidza. Asi usafunga kuti dambudziko iri rinobata vanhu chete: matambudziko akafanana anomuka kana uchiedza kuona mafashama mudura kana gasi rakakundikana. Mamwe masisitimu anogona kunge akarerekera kune ruvara rweganda, mamwe anozorerekera kune Nokia sensors.
• Matambudziko akadaro haasi matsva pakudzidza kwemichina, uye ari kure nekusiyana kwazviri. Kufungidzira kwakashata kunoitwa mune chero chimiro chakaoma, uye kunzwisisa kuti nei imwe sarudzo yakaitwa kunogara kwakaoma. Isu tinofanirwa kurwisa izvi nenzira yakazara: gadzira maturusi nematanho ekusimbisa - uye dzidzisa vashandisi kuti vasaite bofu kutevedzera AI kurudziro. Kudzidza nemuchina kunoita zvimwe zvinhu zvirinani kupfuura zvatingaite - asi imbwa, semuenzaniso, dzinoshanda zvakanyanya kupfuura vanhu pakuona zvinodhaka, icho chisiri chikonzero chekuzvishandisa sezvapupu uye kuita mitongo zvichienderana neuchapupu hwadzo. Uye imbwa, nenzira, dzakangwara kupfuura chero muchina wekudzidza system.

Kudzidza kwemichina ndeimwe yeakanyanya kukosha tekinoroji maitiro nhasi. Iyi ndiyo imwe yenzira huru iyo tekinoroji ichachinja nyika yakatitenderedza mumakore gumi anotevera. Zvimwe zvekuchinja uku ndizvo zvinokonzera kunetseka. Semuyenzaniso, kukanganisa kungangoitika kwemuchina kudzidza pamusika wevashandi, kana kushandiswa kwayo nekuda kwezvikonzero zvisina kunaka (semuenzaniso, nehurumende dzehutongi). Pane rimwe dambudziko rinogadziriswa nepositi iyi: artificial intelligence bias.

Iyi haisi nyaya iri nyore.

Google's AI inogona kuwana katsi. Iyi nhau kubva muna 2012 yaive chinhu chakakosha kare ikako.

Chii chinonzi "AI Bias"?

"Raw data" ndeye oxymoron uye pfungwa yakaipa; data inofanira kugadzirirwa zvakanaka uye nokungwarira. —Geoffrey Boker

Pane imwe nzvimbo pamberi pe2013, kuitira kuti uite hurongwa hunoti, hunoziva katsi mumifananidzo, waifanira kutsanangura matanho ane musoro. Nzira yekuwana sei makona mumufananidzo, ziva maziso, ongorora manyorerwo emvere, kuverenga paws, zvichingodaro. Wobva waisa zvinhu zvese pamwe chete uye woona kuti hazvishande. Zvakawanda sebhiza remuchina - ne theoretically rinogona kugadzirwa, asi mukuita rakanyanya kuoma kutsanangura. Mhedzisiro ndeye mazana (kana kunyange zviuru) zvemitemo yakanyorwa nemaoko. Uye kwete imwechete inoshanda modhi.

Nekuuya kwemichina yekudzidza, takamira kushandisa "manual" mitemo yekuziva chimwe chinhu. Pane kudaro, tinotora zviuru zvemasampuli e "iyi", X, zviuru zve "zvimwe", Y, uye kombiyuta ivake muenzaniso zvichienderana nekuongorora kwavo nhamba. Isu tinozopa iyi modhi imwe data yemuenzaniso uye inosarudza nehumwe hunhu kuti inokodzera imwe yemaseti. Kudzidza kwemichina kunogadzira modhi kubva kune data kwete kubva kumunhu kuinyora. Mhedzisiro yacho inokatyamadza, kunyanya mundima yemufananidzo uye kucherechedzwa kwepateni, ndosaka iyo indasitiri yese yetekinoroji yave kuenda kumuchina kudzidza (ML).

Asi hazvisi nyore kudaro. Munyika chaiyo, zviuru zvako zvemienzaniso ye X kana Y inewo A, B, J, L, O, R, uye kunyange L. Izvi zvinogona kunge zvisina kugovaniswa zvakaenzana, uye zvimwe zvinogona kuitika kakawanda zvekuti system inobhadhara zvakawanda. kuchenjerera kwavari kupfuura zvinhu zvinokufadza.

Izvi zvinorevei mukuita? Muenzaniso wangu wandinoda ndewe kana masisitimu ekuzivikanwa kwemifananidzo tarisa chikomo chine huswa uti, "makwai". Zviri pachena kuti nei: mizhinji yemuenzaniso mapikicha e "makwai" anotorwa mumafuro maanogara, uye mumifananidzo iyi uswa hunotora nzvimbo yakawanda kupfuura diki chena fluffs, uye ndihwo huswa hunotorwa nehurongwa hwakakosha. .

Kune mimwe mienzaniso yakakomba. Imwe yazvino ronga yekuona gomarara reganda mumifananidzo. Zvakazoitika kuti dermatologists vanowanzotora mufananidzo mutongi pamwe chete nekuratidzwa kwekenza yeganda kunyora hukuru hwemaumbirwo. Iko hakuna vatongi mumuenzaniso mapikicha eganda rine hutano. Kune AI system, vatongi vakadaro (kunyanya, mapikseli atinotsanangura se "mutongi") ave mumwe wemisiyano pakati pemaseti emienzaniso, uye dzimwe nguva inonyanya kukosha kupfuura kuputika kuduku paganda. Saka hurongwa hwakagadzirwa kuona gomarara reganda dzimwe nguva hunozivikanwa vatongi panzvimbo.

Chinhu chakakosha apa ndechekuti sisitimu haina kunzwisisa semantic yezvairi kutarisa. Isu tinotarisa seti yepixels uye tinoona mairi gwai, ganda kana vatongi, asi iyo system ingori nhamba mutsara. Haaone nzvimbo ine mativi matatu, haaone zvinhu, maumbirwo, kana hwai. Anongoona maitiro mune data.

Chinetso chekuongorora matambudziko akadai ndechekuti neural network (iyo modhi inogadzirwa nemuchina wako wekudzidza system) ine zviuru zvemazana ezviuru zvemanodhi. Iko hakuna nzira iri nyore yekutarisa mumuenzaniso uye kuona kuti inoita sei sarudzo. Kuve nenzira yakadaro kunoreva kuti maitiro ari nyore zvakakwana kutsanangura mitemo yese nemaoko, pasina kushandisa muchina kudzidza. Vanhu vanonetsekana kuti kudzidza kwemichina kwave chimwe chinhu chebhokisi dema. (Ndichatsanangura zvishoma gare gare nei kuenzanisa uku kuchiri kwakanyanya.)

Iri, kazhinji, idambudziko rekusarura muhungwaru hwekugadzira kana kudzidza muchina: hurongwa hwekutsvaga mapatani mune data inogona kuwana maitiro asina kunaka, uye iwe ungasazviona. Uyu ndiwo hunhu hwakakosha hwetekinoroji, uye zviri pachena kune wese anoshanda nayo muzvidzidzo uye kumakambani makuru etekinoroji. Asi mhedzisiro yaro yakaoma, uye ndozvakaitawo mhinduro dzedu dzinogoneka kumhedzisiro iyoyo.

Ngatitange tataura nezvemhedzisiro.

AI inokwanisa, zvamazvirokwazvo kwatiri, kuita sarudzo mukufarira mamwe mapoka evanhu, zvichibva pahuwandu hukuru hwezviratidzo zvisingaoneki.

AI Bias Scenarios

Zvikuru zviri pachena uye zvinotyisa, dambudziko iri rinogona kuzviratidza pacharo kana zvasvika pakusiyana kwevanhu. Munguva pfupi yapfuura kwaiva nerunyerekupekuti Amazon yakaedza kuvaka muchina wekudzidza sisitimu yekutanga kuongororwa kwevanoda basa. Sezvo kune varume vakawanda pakati pevashandi veAmazon, mienzaniso ye "kuhaya vakabudirira" inowanzoitawo varume, uye pakanga paine varume vakawanda mukusarudzwa kwezvekutanga zvakakurudzirwa nehurongwa. Amazon yakaona izvi uye haina kuburitsa sisitimu mukugadzira.

Chinonyanya kukosha mumuenzaniso uyu ndechekuti hurongwa hwaive nerunyerekupe hwekufarira vanyoreri vechirume, zvisinei nekuti murume kana mukadzi haana kutaurwa pazvekutanga. Iyo sisitimu yakaona mamwe mapatani mumienzaniso ye "mihoro yakanaka": semuenzaniso, vakadzi vanogona kushandisa mazwi akakosha kutsanangura zvakaitwa, kana kuita zvinhu zvekuzvivaraidza nazvo. Ehe, sisitimu yaisaziva kuti "hockey" yaive chii, kana kuti "vanhu" vaive ani, kana kuti "budiriro" yaive yei - yakangoita ongororo yezvinyorwa. Asi maitiro aakaona angangotadza kucherechedzwa nevanhu, uye mamwe acho (somuenzaniso, idi rekuti vanhu vevarume vakasiyana vanotsanangura kubudirira nenzira dzakasiyana) zvimwe zvingatiomera kuona kunyangwe tikazvitarisa.

Kuwedzera - kuipa. Muchina wekudzidza sisitimu yakanaka kwazvo pakutsvaga gomarara paganda rakachenuruka inogona kusaita zvakanaka paganda rakasviba, kana zvinopesana. Kwete hazvo nekuda kwekurerekera, asi nekuti iwe ungangoda kuvaka yakaparadzana modhi kune rakasiyana ruvara rweganda, uchisarudza akasiyana maitiro. Michina yekudzidza masisitimu haichinjike kunyangwe munzvimbo yakamanikana sekuzivikanwa kwemufananidzo. Iwe unofanirwa kugadzirisa iyo sisitimu, dzimwe nguva nekungoyedza uye kukanganisa, kuti uwane yakanaka kubata pane iyo data yauri kufarira kudzamara wawana iko kurongeka kwaunoda. Asi izvo zvaungasacherechedze ndezvekuti hurongwa hwakarurama 98% yenguva neboka rimwe chete, uye chete 91% (kunyangwe yakanyatsojeka kupfuura kuongororwa kwevanhu) neimwe.

Parizvino ndashandisa kunyanya mienzaniso ine chekuita nevanhu nehunhu hwavo. Hurukuro yakapoteredza dambudziko iri inonyanya kutarisisa pane iyi nyaya. Asi zvakakosha kunzwisisa kuti kurerekera kune vanhu chikamu chedambudziko. Tichange tichishandisa muchina kudzidza kune zvakawanda zvinhu, uye sampling kukanganisa kuchave kwakakodzera kune ese. Kune rumwe rutivi, kana ukashanda nevanhu, kusarura mune data kunogona kunge kusina hukama navo.

Kuti tinzwisise izvi, ngatidzokerei kumuenzaniso wegomarara reganda uye tifunge nezvemikana mitatu yekufungidzira yekutadza kwehurongwa.

1. Heterogeneous distribution of people: nhamba isina kuenzana yemapikicha emhando dzakasiyana dzeganda, zvinotungamira kune manyepo enhema kana manyepo ekunyepa nekuda kwe pigmentation.
2. Iyo data inodzidziswa iyo system ine chinhu chinowanzoitika uye chakagoverwa zvakasiyana-siyana chisingabatanidzwe nevanhu uye chisina kukosha kwekuongorora: mutongi mumifananidzo yegomarara reganda kana uswa mumifananidzo yemakwai. Muchiitiko ichi, mhedzisiro ichave yakasiyana kana iyo system ikawana mapixels mumufananidzo wechimwe chinhu icho ziso remunhu rinozivisa se "mutongi".
3. Iyo data ine yechitatu-bato hunhu iyo munhu asingakwanisi kuona kunyangwe akaitsvaga.

Zvinorevei? Isu tinoziva a priori kuti data inogona kumiririra mapoka akasiyana evanhu zvakasiyana, uye padiki tinogona kuronga kutsvaga kusarudzika kwakadaro. Mune mamwe mazwi, pane zvakawanda zvezvikonzero zvemagariro zvekufunga kuti data nezvemapoka evanhu yatove nekumwe kusarura. Kana tikatarisa mufananidzo nemutongi, tichaona mutongi uyu - isu takangozvikanganwa kare, tichiziva kuti hazvina basa, uye tichikanganwa kuti hurongwa hahuzivi chinhu.

Asi zvakadini kana mapikicha ako ose eganda risina kunaka akatorwa muhofisi pasi pechiedza che incandescent, uye ganda rako rine utano rakatorwa pasi pechiedza chefluorescent? Ko kana, mushure mekunge wapedza kupfura ganda rine hutano, usati wapfura ganda risina kunaka, iwe wakagadziridza sisitimu yekushandisa pafoni yako, uye Apple kana Google yakachinja zvishoma algorithm yekudzikisa ruzha? Munhu haagoni kuona izvi, zvisinei kuti anotarisa sei zvinhu zvakadaro. Asi iyo muchina-wekushandisa sisitimu inozoona nekukurumidza uye kushandisa izvi. Hapana chaanoziva.

Parizvino tataura nezve spurious corelations, asi zvinogona zvakare kuti data racho rakarurama uye mhedzisiro yacho ndeyechokwadi, asi haudi kuishandisa nekuda kwehunhu, zvemutemo, kana manejimendi zvikonzero. Mamwe matunhu, somuenzaniso, haabvumiri vakadzi kuwana kuderedzwa painishuwarenzi yavo, kunyange zvazvo vakadzi vangava vatyairi vakachengeteka. Tinogona kufungidzira zviri nyore hurongwa hwekuti, kana uchiongorora nhoroondo yedata, yaizopa yakaderera njodzi kune mazita echikadzi. Zvakanaka, ngatibvise mazita kubva pakusarudza. Asi rangarira muenzaniso weAmazon: iyo sisitimu inogona kuona kuti murume kana mukadzi zvichienderana nezvimwe zvinhu (kunyangwe isingazive kuti murume kana mukadzi chii, kana kuti mota chii), uye hauzozvicherechedze izvi kudzamara mutongi aongorora zvakare mutero waunotora. kupa uye mhosva iwe ucharipiswa.

Pakupedzisira, zvinowanzofungidzirwa kuti tichashandisa maitiro akadaro chete kumapurojekiti anosanganisira vanhu uye kushamwaridzana kwevanhu. Izvi zvakaipa. Kana iwe ukagadzira maturbine egasi, iwe ungangoda kuisa muchina kudzidza kune telemetry inofambiswa nemakumi kana mazana emasensa pachigadzirwa chako (odhiyo, vhidhiyo, tembiricha, uye chero mamwe ma sensors anoburitsa data rinogona kuchinjika nyore nyore kugadzira muchina. kudzidza modhi). Nehunyengeri, unogona kuti, "Heino data kubva kune chiuru cheturbines yakatadza isati yakundikana, uye heino data kubva kune chiuru cheturbines isina kukundikana. Vaka modhi yekutaura kuti musiyano uri pakati pavo ndeupi. " Zvakanaka, zvino fungidzira kuti Siemens sensors yakaiswa pa 75% ye turbines yakaipa, uye chete 12% yezvakanaka (hapana hukama nekukundikana). Iyo sisitimu ichavaka modhi yekutsvaga turbines ine Nokia sensors. Maiwe!

Mufananidzo — Moritz Hardt, UC Berkeley

Kugadzirisa AI Bias

Chii chatingaita nezvazvo? Iwe unogona kutarisana nenyaya yacho kubva pamakona matatu:

1. Methodological rigor mukuunganidza uye kutonga data yekudzidzisa sisitimu.
2. Zvishandiso zvehunyanzvi zvekuongorora uye kuongorora maitiro emuenzaniso.
3. Rovedza, dzidzisa, uye chenjera paunenge uchishandisa muchina kudzidza muzvigadzirwa.

Pane joke mubhuku raMolière "The Bourgeois in the Nobility": mumwe murume akaudzwa kuti mabhuku akakamurwa kuva prose uye nhetembo, uye akafara kuona kuti akanga achitaura muprose hupenyu hwake hwose, asingazivi. Izvi zvingangove manzwiro anoita vezviverengero nhasi: vasingazvizive, vakatsaurira mabasa avo kune hungwaru hwekugadzira uye sampling kukanganisa. Kutsvaga sampling chikanganiso uye kunetsekana nacho harisi dambudziko idzva, isu tinongoda kurongeka takananga mhinduro yayo. Sezvataurwa pamusoro apa, mune zvimwe zviitiko zviri nyore kuita izvi nekudzidza matambudziko ane chekuita nevanhu data. Isu a priori vanofungidzira kuti tinogona kuva nerusaruro maererano nemapoka akasiyana-siyana evanhu, asi zvakaoma kuti isu kunyange kufungidzira kusarura pamusoro Siemens sensors.

Chii chitsva pane zvese izvi, chokwadi, ndechekuti vanhu havachaite ongororo yenhamba zvakananga. Inoitwa nemichina inogadzira yakakura, yakaoma modhi iyo yakaoma kunzwisisa. Nyaya yekubuda pachena ndiyo imwe yenyaya huru dzedambudziko rekurerekera divi rimwe chete. Isu tinotya kuti iyo sisitimu haina kungorerekera, asi kuti hapana nzira yekuona yakarerekera, uye kuti kudzidza kwemuchina kwakasiyana nedzimwe nzira dze otomatiki, dzinofanirwa kunge dzine matanho akajeka anonzwisisika anogona kuongororwa.

Pane matambudziko maviri pano. Tinogona kuramba tichikwanisa kuita imwe mhando yekuongorora yemashini ekudzidza masisitimu. Uye kuongorora chero imwe sisitimu haisi nyore.

Chekutanga, imwe yenzira yekutsvagisa yemazuva ano mumunda wekudzidza muchina ndeye kutsvaga nzira dzekuona kwakakosha mashandiro emashini ekudzidza masisitimu. Zvakadaro, kudzidza muchina (mumamiriro ayo azvino) inzvimbo nyowani yesainzi iri kuchinja nekukurumidza, saka usafunge kuti zvinhu zvisingagoneke nhasi hazvigone kukurumidza kuitika. Project OpenAI - muenzaniso unonakidza weizvi.

Chechipiri, iyo pfungwa yekuti iwe unogona kuyedza uye kunzwisisa maitiro ekuita sarudzo yearipo masisitimu kana masangano akanaka mudzidziso, asi saka-saizvozvo mukuita. Kunzwisisa kuti sarudzo dzinoitwa sei musangano guru hakusi nyore. Kunyangwe paine hurongwa hwekuita sarudzo, hazviratidzi kuti vanhu vanodyidzana sei, uye ivo pachavo kazhinji havana maitiro ane musoro, akarongeka ekuita sarudzo. Sezvakataurwa nemumwe wangu Vijay Pande, vanhuwo mabhokisi matema.

Tora chiuru chevanhu mumakambani akawanda anopindirana uye masangano, uye dambudziko rinowedzera kuoma. Isu tinoziva mushure mekuda kwekuti Space Shuttle yaizoparadzana pakudzoka, uye vanhu mukati meNASA vaive neruzivo rwakavapa chikonzero chekufunga kuti chimwe chinhu chakashata chingaitika, asi iyo system. kazhinji Ndaisaziva izvi. NASA ichangopfuura nekuongorora kwakafanana mushure mekurasikirwa nevhavha yayo yapfuura, uye zvakadaro yakarasikirwa neimwe nekuda kwechikonzero chakafanana. Zviri nyore kuita nharo kuti masangano nevanhu vanotevera mitemo yakajeka, ine musoro inogona kuedzwa, kunzwisiswa, uye kuchinjwa—asi ruzivo runoratidza zvakasiyana. Izvi"Gosplan's delusion".

Ini ndinowanzo enzanisa kudzidza kwemichina kune dhatabhesi, kunyanya hukama - tekinoroji nyowani yakakosha yakashandura kugona kwesainzi yekombuta nenyika yakaitenderedza, iyo yava chikamu chezvose, izvo zvatinoshandisa nguva dzose tisingazvizivi. Databases zvakare ine matambudziko, uye iwo emhando yakafanana: iyo sisitimu inogona kuvakwa pafungidziro dzakaipa kana data rakashata, asi zvichave zvakaoma kuona, uye vanhu vanoshandisa iyo system vanozoita zvainovaudza vasingabvunze mibvunzo. Kune majee akawanda ekare pamusoro pevateresi vakamboperetera zita rako zvisizvo, uye kuvanyengetedza kuti vagadzirise chikanganiso chacho kwakaoma zvikuru kupfuura kusandura zita rako chaizvoizvo. Kune nzira dzakawanda dzekufunga pamusoro peizvi, asi hazvisi pachena kuti ndezvipi zviri nani: sedambudziko rehunyanzvi muSQL, kana sebug mukuburitswa kweOracle, kana sekutadza kwemasangano ehurumende? Zvakaoma sei kuwana bug mukuita izvo zvaita kuti sisitimu ishaye typo yekururamisa chimiro? Izvi zvingave zvakaonekwa vanhu vasati vatanga kunyunyuta here?

Dambudziko iri rinoratidzwa zvakanyanya nengano kana vatyairi vanotyaira munzizi nekuda kwedata rekare mufambisi. Zvakanaka, mepu dzinoda kuvandudzwa nguva dzose. Asi TomTom ane mhosva yakawanda sei kuti mota yako yapeperetswa mugungwa?

Chikonzero chandinotaura izvi ndechekuti hongu, kusarudzika kwekudzidza kwemichina kuchaunza matambudziko. Asi matambudziko aya achafanana neaya atakatarisana nawo munguva yakapfuura, uye anogona kucherechedzwa nekugadziriswa (kana kusagadziriswa) nezve sezvataikwanisa kuita munguva yakapfuura. Naizvozvo, mamiriro ezvinhu umo AI bias inokonzeresa kukuvadza haigone kuitika kune vakuru vatsvagiri vanoshanda musangano rakakura. Zvingangodaro, imwe isingakoshi tekinoroji kontrakta kana software mutengesi anonyora chimwe chinhu pamabvi avo, vachishandisa yakavhurika sosi zvikamu, maraibhurari uye zvishandiso zvavasinganzwisise. Uye mutengi asina rombo rakanaka anotenga mutsara wekuti "artificial intelligence" mune tsananguro yechigadzirwa uye, pasina kubvunza chero mibvunzo, anoigovera kune vashandi vake vanobhadhara zvishoma, achivaraira kuti vaite zvinotaurwa neAI. Izvi ndizvo chaizvo zvakaitika nemadatabase. Iri harisi dambudziko rehungwaru hwekugadzira, kana kunyange dambudziko resoftware. Ichi ndicho chikonzero chemunhu.

mhedziso

Kudzidza kwemichina kunogona kuita chero chinhu chaunogona kudzidzisa imbwa - asi haungambove nechokwadi kuti chii chaizvo chawakadzidzisa imbwa.

Ndinowanzonzwa sekuti izwi rekuti "artificial intelligence" rinongopinda munzira yekukurukurirana seizvi. Izwi iri rinopa fungidziro yenhema yekuti takazvigadzira - kungwara uku. Kuti tiri munzira kuenda kuHAL9000 kana Skynet - chimwe chinhu chaicho anonzwisisa. Asi kwete. Iyi ingori michina, uye yakanyanya kunyatso kuenzanisa, toti, muchina wekuwachisa. Anowacha hembe zviri nani kudarika munhu, asi ukamuisira ndiro pane kuwacha, anozowacha. Ndiro dzacho dzichatochena. Asi izvi hazvizove izvo zvawaitarisira, uye izvi hazvizoitika nekuti sisitimu ine rusaruro maererano nendiro. Muchina wekuwachisa hauzive kuti zvii kana zvipfeko zvipi - ingori muenzaniso weotomatiki, zvine mutsindo hapana kusiyana nemaitiro aiitwa otomatiki kare.

Kunyangwe tichitaura nezve mota, ndege, kana dhatabhesi, masisitimu aya achave ane simba uye ane mashoma. Vachaenderana zvachose nemashandisiro anoita vanhu masisitimu aya, kuti vavariro dzavo dzakanaka here kana kuti dzakaipa, uye kuti vanonzwisisa zvakadii mashandiro avanoita.

Naizvozvo, kutaura kuti "ungwaru hwekugadzira masvomhu, saka haugone kuve nekurerekera" manyepo zvachose. Asi inhema zvakaenzana kutaura kuti kudzidza muchina "kunozviisa pasi pechisikigo." Kudzidza kwemichina kunowana mapatani mu data, uye kuti ndeapi mapatani aanowana zvinoenderana nedata, uye data rinoenderana nesu. Kungofanana nezvatinoita navo. Kudzidza nemuchina kunoita zvimwe zvinhu zvirinani kupfuura zvatingaite - asi imbwa, semuenzaniso, dzinoshanda zvakanyanya kupfuura vanhu pakuona zvinodhaka, icho chisiri chikonzero chekuzvishandisa sezvapupu uye kuita mitongo zvichienderana neuchapupu hwadzo. Uye imbwa, nenzira, dzakangwara kupfuura chero muchina wekudzidza system.

Shanduro: Diana Letskaya.
Kugadzirisa: Aleksey Ivanov.
Nharaunda @PonchikNews.

Source: www.habr.com