{"id":32781,"date":"2019-10-31T21:48:54","date_gmt":"2019-10-31T18:48:54","guid":{"rendered":"https:\/\/prohoster.info\/blog\/o-predvzyatosti-iskusstvennogo-intellekta\/"},"modified":"2019-10-31T21:48:54","modified_gmt":"2019-10-31T18:48:54","slug":"o-predvzyatosti-iskusstvennogo-intellekta","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sk\/blog\/news\/o-predvzyatosti-iskusstvennogo-intellekta","title":{"rendered":"O predpojatosti umelej inteligencie","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"<p><img decoding=\"async\" alt=\"O predpojatosti umelej inteligencie\" src=\"\/wp-content\/uploads\/2019\/04\/05a731677e1231acceff953db8e1e26a.jpg\" style=\"display:block;margin: 0 auto;\" \/><\/p>\n<p><\/p>\n<h2 id=\"tldr\">zhrnutie:<\/h2>\n<p><\/p>\n<ul>\n<li>Strojov\u00e9 u\u010denie h\u013ead\u00e1 vzory v d\u00e1tach. Ale umel\u00e1 inteligencia m\u00f4\u017ee by\u0165 \u201epredpojat\u00e1\u201c \u2013 to znamen\u00e1, \u017ee n\u00e1jde nespr\u00e1vne vzory. Napr\u00edklad, syst\u00e9m na detekciu rakoviny ko\u017ee z fotografie m\u00f4\u017ee osobitne zoh\u013eadni\u0165 sn\u00edmky uroben\u00e9 v lek\u00e1rskej ordin\u00e1cii. Strojov\u00e9 u\u010denie nevie <em>rozumie\u0165<\/em>: jeho algoritmy len odha\u013euj\u00fa vzory v \u010d\u00edslach a ak d\u00e1ta nie s\u00fa reprezentat\u00edvne, tak\u00fd bude aj v\u00fdsledok ich spracovania. A odha\u013eova\u0165 tak\u00e9to chyby m\u00f4\u017ee by\u0165 zlo\u017eit\u00e1 \u00faloha kv\u00f4li samotnej mechanike strojov\u00e9ho u\u010denia. <noindex><a rel=\"nofollow\" name=\"habracut\"><\/a><\/noindex><\/li>\n<li>Najzrejmej\u0161ou a najstra\u0161nej\u0161ou probl\u00e9movou oblastou je \u013eudsk\u00e1 rozmanitos\u0165. Existuje mno\u017estvo d\u00f4vodov, pre\u010do d\u00e1ta o \u013eu\u010foch m\u00f4\u017eu str\u00e1ca\u0165 objekt\u00edvnos\u0165 u\u017e v f\u00e1ze zberu. Ale netreba si myslie\u0165, \u017ee tento probl\u00e9m sa t\u00fdka len \u013eud\u00ed: podobn\u00e9 \u0165a\u017ekosti sa objavuj\u00fa aj pri pokuse odhali\u0165 povode\u0148 v sklade alebo poruchu plynov\u00e9ho turb\u00edny. Niektor\u00e9 syst\u00e9my m\u00f4\u017eu ma\u0165 predsudky vo\u010di farbe poko\u017eky, in\u00e9 bud\u00fa ma\u0165 zaujat\u00fd pr\u00edstup k senzorom Siemens. <\/li>\n<li>Tieto probl\u00e9my nie s\u00fa nov\u00e9 pre strojov\u00e9 u\u010denie a nie s\u00fa typick\u00e9 len pre neho. Nespr\u00e1vne predpoklady sa robia v ak\u00fdchko\u013evek zlo\u017eit\u00fdch \u0161trukt\u00farach a porozumie\u0165, pre\u010do bolo prijat\u00e9 konkr\u00e9tne rozhodnutie, je v\u017edy zlo\u017eit\u00e9. S t\u00fdmto treba bojova\u0165 komplexne: vytv\u00e1ra\u0165 n\u00e1stroje a procesy na overovanie \u2013 a vzdel\u00e1va\u0165 pou\u017e\u00edvate\u013eov, aby neslepo nasledovali odpor\u00fa\u010dania AI. Strojov\u00e9 u\u010denie naozaj rob\u00ed niektor\u00e9 veci ove\u013ea lep\u0161ie ako my \u2013 ale psy, napr\u00edklad, s\u00fa omnoho efekt\u00edvnej\u0161ie ako \u013eudia pri odha\u013eovan\u00ed drog, \u010do nie je d\u00f4vod na ich anga\u017eovanie ako svedkov a na z\u00e1klade ich v\u00fdpoved\u00ed vyn\u00e1\u0161anie rozsudkov. A psy, ak m\u00f4\u017eem podotkn\u00fa\u0165, s\u00fa ove\u013ea m\u00fadrej\u0161ie ako ak\u00fdko\u013evek syst\u00e9m strojov\u00e9ho u\u010denia. <\/li>\n<\/ul>\n<p>Strojov\u00e9 u\u010denie dnes predstavuje jeden z najd\u00f4le\u017eitej\u0161\u00edch z\u00e1kladn\u00fdch technick\u00fdch trendov. Je to jeden z hlavn\u00fdch sp\u00f4sobov, ktor\u00fdmi technol\u00f3gia zmen\u00ed okolit\u00fd svet v nasleduj\u00facom desa\u0165ro\u010d\u00ed. Niektor\u00e9 aspekty t\u00fdchto zmien vzbudzuj\u00fa obavy. Napr\u00edklad potenci\u00e1lny vplyv strojov\u00e9ho u\u010denia na trh pr\u00e1ce alebo jeho vyu\u017eitie na neetick\u00e9 \u00fa\u010dely (napr\u00edklad autorit\u00e1rskymi re\u017eimami). Existuje e\u0161te jeden probl\u00e9m, ktor\u00e9mu sa tento pr\u00edspevok venuje: <strong>predpojatost' umelej inteligencie<\/strong>. <\/p>\n<p><\/p>\n<p>Toto nie je jednoduch\u00fd pr\u00edbeh. <\/p>\n<p><\/p>\n<p><img decoding=\"async\" alt=\"O predpojatosti umelej inteligencie\" src=\"\/wp-content\/uploads\/2019\/04\/d692fbfff47614758ab1d8e1516b2742.jpg\" style=\"display:block;margin: 0 auto;\" \/><br \/>\n<em>AI od Google dok\u00e1\u017ee n\u00e1js\u0165 ma\u010dky. T\u00e1to spr\u00e1va z roku 2012 bola v tom \u010dase nie\u010d\u00edm v\u00fdnimo\u010dn\u00fdm.<\/em><\/p>\n<p><\/p>\n<h2 id=\"chto-takoe-predvzyatost-ii\">\u010co je \u201epredpojatos\u0165 AI\u201c?<\/h2>\n<p><\/p>\n<blockquote><p><em>\u201eSurov\u00e9 d\u00e1ta\u201c s\u00fa z\u00e1rove\u0148 oxymoron a zl\u00fd n\u00e1pad; d\u00e1ta je potrebn\u00e9 dobre a starostlivo spracova\u0165.<\/em> \u2014Jeffrey Boker <\/p><\/blockquote>\n<p>A\u017e do niekedy v roku 2013 bolo potrebn\u00e9 pri vytv\u00e1ran\u00ed syst\u00e9mu, ktor\u00fd napr\u00edklad rozpozn\u00e1va ma\u010dky na fotografi\u00e1ch, popisova\u0165 logick\u00e9 kroky. Ako n\u00e1js\u0165 rohy na obr\u00e1zku, rozpozna\u0165 o\u010di, analyzova\u0165 text\u00fary na pr\u00edtomnos\u0165 ko\u017eu\u0161iny, spo\u010d\u00edta\u0165 labky a tak \u010falej. Potom zhroma\u017edi\u0165 v\u0161etky komponenty \u2014 a zisti\u0165, \u017ee to v\u0161etko v podstate nefunguje. Pripom\u00ednalo to mechanick\u00e9ho ko\u0148a \u2014 teoreticky to je mo\u017en\u00e9 vyrobi\u0165, ale v praxi je to pr\u00edli\u0161 zlo\u017eit\u00e9 na opis. Na v\u00fdstupe m\u00e1te stovky (alebo dokonca tis\u00edce) ru\u010dne nap\u00edsan\u00fdch pravidiel. A ani jednu funguj\u00facu model. <\/p>\n<p><\/p>\n<p>S pr\u00edchodom strojov\u00e9ho u\u010denia sme prestali pou\u017e\u00edva\u0165 \u201eru\u010dn\u00e9\u201c pravidl\u00e1 na rozpozn\u00e1vanie toho-ktor\u00e9ho objektu. Namiesto toho vezmeme tis\u00edc vzoriek \u201etoho\u201c, X, tis\u00edc vzoriek \u201ein\u00e9ho\u201c, Y, a prin\u00fatime po\u010d\u00edta\u010d vytvori\u0165 model na z\u00e1klade ich \u0161tatistickej anal\u00fdzy. Potom tejto modelu poskytneme nejak\u00fd pr\u00edklad d\u00e1t a ona s ur\u010ditou presnos\u0165ou ur\u010d\u00ed, \u010di patr\u00ed do jednej z mno\u017e\u00edn. Strojov\u00e9 u\u010denie generuje model na z\u00e1klade d\u00e1t, nie pomocou \u010dloveka, ktor\u00fd ho nap\u00ed\u0161e. V\u00fdsledky s\u00fa p\u00f4sobiv\u00e9, najm\u00e4 v oblasti rozpozn\u00e1vania obr\u00e1zkov a vzorov, a preto sa cel\u00e1 technologick\u00e1 industr\u00eda teraz pres\u00fava k strojov\u00e9mu u\u010deniu (ML). <\/p>\n<p><\/p>\n<p>Ale nie je v\u0161etko tak jednoduch\u00e9. V re\u00e1lnom svete va\u0161e tis\u00edce pr\u00edkladov X alebo Y obsahuj\u00fa aj A, B, J, L, O, R a dokonca aj L. M\u00f4\u017eu by\u0165 rozlo\u017een\u00e9 nerovnomerne, a niektor\u00e9 z nich sa m\u00f4\u017eu vyskytova\u0165 tak \u010dasto, \u017ee syst\u00e9m na ne prirad\u00ed v\u00e4\u010d\u0161iu pozornos\u0165 ako objektom, ktor\u00e9 v\u00e1s zauj\u00edmaj\u00fa.<\/p>\n<p><\/p>\n<p>\u010co to znamen\u00e1 v praxi? M\u00f4j ob\u013e\u00faben\u00fd pr\u00edklad je, ke\u010f syst\u00e9my rozpozn\u00e1vania obr\u00e1zkov <noindex><a rel=\"nofollow\" href=\"http:\/\/aiweirdness.com\/post\/171451900302\/do-neural-nets-dream-of-electric-sheep\">d\u00edvaj\u00fa sa na tr\u00e1vnat\u00fd kopec a hovoria: \u201eovca\u201c<\/a><\/noindex>. Je jasn\u00e9, pre\u010do: v\u00e4\u010d\u0161ina fotografi\u00ed-pr\u00edkladov \u201eovce\u201c bola vyroben\u00e1 na l\u00fakach, kde \u017eij\u00fa, a na t\u00fdchto obr\u00e1zkoch tr\u00e1va zaber\u00e1 ove\u013ea viac miesta ako mal\u00e9 biele nad\u00fdchan\u00e9 tvory, a pr\u00e1ve tr\u00e1vu syst\u00e9my pova\u017euj\u00fa za najd\u00f4le\u017eitej\u0161iu. <\/p>\n<p><\/p>\n<p>S\u00fa aj v\u00e1\u017enej\u0161ie pr\u00edklady. Z ned\u00e1vnych \u2014 jeden <noindex><a rel=\"nofollow\" href=\"https:\/\/www.medpagetoday.com\/dermatology\/generaldermatology\/70513\">projekt<\/a><\/noindex> o detekciu rakoviny ko\u017ee na fotografi\u00e1ch. Uk\u00e1zalo sa, \u017ee dermatol\u00f3govia \u010dasto fotografuj\u00fa prav\u00edtko spolu s prejavmi rakoviny ko\u017ee, aby zdokumentovali ve\u013ekos\u0165 \u00fatvarov. Na pr\u00edkladoch fotografi\u00ed zdravej ko\u017ee prav\u00edtka nie s\u00fa. Pre syst\u00e9m AI sa tieto prav\u00edtka (presnej\u0161ie, pixely, ktor\u00e9 ur\u010d\u00edme ako \u201eprav\u00edtko\u201c) stali jedn\u00fdm z rozli\u0161ovac\u00edch znakov medzi s\u00fabormi vzoriek, a niekedy s\u00fa d\u00f4le\u017eitej\u0161ie ako mal\u00e9 vyr\u00e1\u017eky na ko\u017ei. Tak syst\u00e9m navrhnut\u00fd na rozpozn\u00e1vanie rakoviny ko\u017ee niekedy namiesto toho rozpoznal prav\u00edtka. <\/p>\n<p><\/p>\n<p>K\u013e\u00fa\u010dov\u00fdm aspektom je to, \u017ee syst\u00e9m nem\u00e1 semantick\u00e9 pochopenie toho, na \u010do sa pozer\u00e1. My sa pozer\u00e1me na s\u00fabor pixelov a vid\u00edme v nich ovcu, ko\u017eu alebo prav\u00edtka, zatia\u013e \u010do syst\u00e9m vid\u00ed len \u010d\u00edseln\u00fd re\u0165azec. On nevid\u00ed trojrozmern\u00fd priestor, nevid\u00ed ani objekty, ani text\u00fary, ani ovce. Vn\u00edma len vzory v d\u00e1tach. <\/p>\n<p><\/p>\n<p>Zlo\u017eitost diagnostiky tak\u00fdchto probl\u00e9mov spo\u010d\u00edva v tom, \u017ee neur\u00f3nov\u00e1 sie\u0165 (model vytvoren\u00fd va\u0161\u00edm syst\u00e9mom strojov\u00e9ho u\u010denia) pozost\u00e1va zo tis\u00edcov, stotis\u00edcov uzlov. Neexistuje jednoduch\u00fd sp\u00f4sob, ako nazrie\u0165 do modelu a zisti\u0165, ako prij\u00edma rozhodnutia. Existencia tak\u00e9ho sp\u00f4sobu by znamenala, \u017ee proces je dostato\u010dne jednoduch\u00fd na to, aby sa dali v\u0161etky pravidl\u00e1 pop\u00edsa\u0165 manu\u00e1lne, bez pou\u017eitia strojov\u00e9ho u\u010denia. \u013dudia sa ob\u00e1vaj\u00fa, \u017ee strojov\u00e9 u\u010denie sa stalo ak\u00fdmsi \u201e\u010diernym boxom\u201c. (Nesk\u00f4r vysvetl\u00edm, pre\u010do je toto porovnanie trochu prehnan\u00e9.) <\/p>\n<p><\/p>\n<p>To, v z\u00e1kladn\u00fdch rysoch, je probl\u00e9m zaujatosti umel\u00e9ho inteligencie alebo strojov\u00e9ho u\u010denia: syst\u00e9m na h\u013eadanie vzorov v d\u00e1tach m\u00f4\u017ee n\u00e1js\u0165 nespr\u00e1vne vzory, a vy si toho nemus\u00edte v\u0161imn\u00fa\u0165. To je z\u00e1kladn\u00e1 vlastnos\u0165 technol\u00f3gie a je to jasn\u00e9 ka\u017ed\u00e9mu, kto s \u0148ou pracuje v vedeck\u00fdch kruhoch a vo ve\u013ek\u00fdch technologick\u00fdch firm\u00e1ch. Ale jej n\u00e1sledky s\u00fa zlo\u017eit\u00e9 a na\u0161e mo\u017en\u00e9 rie\u0161enia t\u00fdchto n\u00e1sledkov s\u00fa tie\u017e. <\/p>\n<p><\/p>\n<p>Najprv sa pozrime na n\u00e1sledky. <\/p>\n<p><\/p>\n<p><img decoding=\"async\" alt=\"O predpojatosti umelej inteligencie\" src=\"\/wp-content\/uploads\/2019\/04\/43748412bd82ae53a9c2e47afec9c655.jpg\" style=\"display:block;margin: 0 auto;\" \/><br \/>\n<em>AI m\u00f4\u017ee implicitne robi\u0165 rozhodnutia v prospech ur\u010dit\u00fdch kateg\u00f3ri\u00ed \u013eud\u00ed, zalo\u017een\u00e9 na mno\u017estve nepostrehnute\u013en\u00fdch sign\u00e1lov.<\/em><\/p>\n<p><\/p>\n<h2 id=\"scenarii-predvzyatosti-ii\">Scen\u00e1re zaujatosti AI<\/h2>\n<p><\/p>\n<p>Najzjavnej\u0161ie a najdesivej\u0161ie je, \u017ee tento probl\u00e9m sa m\u00f4\u017ee prejavi\u0165, ke\u010f ide o \u013eudsk\u00fa rozmanitos\u0165. Ned\u00e1vno <noindex><a rel=\"nofollow\" href=\"https:\/\/www.reuters.com\/article\/us-amazon-com-jobs-automation-insight\/amazon-scraps-secret-ai-recruiting-tool-that-showed-bias-against-women-idUSKCN1MK08G\">sa objavili spr\u00e1vy<\/a><\/noindex>, \u017ee Amazon sa pok\u00fasil vytvori\u0165 syst\u00e9m strojov\u00e9ho u\u010denia na prvotn\u00e9 hodnotenie uch\u00e1dza\u010dov o zamestnanie. Ke\u010f\u017ee medzi zamestnancami Amazonu je viac mu\u017eov, pr\u00edklady \"\u00faspe\u0161n\u00e9ho zamestnania\" s\u00fa taktie\u017e \u010dastej\u0161ie mu\u017esk\u00e9ho pohlavia, a v s\u00fabore \u017eivotopisov, ktor\u00e9 syst\u00e9m navrhol, bolo viac mu\u017eov. Amazon si to v\u0161imol a syst\u00e9m neuviedol do produkcie. <\/p>\n<p><\/p>\n<p>Najd\u00f4le\u017eitej\u0161ie na tomto pr\u00edklade je, \u017ee syst\u00e9m sa \u00fadajne uprednost\u0148oval uch\u00e1dza\u010dov mu\u017esk\u00e9ho pohlavia, aj ke\u010f pohlavie nebolo v \u017eivotopise uveden\u00e9.<strong> Syst\u00e9m zaznamenal in\u00e9 vzory v pr\u00edkladoch \"\u00faspe\u0161n\u00e9ho zamestnania\": napr\u00edklad, \u017eeny m\u00f4\u017eu pou\u017e\u00edva\u0165 \u0161peci\u00e1lne slov\u00e1 na opis svojich \u00faspechov alebo ma\u0165 \u0161peci\u00e1lne kon\u00ed\u010dky. Samozrejme, syst\u00e9m nevedel, \u010do je \"hokej\", ani kto s\u00fa \"\u013eudia\", ani \u010do je \"\u00faspech\" \u2014 robil len \u0161tatistick\u00fa anal\u00fdzu textu.<\/strong> Ale vzory, ktor\u00e9 videl, by pravdepodobne ostali nepov\u0161imnut\u00e9 \u010dlovekom a niektor\u00e9 z nich (napr\u00edklad to, \u017ee mu\u017ei a \u017eeny opisuj\u00fa \u00faspech r\u00f4znymi sp\u00f4sobmi) by sme pravdepodobne \u0165a\u017eko spozorovali, aj keby sme na ne pozerali. <\/p>\n<p><\/p>\n<p>\u010ealej \u2014 je to hor\u0161ie. Syst\u00e9m strojov\u00e9ho u\u010denia, ktor\u00fd ve\u013emi dobre rozpozn\u00e1va rakovinu na blednej pleti, m\u00f4\u017ee fungova\u0165 hor\u0161ie s tmavou ple\u0165ou, alebo naopak. Nie nutne kv\u00f4li zaujatosti, ale preto, \u017ee pravdepodobne potrebujete vytvori\u0165 pre in\u00fd odtie\u0148 pleti samostatn\u00fd model, ktor\u00fd vyberie in\u00e9 charakteristiky.<strong> Syst\u00e9my strojov\u00e9ho u\u010denia nie s\u00fa zamie\u0148ate\u013en\u00e9 ani v takej \u00fazkej oblasti, ako je rozpozn\u00e1vanie obrazov.<\/strong> Mus\u00edte nastavova\u0165 syst\u00e9m, niekedy len cez experimentovanie, aby ste spr\u00e1vne postrehli charakteristiky v \u00fadajoch, ktor\u00e9 v\u00e1s zauj\u00edmaj\u00fa, a\u017e k\u00fdm nedosiahnete po\u017eadovan\u00fa presnos\u0165. Ale nemus\u00edte si v\u0161imn\u00fa\u0165, \u017ee syst\u00e9m je vo 98 % presn\u00fd pri pr\u00e1ci s jednou skupinou a len v 91 % (aj ke\u010f je to presnej\u0161ie ako anal\u00fdza vykonan\u00e1 \u010dlovekom) \u2014 s inou. <\/p>\n<p><\/p>\n<p>Doteraz som pou\u017e\u00edval najm\u00e4 pr\u00edklady t\u00fdkaj\u00face sa \u013eud\u00ed a ich charakterist\u00edk. Na tejto t\u00e9me sa diskusia okolo tohto probl\u00e9mu preva\u017ene s\u00fastre\u010fuje. Ale je d\u00f4le\u017eit\u00e9 pochopi\u0165, \u017ee zaujatost vo\u010di \u013eu\u010fom je len \u010das\u0165ou probl\u00e9mu. Budeme pou\u017e\u00edva\u0165 strojov\u00e9 u\u010denie na mno\u017estvo vec\u00ed a chyba v\u00fdberu bude relevantn\u00e1 pre v\u0161etky z nich. Na druhej strane, ak pracujete s \u013eu\u010fmi, m\u00f4\u017ee by\u0165 zaujatost \u00fadajov spojen\u00e1 nie s nimi. <\/p>\n<p><\/p>\n<p>Aby sme to pochopili, vr\u00e1\u0165me sa k pr\u00edkladu s rakovinou ko\u017ee a zv\u00e1\u017eme tri hypotetick\u00e9 mo\u017enosti zlyhania syst\u00e9mu. <\/p>\n<p><\/p>\n<ol>\n<li>Nerovnomern\u00e9 rozlo\u017eenie \u013eud\u00ed: nevyv\u00e1\u017een\u00fd po\u010det fotografi\u00ed pleti r\u00f4znych odtie\u0148ov, \u010do vedie k falo\u0161ne pozit\u00edvnym alebo falo\u0161ne negat\u00edvnym v\u00fdsledkom t\u00fdkaj\u00facim sa pigment\u00e1cie. <\/li>\n<li>Data, na ktor\u00fdch sa syst\u00e9m tr\u00e9nuje, obsahuj\u00fa \u010dasto sa vyskytuj\u00facu a nerovnomerne rozlo\u017een\u00fa charakteristiku, ktor\u00e1 nie je spojen\u00e1 s \u013eu\u010fmi a nem\u00e1 diagnostick\u00fa hodnotu: prav\u00edtko na fotografi\u00e1ch prejavov rakoviny ko\u017ee alebo tr\u00e1vu na fotografi\u00e1ch oviec. V takom pr\u00edpade sa v\u00fdsledok bude l\u00ed\u0161i\u0165, ak syst\u00e9m na obr\u00e1zku n\u00e1jde pixely nie\u010doho, \u010do by \u013eudsk\u00e9 oko identifikovalo ako \"prav\u00edtko\". <\/li>\n<li>Data obsahuj\u00fa vonkaj\u0161iu charakteristiku, ktor\u00fa \u010dlovek nem\u00f4\u017ee vidie\u0165, ani keby ju h\u013eadal. <\/li>\n<\/ol>\n<p><\/p>\n<p>\u010co to znamen\u00e1? A priory vieme, \u017ee d\u00e1ta m\u00f4\u017eu r\u00f4zne reprezentova\u0165 r\u00f4zne skupiny \u013eud\u00ed, a minim\u00e1lne m\u00f4\u017eeme napl\u00e1nova\u0165 h\u013eadanie podobn\u00fdch v\u00fdnimiek. In\u00fdmi slovami, existuje mnoho soci\u00e1lnych d\u00f4vodov predpoklada\u0165, \u017ee d\u00e1ta o skupin\u00e1ch \u013eud\u00ed u\u017e obsahuj\u00fa ist\u00e9 predsudky. Ak sa pozrieme na fotografiu s prav\u00edtkom, uvid\u00edme toto prav\u00edtko \u2014 jednoducho sme ho predt\u00fdm ignorovali, preto\u017ee sme vedeli, \u017ee nem\u00e1 v\u00fdznam, a zabudli sme, \u017ee syst\u00e9mu ni\u010d nie je zn\u00e1me.<\/p>\n<p><\/p>\n<p>Ale \u010do ak s\u00fa v\u0161etky va\u0161e fotografie nezdravej pleti vyroben\u00e9 v kancel\u00e1rii, kde sa pou\u017e\u00edvaj\u00fa \u017eiarovky, a zdravej \u2014 pri fluorescen\u010dnom svetle? \u010co ak ste, ke\u010f ste skon\u010dili so sn\u00edman\u00edm zdravej pleti, pred sn\u00edman\u00edm nezdravej aktualizovali opera\u010dn\u00fd syst\u00e9m na telef\u00f3ne, a Apple alebo Google trochu zmenil algoritmus potl\u00e1\u010dania \u0161umu? \u010clovek to nemus\u00ed zaregistrova\u0165, bez oh\u013eadu na to, ko\u013eko by tak\u00fdchto vlastnost\u00ed h\u013eadal. Syst\u00e9m strojov\u00e9ho u\u010denia to okam\u017eite uvid\u00ed a vyu\u017eije. On o ni\u010dom nevie. <\/p>\n<p><\/p>\n<p>Zatia\u013e sme hovorili o falo\u0161n\u00fdch korel\u00e1ci\u00e1ch, ale m\u00f4\u017ee sa sta\u0165 aj to, \u017ee \u00fadaje s\u00fa presn\u00e9 a v\u00fdsledky spr\u00e1vne, no z etick\u00fdch, pr\u00e1vnych alebo mana\u017e\u00e9rskych d\u00f4vodov ich nechcete pou\u017e\u00edva\u0165. V niektor\u00fdch jurisdikci\u00e1ch, napr\u00edklad, nie je povolen\u00e9 poskytova\u0165 \u017een\u00e1m z\u013eavu na poistenie, aj ke\u010f \u017eeny m\u00f4\u017eu bezpe\u010dnej\u0161ie \u0161of\u00e9rova\u0165. \u013dahko si m\u00f4\u017eeme predstavi\u0165 syst\u00e9m, ktor\u00fd pri anal\u00fdze historick\u00fdch \u00fadajov pridel\u00ed \u017eensk\u00fdm men\u00e1m ni\u017e\u0161\u00ed koeficient rizika. Dobre, odstr\u00e1\u0148me men\u00e1 z v\u00fdberu. Ale spome\u0148te si na pr\u00edklad s Amazonom: syst\u00e9m m\u00f4\u017ee ur\u010di\u0165 pohlavie na z\u00e1klade in\u00fdch faktorov (aj ke\u010f nevie, \u010do je pohlavie, a \u010do je auto), a vy si toho nev\u0161imnete, k\u00fdm regul\u00e1tor sp\u00e4tne neanalyzuje pon\u00fakan\u00e9 tarifn\u00e9 pl\u00e1ny a nevym\u00f4\u017ee od v\u00e1s pokutu. <\/p>\n<p><\/p>\n<p>Nakoniec sa \u010dasto predpoklad\u00e1, \u017ee tak\u00e9to syst\u00e9my budeme pou\u017e\u00edva\u0165 iba pre projekty, ktor\u00e9 s\u00favisia s \u013eu\u010fmi a soci\u00e1lnymi interakciami. To nie je pravda. Ak vyr\u00e1bate plynov\u00e9 turb\u00edny, ur\u010dite budete chcie\u0165 aplikova\u0165 strojov\u00e9 u\u010denie na telemetriu, ktor\u00fa pren\u00e1\u0161a desiatky alebo stovky senzorov na va\u0161om produkte (audio-, video-, teplotn\u00e9 a ak\u00e9ko\u013evek in\u00e9 senzory generuj\u00fa \u00fadaje, ktor\u00e9 sa ve\u013emi \u013eahko daj\u00fa prisp\u00f4sobi\u0165 na vytvorenie modelu strojov\u00e9ho u\u010denia). Hypoteticky m\u00f4\u017eete poveda\u0165: \"Tu s\u00fa \u00fadaje o tis\u00edcke poruchov\u00fdch turb\u00edn, z\u00edskan\u00e9 pred ich poruchou, a tu s\u00fa \u00fadaje z tis\u00edcky turb\u00edn, ktor\u00e9 sa neporuchovali. Vytvorte model, ktor\u00fd ur\u010d\u00ed, v \u010dom je medzi nimi rozdiel.\" A teraz si predstavte, \u017ee senzory Siemens s\u00fa na 75 % zl\u00fdch turb\u00edn a iba na 12 % dobr\u00fdch (pri\u010dom nie je \u017eiadna s\u00favislos\u0165 so zlyhaniami). Syst\u00e9m vytvor\u00ed model, aby na\u0161iel turb\u00edny so senzormi Siemens. Ups!<\/p>\n<p><\/p>\n<p><img decoding=\"async\" alt=\"O predpojatosti umelej inteligencie\" src=\"\/wp-content\/uploads\/2019\/04\/e8097e4913756fdf04826a6628bc6836.jpg\" style=\"display:block;margin: 0 auto;\" \/><br \/>\nObr\u00e1zok \u2014 Moritz Hardt, UC Berkeley<\/p>\n<p><\/p>\n<h2 id=\"upravlenie-predvzyatostyu-ii\">Riadenie predsudkov AI<\/h2>\n<p><\/p>\n<p>\u010co s t\u00fdm m\u00f4\u017eeme urobi\u0165? M\u00f4\u017eeme sa na ot\u00e1zku pozrie\u0165 z troch str\u00e1n:<\/p>\n<p><\/p>\n<ol>\n<li>Metodologick\u00e1 pr\u00edsnos\u0165 pri zbieran\u00ed a spr\u00e1ve \u00fadajov pre u\u010denie syst\u00e9mu.<\/li>\n<li>Technick\u00e9 n\u00e1stroje na anal\u00fdzu a diagnostiku spr\u00e1vania modelu.<\/li>\n<li>\u0160kolenie, vzdel\u00e1vanie a opatrnos\u0165 pri implement\u00e1cii strojov\u00e9ho u\u010denia do produktov.<\/li>\n<\/ol>\n<p><\/p>\n<p>V knihe \u201eMestsk\u00fd \u010dlovek v \u0161\u013eachte\u201c od Moli\u00e8ra je vtip: jedn\u00e9mu mu\u017eovi povedali, \u017ee literat\u00fara sa del\u00ed na pr\u00f3zu a po\u00e9ziu, a on s obdivom zist\u00ed, \u017ee cel\u00fd \u017eivot hovoril pr\u00f3zou, ani o tom nevediac. Pravdepodobne sa \u0161tatistiky c\u00edtia dnes podobne: bez toho, aby si to uvedomili, zasv\u00e4tili svoje kari\u00e9ry umelej inteligencii a chybe vzorkovania. H\u013eada\u0165 chybu vzorkovania a tr\u00e1pi\u0165 sa kv\u00f4li nej \u2014 to nie je nov\u00e1 ot\u00e1zka, mus\u00edme sa k jej rie\u0161eniu postavi\u0165 systematicky. Ako u\u017e bolo spomenut\u00e9, v niektor\u00fdch pr\u00edpadoch je naozaj jednoduch\u0161ie to robi\u0165, ke\u010f sk\u00famame probl\u00e9my spojen\u00e9 s \u00fadajmi o \u013eu\u010foch. A priori predpoklad\u00e1me, \u017ee m\u00f4\u017eeme ma\u0165 predsudky vo\u010di r\u00f4znym skupin\u00e1m \u013eud\u00ed, ale predsudok vo\u010di senzorom Siemens si \u0165a\u017eko dok\u00e1\u017eeme predstavi\u0165.<\/p>\n<p><\/p>\n<p>Nov\u00e9 na tom v\u0161etkom je, \u017ee \u013eudia sa u\u017e priamo nezap\u00e1jaj\u00fa do \u0161tatistickej anal\u00fdzy. T\u00fa vykon\u00e1vaj\u00fa stroje, ktor\u00e9 vytv\u00e1raj\u00fa ve\u013ek\u00e9 zlo\u017een\u00e9 modely, \u0165a\u017eko pochopite\u013en\u00e9. Ot\u00e1zka transparentnosti je jedn\u00fdm z hlavn\u00fdch aspektov probl\u00e9mu zaujatosti. Boj\u00edme sa, \u017ee syst\u00e9m nie je len zaujat\u00fd, ale \u017ee neexistuje \u017eiadna mo\u017enos\u0165 odhali\u0165 jeho predpojatosti, a \u017ee strojov\u00e9 u\u010denie sa t\u00fdm l\u00ed\u0161i od in\u00fdch foriem automatiz\u00e1cie, ktor\u00e9 sa predpokladaj\u00fa, \u017ee pozost\u00e1vaj\u00fa z jasn\u00fdch logick\u00fdch krokov, ktor\u00e9 je mo\u017en\u00e9 overi\u0165. <\/p>\n<p><\/p>\n<p>S\u00fa tu dve probl\u00e9my. Mo\u017eno by sme mohli vykona\u0165 ur\u010dit\u00fd audit syst\u00e9mov strojov\u00e9ho u\u010denia. A audit akejko\u013evek inej s\u00fastavy nie je vlastne o ni\u010d jednoduch\u0161\u00ed.<\/p>\n<p><\/p>\n<p>Po prv\u00e9, jedn\u00fdm z smerov s\u00fa\u010dasn\u00e9ho v\u00fdskumu v oblasti strojov\u00e9ho u\u010denia je h\u013eadanie met\u00f3d, ako identifikova\u0165 d\u00f4le\u017eit\u00e9 funkcie syst\u00e9mov strojov\u00e9ho u\u010denia. Strojov\u00e9 u\u010denie (v jeho aktu\u00e1lnom stave) je \u00faplne nov\u00e1 oblas\u0165 vedy, ktor\u00e1 sa r\u00fdchlo men\u00ed, tak\u017ee netreba predpoklada\u0165, \u017ee dnes nemo\u017en\u00e9 veci sa \u010doskoro nem\u00f4\u017eu sta\u0165 celkom skuto\u010dn\u00fdmi. Projekt <noindex><a rel=\"nofollow\" href=\"https:\/\/openai.com\/\">OpenAI<\/a><\/noindex> \u2014 je zauj\u00edmav\u00fd t\u00fdmto pr\u00edkladom. <\/p>\n<p><\/p>\n<p>Po druh\u00e9, my\u0161lienka, \u017ee je mo\u017en\u00e9 overova\u0165 a rozumie\u0165 procesu rozhodovania v existuj\u00facich syst\u00e9moch alebo organiz\u00e1ci\u00e1ch, je dobr\u00e1 teoreticky, ale na prax nie ve\u013emi. Pochopi\u0165, ako sa rozhoduje vo ve\u013ekej organiz\u00e1cii, v\u00f4bec nie je jednoduch\u00e9. Aj ke\u010f existuje form\u00e1lny proces rozhodovania, neodr\u00e1\u017ea to, ako sa \u013eudia v skuto\u010dnosti interactuj\u00fa, a sami \u010dasto nemaj\u00fa logick\u00fd syst\u00e9mov\u00fd pr\u00edstup k prij\u00edmaniu svojich rozhodnut\u00ed. Ako povedal m\u00f4j kolega <noindex><a rel=\"nofollow\" href=\"https:\/\/en.wikipedia.org\/wiki\/Vijay_S._Pande\">Vidhay Pande<\/a><\/noindex>, <strong>\u013eudia s\u00fa tie\u017e \u010dierne skrinky<\/strong>. <\/p>\n<p><\/p>\n<p>Zoberte tis\u00edc \u013eud\u00ed z nieko\u013ek\u00fdch prekr\u00fdvaj\u00facich sa spolo\u010dnost\u00ed a in\u0161tit\u00faci\u00ed a probl\u00e9m sa stane e\u0161te zlo\u017eitej\u0161\u00edm. Po skuto\u010dnosti vieme, \u017ee \u201eSpace Shuttle\u201c bolo predur\u010den\u00e9 rozpadn\u00fa\u0165 sa na kusy pri n\u00e1vrate, a jednotliv\u00ed \u013eudia v NASA mali inform\u00e1cie, ktor\u00e9 im d\u00e1vali d\u00f4vod myslie\u0165 si, \u017ee sa m\u00f4\u017ee sta\u0165 nie\u010do zl\u00e9, ale syst\u00e9m <em>vo v\u0161eobecnosti<\/em> to nevedel. NASA pr\u00e1ve pre\u0161la podobn\u00fdm auditom, ke\u010f stratila predch\u00e1dzaj\u00faci shuttle, a predsa stratila \u010fal\u0161\u00ed \u2014 z ve\u013emi podobn\u00e9ho d\u00f4vodu. Je \u013eahk\u00e9 tvrdi\u0165, \u017ee organiz\u00e1cie a \u013eudia dodr\u017eiavaj\u00fa pr\u00edsne logick\u00e9 pravidl\u00e1, ktor\u00e9 je mo\u017en\u00e9 overi\u0165, pochopi\u0165 a zmeni\u0165 \u2014 ale sk\u00fasenosti dokazuj\u00fa opak. Toto je \"<noindex><a rel=\"nofollow\" href=\"https:\/\/ru.wikipedia.org\/wiki\/%D0%93%D0%BE%D1%81%D0%BF%D0%BB%D0%B0%D0%BD_%D0%A1%D0%A1%D0%A1%D0%A0\">omyl Gospod\u00e1ra<\/a><\/noindex>\u00bb.<\/p>\n<p><\/p>\n<p><strong>\u010casto porovn\u00e1vam strojov\u00e9 u\u010denie s datab\u00e1zami, najm\u00e4 s rela\u010dn\u00fdmi \u2014 novou z\u00e1kladnou technol\u00f3giou, ktor\u00e1 zmenila mo\u017enosti informatiky a svet okolo nej, ktor\u00e1 sa stala s\u00fa\u010das\u0165ou v\u0161etk\u00e9ho, \u010do neust\u00e1le pou\u017e\u00edvame, ani\u017e by sme si to uvedomovali.<\/strong> V datab\u00e1zach s\u00fa tie\u017e probl\u00e9my podobn\u00e9ho charakteru: syst\u00e9m m\u00f4\u017ee by\u0165 postaven\u00fd na nespr\u00e1vnych predpokladoch alebo na zl\u00fdch \u00fadajoch, ale \u0165a\u017eko sa to d\u00e1 postrehn\u00fa\u0165 a \u013eudia, ktor\u00ed syst\u00e9m pou\u017e\u00edvaj\u00fa, bud\u00fa robi\u0165 to, \u010do im hovor\u00ed, bez kladenia ot\u00e1zok. Existuje mno\u017estvo star\u00fdch vtipov o pracovn\u00edkoch da\u0148ov\u00e9ho \u00faradu, ktor\u00ed raz nespr\u00e1vne zap\u00edsali va\u0161e meno, a presved\u010di\u0165 ich, aby chybu opravili, je omnoho \u0165a\u017e\u0161ie, ne\u017e skuto\u010dne zmeni\u0165 meno. O tom sa d\u00e1 uva\u017eova\u0165 r\u00f4zne, ale nie je jasn\u00e9, ako lep\u0161ie: ako o technickom probl\u00e9me v SQL, alebo ako o chybe vo vydan\u00ed Oracle, alebo ako o poruche byrokratick\u00fdch in\u0161tit\u00fatov? Ak\u00e9 \u0165a\u017ek\u00e9 je n\u00e1js\u0165 chybu v procese, ktor\u00fd viedol k tomu, \u017ee syst\u00e9m nem\u00e1 tak\u00fa funkciu, ako oprava preklepov? Mohlo sa to zisti\u0165 sk\u00f4r, ne\u017e za\u010dali \u013eudia s\u0165a\u017eova\u0165? <\/p>\n<p><\/p>\n<p>T\u00fato problematiku e\u0161te jednoduch\u0161ie ilustruj\u00fa pr\u00edbehy, ke\u010f vodi\u010di kv\u00f4li zastaran\u00fdm \u00fadajom v navig\u00e1tore odbo\u010dia do riek. Dobre, mapy by sa mali neust\u00e1le aktualizova\u0165. Ale do akej miery m\u00f4\u017ee by\u0165 TomTom zodpovedn\u00fd za to, \u017ee va\u0161u auto odv\u00e1\u017ea do mora?<\/p>\n<p><\/p>\n<p>Hovor\u00edm to preto, \u017ee \u00e1no \u2014 predsudky strojov\u00e9ho u\u010denia prines\u00fa probl\u00e9my. Ale tieto probl\u00e9my bud\u00fa podobn\u00e9 t\u00fdm, ktor\u00fdm sme \u010delili v minulosti, a d\u00e1 sa ich v\u0161imn\u00fa\u0165 a vyrie\u0161i\u0165 (alebo nie) zhruba tak dobre, ako to bola na\u0161a predch\u00e1dzaj\u00faca sk\u00fasenos\u0165. Preto je pravdepodobn\u00e9, \u017ee scen\u00e1r, v ktorom predsudok AI sp\u00f4sob\u00ed \u0161kodu, sa zrejme nevyskytne u popredn\u00fdch v\u00fdskumn\u00edkov pracuj\u00facich vo ve\u013ekej organiz\u00e1cii. Sk\u00f4r sa nejak\u00fd bezv\u00fdznamn\u00fd technologick\u00fd dod\u00e1vate\u013e alebo dod\u00e1vate\u013e softv\u00e9ru rozhodne nap\u00edsa\u0165 nie\u010do na kolene, pri\u010dom pou\u017eije nezrozumite\u013en\u00e9 open source komponenty, kni\u017enice a n\u00e1stroje. A ne\u0161\u0165astn\u00fd klient sa nech\u00e1 oklama\u0165 fr\u00e1zou \u201eumel\u00e1 inteligencia\u201c v opise produktu a bez zbyto\u010dn\u00fdch ot\u00e1zok ho rozd\u00e1 svojim n\u00edzkopr\u00edjmov\u00fdm zamestnancom, n\u00fatiac ich robi\u0165 to, \u010do povie AI. To sa presne stalo s datab\u00e1zami. Nie je to probl\u00e9m umelej inteligencie ani probl\u00e9m softv\u00e9ru. Je to \u013eudsk\u00fd faktor.<\/p>\n<p><\/p>\n<h2 id=\"zaklyuchenie\">Z\u00e1ver<\/h2>\n<p><\/p>\n<blockquote><p><em>Strojov\u00e9 u\u010denie m\u00f4\u017ee dosiahnu\u0165 v\u0161etko, \u010do m\u00f4\u017eete nau\u010di\u0165 psa \u2014 ale nikdy nem\u00f4\u017eete by\u0165 si ist\u00ed, \u010do presne ste tohto psa nau\u010dili.<\/em><\/p><\/blockquote>\n<p>\u010casto m\u00e1m pocit, \u017ee term\u00edn \u201eumel\u00e1 inteligencia\u201c len br\u00e1ni diskusi\u00e1m ako je t\u00e1to. Tento term\u00edn vytv\u00e1ra falo\u0161n\u00fd dojem, \u017ee sme ju skuto\u010dne vytvorili \u2013 t\u00fato inteligenciu. \u017de sme na ceste k HAL9000 alebo Skynet \u2013 k nie\u010domu, \u010do naozaj <em>ch\u00e1pe<\/em>. Ale nie. S\u00fa to len stroje a je ove\u013ea presnej\u0161ie ich porovn\u00e1va\u0165 napr\u00edklad s pr\u00e1\u010dkou. Pr\u00e1\u010dka sa ove\u013ea lep\u0161ie postar\u00e1 o pranie ako \u010dlovek, ale ak do nej d\u00e1te riad namiesto oble\u010denia, ten riad\u2026 opra\u0165. Riad bude dokonca \u010dist\u00fd. Ale nebude to to, \u010do ste o\u010dak\u00e1vali, a stane sa to nie preto, \u017ee by syst\u00e9m mal nejak\u00e9 predsudky vo\u010di riadu. Pr\u00e1\u010dka nevie, \u010do je riad, ani \u010do je oble\u010denie \u2013 je to len pr\u00edklad automatiz\u00e1cie, ktor\u00fd sa konceptu\u00e1lne nel\u00ed\u0161i od toho, ako sa procesy automatizovali predt\u00fdm.<\/p>\n<p><\/p>\n<p>O \u010domko\u013evek sa hovor\u00ed \u2013 o strojoch, lietadl\u00e1ch alebo datab\u00e1zach \u2013 tieto syst\u00e9my bud\u00fa z\u00e1rove\u0148 ve\u013emi mocn\u00e9 a ve\u013emi obmedzen\u00e9. Bud\u00fa \u00faplne z\u00e1visl\u00e9 od toho, ako ich \u013eudia pou\u017e\u00edvaj\u00fa, ak\u00e9 dobr\u00e9 alebo zl\u00e9 maj\u00fa pri tom \u00famysly a do akej miery rozumej\u00fa ich fungovaniu.<\/p>\n<p><\/p>\n<p>Preto je tvrdi\u0165, \u017ee \u201eumel\u00e1 inteligencia je matematika, tak\u017ee nem\u00f4\u017ee ma\u0165 predsudky\u201c, \u00faplne nespr\u00e1vne. Ale rovnako nepresn\u00e9 je tvrdi\u0165, \u017ee strojov\u00e9 u\u010denie je \u201esubjekt\u00edvne vo svojej podstate\u201c. Strojov\u00e9 u\u010denie nach\u00e1dza vzory v d\u00e1tach a ak\u00e9 vzory n\u00e1jde, z\u00e1vis\u00ed od d\u00e1t, a d\u00e1ta z\u00e1visia od n\u00e1s. Rovnako ako to, \u010do s nimi rob\u00edme. Strojov\u00e9 u\u010denie naozaj rob\u00ed niektor\u00e9 veci ove\u013ea lep\u0161ie ako my \u2013 ale psy, napr\u00edklad, s\u00fa ove\u013ea efekt\u00edvnej\u0161ie ako \u013eudia pri odha\u013eovan\u00ed drog, \u010do v\u00f4bec nie je d\u00f4vod na to, aby sme ich pova\u017eovali za svedkov a vyn\u00e1\u0161ali rozsudky na z\u00e1klade ich v\u00fdpoved\u00ed. A psy, mimochodom, s\u00fa ove\u013ea inteligentnej\u0161ie ako ak\u00fdko\u013evek syst\u00e9m strojov\u00e9ho u\u010denia.<\/p>\n<p><strong>Preklad:<\/strong>\u00a0<noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/en\/users\/757NF\/\">Diana Leck\u00e1<\/a><\/noindex>.<br \/>\n<strong>Redakcia:<\/strong>\u00a0<noindex><a rel=\"nofollow\" href=\"http:\/\/teleg.one\/ponchiknews\">Alexej Ivanov<\/a><\/noindex>.<br \/>\n<strong>Komunita:<\/strong> <noindex><a rel=\"nofollow\" href=\"http:\/\/teleg.one\/ponchiknews\">@PonchikNews<\/a><\/noindex>.<\/p>\n<p>Zdroj: <a content=\"nofollow\" rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/449224\/\">habr.com<\/a><\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"<p>tl;dr: \u041c\u0430\u0448\u0438\u043d\u043d\u043e\u0435 \u043e\u0431\u0443\u0447\u0435\u043d\u0438\u0435 \u0438\u0449\u0435\u0442 \u0437\u0430\u043a\u043e\u043d\u043e\u043c\u0435\u0440\u043d\u043e\u0441\u0442\u0438 \u0432 \u0434\u0430\u043d\u043d\u044b\u0445. \u041d\u043e \u0438\u0441\u043a\u0443\u0441\u0441\u0442\u0432\u0435\u043d\u043d\u044b\u0439 \u0438\u043d\u0442\u0435\u043b\u043b\u0435\u043a\u0442 \u043c\u043e\u0436\u0435\u0442 \u0431\u044b\u0442\u044c \u00ab\u043f\u0440\u0435\u0434\u0432\u0437\u044f\u0442\u00bb \u2014\u00a0\u0442\u043e \u0435\u0441\u0442\u044c, \u043d\u0430\u0445\u043e\u0434\u0438\u0442\u044c \u043d\u0435\u0432\u0435\u0440\u043d\u044b\u0435 \u043f\u0430\u0442\u0442\u0435\u0440\u043d\u044b. \u041a \u043f\u0440\u0438\u043c\u0435\u0440\u0443, \u0441\u0438\u0441\u0442\u0435\u043c\u0430 \u043e\u0431\u043d\u0430\u0440\u0443\u0436\u0435\u043d\u0438\u044f \u0440\u0430\u043a\u0430 \u043a\u043e\u0436\u0438 \u043f\u043e \u0444\u043e\u0442\u043e\u0433\u0440\u0430\u0444\u0438\u0438 \u043c\u043e\u0436\u0435\u0442 \u043e\u0431\u0440\u0430\u0449\u0430\u0442\u044c \u043e\u0441\u043e\u0431\u043e\u0435 \u0432\u043d\u0438\u043c\u0430\u043d\u0438\u0435 \u043d\u0430 \u0441\u043d\u0438\u043c\u043a\u0438, \u0441\u0434\u0435\u043b\u0430\u043d\u043d\u044b\u0435 \u0432\u043e \u0432\u0440\u0430\u0447\u0435\u0431\u043d\u043e\u043c \u043a\u0430\u0431\u0438\u043d\u0435\u0442\u0435. \u041c\u0430\u0448\u0438\u043d\u043d\u043e\u0435 \u043e\u0431\u0443\u0447\u0435\u043d\u0438\u0435 \u043d\u0435 \u0443\u043c\u0435\u0435\u0442 \u043f\u043e\u043d\u0438\u043c\u0430\u0442\u044c: \u0435\u0433\u043e \u0430\u043b\u0433\u043e\u0440\u0438\u0442\u043c\u044b \u043b\u0438\u0448\u044c \u0432\u044b\u044f\u0432\u043b\u044f\u044e\u0442 \u0437\u0430\u043a\u043e\u043d\u043e\u043c\u0435\u0440\u043d\u043e\u0441\u0442\u0438 \u0432 \u0447\u0438\u0441\u043b\u0430\u0445, \u0438 \u0435\u0441\u043b\u0438 \u0434\u0430\u043d\u043d\u044b\u0435 \u043d\u0435 \u0440\u0435\u043f\u0440\u0435\u0437\u0435\u043d\u0442\u0430\u0442\u0438\u0432\u043d\u044b, \u0442\u0430\u043a\u0438\u043c \u0431\u0443\u0434\u0435\u0442 [&hellip;]<\/p>\n","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"author":1,"featured_media":24558,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[702],"tags":[],"class_list":["post-32781","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news"],"aioseo_notices":[],"aioseo_head":"\n\t\t<!-- All in One SEO 5.0.3 - aioseo.com -->\n\t<meta name=\"description\" content=\"tl;dr: .\" \/>\n\t<meta name=\"robots\" content=\"max-image-preview:large\" \/>\n\t<meta name=\"author\" content=\"Yuri Gagarin\"\/>\n\t<link rel=\"canonical\" href=\"https:\/\/prohoster.info\/sk\/blog\/news\/o-predvzyatosti-iskusstvennogo-intellekta\" \/>\n\t\t<meta name=\"generator\" content=\"All in One SEO (AIOSEO) 5.0.3\" \/>\n\t\t<meta property=\"og:locale\" content=\"sk_SK\" \/>\n\t\t<meta property=\"og:site_name\" content=\"ProHoster | \u041a\u0443\u043f\u0438\u0442\u044c \u043d\u0430\u0434\u0435\u0436\u043d\u044b\u0439 \u0445\u043e\u0441\u0442\u0438\u043d\u0433 \u0434\u043b\u044f \u0441\u0430\u0439\u0442\u043e\u0432 \u0441 \u0437\u0430\u0449\u0438\u0442\u043e\u0439 \u043e\u0442 DDoS, VPS VDS \u0441\u0435\u0440\u0432\u0435\u0440\u044b\" \/>\n\t\t<meta property=\"og:type\" content=\"article\" \/>\n\t\t<meta property=\"og:title\" content=\"\ud83e\udd47\u041e \u043f\u0440\u0435\u0434\u0432\u0437\u044f\u0442\u043e\u0441\u0442\u0438 \u0438\u0441\u043a\u0443\u0441\u0441\u0442\u0432\u0435\u043d\u043d\u043e\u0433\u043e \u0438\u043d\u0442\u0435\u043b\u043b\u0435\u043a\u0442\u0430 | ProHoster\" \/>\n\t\t<meta property=\"og:description\" content=\"tl;dr: .\" \/>\n\t\t<meta property=\"og:url\" content=\"https:\/\/prohoster.info\/sk\/blog\/news\/o-predvzyatosti-iskusstvennogo-intellekta\" \/>\n\t\t<meta property=\"og:image\" content=\"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg\" \/>\n\t\t<meta property=\"og:image:secure_url\" content=\"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg\" \/>\n\t\t<meta property=\"og:image:width\" content=\"350\" \/>\n\t\t<meta property=\"og:image:height\" content=\"350\" \/>\n\t\t<meta property=\"article:published_time\" content=\"2019-10-31T18:48:54+00:00\" \/>\n\t\t<meta property=\"article:modified_time\" content=\"2019-10-31T18:48:54+00:00\" \/>\n\t\t<meta property=\"article:publisher\" content=\"https:\/\/www.facebook.com\/prohoster\" \/>\n\t\t<meta property=\"article:author\" content=\"https:\/\/www.facebook.com\/prohoster\" \/>\n\t\t<!-- All in One SEO -->\n\n","aioseo_head_json":{"title":"\ud83e\udd47O predsudkoch umelej inteligencie | ProHoster","description":"tl;dr: .","canonical_url":"https:\/\/prohoster.info\/sk\/blog\/news\/o-predvzyatosti-iskusstvennogo-intellekta","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"sk_SK","og:site_name":"ProHoster | \u041a\u0443\u043f\u0438\u0442\u044c \u043d\u0430\u0434\u0435\u0436\u043d\u044b\u0439 \u0445\u043e\u0441\u0442\u0438\u043d\u0433 \u0434\u043b\u044f \u0441\u0430\u0439\u0442\u043e\u0432 \u0441 \u0437\u0430\u0449\u0438\u0442\u043e\u0439 \u043e\u0442 DDoS, VPS VDS \u0441\u0435\u0440\u0432\u0435\u0440\u044b","og:type":"article","og:title":"\ud83e\udd47\u041e \u043f\u0440\u0435\u0434\u0432\u0437\u044f\u0442\u043e\u0441\u0442\u0438 \u0438\u0441\u043a\u0443\u0441\u0441\u0442\u0432\u0435\u043d\u043d\u043e\u0433\u043e \u0438\u043d\u0442\u0435\u043b\u043b\u0435\u043a\u0442\u0430 | ProHoster","og:description":"tl;dr: .","og:url":"https:\/\/prohoster.info\/sk\/blog\/news\/o-predvzyatosti-iskusstvennogo-intellekta","og:image":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:secure_url":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:width":350,"og:image:height":350,"article:published_time":"2019-10-31T18:48:54+00:00","article:modified_time":"2019-10-31T18:48:54+00:00","article:publisher":"https:\/\/www.facebook.com\/prohoster","article:author":"https:\/\/www.facebook.com\/prohoster"},"aioseo_meta_data":{"post_id":"32781","title":null,"description":null,"keywords":null,"keyphrases":null,"primary_term":null,"canonical_url":null,"og_title":null,"og_description":null,"og_object_type":"default","og_image_type":"default","og_image_url":null,"og_image_width":null,"og_image_height":null,"og_image_custom_url":null,"og_image_custom_fields":null,"og_video":null,"og_custom_url":null,"og_article_section":null,"og_article_tags":null,"twitter_use_og":false,"twitter_card":"default","twitter_image_type":"default","twitter_image_url":null,"twitter_image_custom_url":null,"twitter_image_custom_fields":null,"twitter_title":null,"twitter_description":null,"schema":{"blockGraphs":[],"customGraphs":[],"default":{"data":{"Article":[],"Course":[],"Dataset":[],"FAQPage":[],"Movie":[],"Person":[],"Product":[],"ProductReview":[],"Car":[],"Recipe":[],"Service":[],"SoftwareApplication":[],"WebPage":[]},"graphName":"","isEnabled":true},"graphs":[]},"schema_type":null,"schema_type_options":null,"pillar_content":false,"robots_default":true,"robots_noindex":false,"robots_noarchive":false,"robots_nosnippet":false,"robots_nofollow":false,"robots_noimageindex":false,"robots_noodp":false,"robots_notranslate":false,"robots_max_snippet":null,"robots_max_videopreview":null,"robots_max_imagepreview":"large","priority":null,"frequency":null,"local_seo":null,"seo_analyzer_scan_date":"2026-01-21 12:30:22","breadcrumb_settings":null,"limit_modified_date":false,"reviewed_by":null,"ai":null,"created":"2021-03-01 02:52:26","updated":"2026-01-21 12:30:22","focus_keyword":null,"additional_keywords":null,"truseo_locale":null},"gt_translate_keys":[{"key":"link","format":"url"}],"_links":{"self":[{"href":"https:\/\/prohoster.info\/sk\/wp-json\/wp\/v2\/posts\/32781","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/sk\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/sk\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/sk\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/sk\/wp-json\/wp\/v2\/comments?post=32781"}],"version-history":[{"count":0,"href":"https:\/\/prohoster.info\/sk\/wp-json\/wp\/v2\/posts\/32781\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/prohoster.info\/sk\/wp-json\/wp\/v2\/media\/24558"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/sk\/wp-json\/wp\/v2\/media?parent=32781"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/sk\/wp-json\/wp\/v2\/categories?post=32781"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/sk\/wp-json\/wp\/v2\/tags?post=32781"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}