{"id":39055,"date":"2019-10-31T22:27:36","date_gmt":"2019-10-31T19:27:36","guid":{"rendered":"https:\/\/prohoster.info\/blog\/my-ne-mozhem-doveryat-ii-sistemam-postroennym-na-odnom-lish-glubokom-obuchenii\/"},"modified":"2019-10-31T22:27:36","modified_gmt":"2019-10-31T19:27:36","slug":"my-ne-mozhem-doveryat-ii-sistemam-postroennym-na-odnom-lish-glubokom-obuchenii","status":"publish","type":"post","link":"https:\/\/prohoster.info\/cs\/blog\/news\/my-ne-mozhem-doveryat-ii-sistemam-postroennym-na-odnom-lish-glubokom-obuchenii","title":{"rendered":"Nem\u016f\u017eeme d\u016fv\u011b\u0159ovat AI syst\u00e9m\u016fm postaven\u00fdm pouze na hlubok\u00e9m u\u010den\u00ed","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

\"Nem\u016f\u017eeme
\n
\nTento text nen\u00ed v\u00fdsledkem v\u011bdeck\u00e9ho v\u00fdzkumu, ale jedn\u00edm z mnoha n\u00e1zor\u016f t\u00fdkaj\u00edc\u00edch se na\u0161eho nejbli\u017e\u0161\u00edho technologick\u00e9ho rozvoje. A z\u00e1rove\u0148 v\u00fdzvou k diskusi.<\/i><\/p>\n

Gary Marcus, profesor na Newyorsk\u00e9 univerzit\u011b, je p\u0159esv\u011bd\u010den, \u017ee hlubok\u00e9 u\u010den\u00ed hraje d\u016fle\u017eitou roli ve v\u00fdvoji AI. Ale tak\u00e9 v\u011b\u0159\u00ed, \u017ee p\u0159ehnan\u00e9 nad\u0161en\u00ed pro tuto metodu m\u016f\u017ee v\u00e9st k jej\u00ed diskreditaci.<\/p>\n

Ve sv\u00e9 knize Rebooting AI: Building artificial intelligence we can trust<\/i> Marcus, neurolog vzd\u011bl\u00e1n\u00edm, kter\u00fd vybudoval kari\u00e9ru na p\u0159edn\u00edch v\u00fdzkumech v oblasti AI, se zam\u011b\u0159uje na technick\u00e9 a etick\u00e9 aspekty. Z technologick\u00e9ho hlediska m\u016f\u017ee hlubok\u00e9 u\u010den\u00ed \u00fasp\u011b\u0161n\u011b napodobovat \u0159e\u0161en\u00ed probl\u00e9m\u016f vn\u00edm\u00e1n\u00ed, kter\u00e1 vykon\u00e1v\u00e1 n\u00e1\u0161 mozek: nap\u0159\u00edklad rozpozn\u00e1v\u00e1n\u00ed obraz\u016f nebo \u0159e\u010di. Ale k vy\u0159e\u0161en\u00ed jin\u00fdch \u00fakol\u016f, jako je porozum\u011bn\u00ed konverzac\u00edm nebo ur\u010den\u00ed p\u0159\u00ed\u010dinn\u00fdch souvislost\u00ed, nen\u00ed hlubok\u00e9 u\u010den\u00ed vhodn\u00e9. Aby bylo mo\u017en\u00e9 vytvo\u0159it pokro\u010dilej\u0161\u00ed inteligentn\u00ed stroje, kter\u00e9 dok\u00e1\u017eou \u0159e\u0161it \u0161ir\u0161\u00ed okruh \u00fakol\u016f \u2013 \u010dasto naz\u00fdvan\u00fdch obecnou um\u011blou inteligenc\u00ed \u2013 je t\u0159eba kombinovat hlubok\u00e9 u\u010den\u00ed s jin\u00fdmi metodikami.
\n<\/a><\/noindex>
\nPokud AI syst\u00e9m ve skute\u010dnosti nerozum\u00ed sv\u00fdm \u00fakol\u016fm nebo okoln\u00edmu sv\u011btu, m\u016f\u017ee to v\u00e9st k nebezpe\u010dn\u00fdm n\u00e1sledk\u016fm. I ty nejmen\u0161\u00ed neo\u010dek\u00e1van\u00e9 zm\u011bny v prost\u0159ed\u00ed syst\u00e9mu mohou vyvolat chybnou reakci. Takov\u00fdch p\u0159\u00edklad\u016f u\u017e bylo mnoho: klasifik\u00e1tory nevhodn\u00fdch v\u00fdraz\u016f, kter\u00e9 lze snadno oklamat; syst\u00e9my pro hled\u00e1n\u00ed zam\u011bstn\u00e1n\u00ed, kter\u00e9 neust\u00e1le diskriminuj\u00ed; autonomn\u00ed vozidla, kter\u00e1 havaruj\u00ed a ob\u010das zra\u0148uj\u00ed \u0159idi\u010de nebo chodce. Vytvo\u0159en\u00ed obecn\u00e9 um\u011bl\u00e9 inteligence nen\u00ed jen zaj\u00edmav\u00fd v\u00fdzkumn\u00fd probl\u00e9m, m\u00e1 tak\u00e9 mnoh\u00e1 praktick\u00e1 vyu\u017eit\u00ed.<\/p>\n

Ve sv\u00e9 knize Marcus a jeho spoluautor Ernest Davis navrhuj\u00ed jinou cestu. Domn\u00edvaj\u00ed se, \u017ee jsme je\u0161t\u011b daleko od vytvo\u0159en\u00ed obecn\u00e9 AI, ale jsou si jisti, \u017ee se d\u0159\u00edve \u010di pozd\u011bji poda\u0159\u00ed ji vytvo\u0159it.<\/p>\n

Pro\u010d pot\u0159ebujeme obecnou AI? Specializovan\u00e9 verze u\u017e byly vytvo\u0159eny a p\u0159in\u00e1\u0161ej\u00ed zna\u010dn\u00fd prosp\u011bch.<\/b><\/p>\n

Ano, a p\u0159\u00ednos bude je\u0161t\u011b v\u011bt\u0161\u00ed. Existuje v\u0161ak mnoho \u00faloh, kter\u00e9 specializovan\u00e9 AI nedok\u00e1\u017ee vy\u0159e\u0161it. Nap\u0159\u00edklad porozum\u011bn\u00ed b\u011b\u017en\u00e9 \u0159e\u010di, nebo obecn\u00e1 pomoc ve virtu\u00e1ln\u00edm sv\u011bt\u011b, nebo robot, kter\u00fd pom\u00e1h\u00e1 s \u00faklidem a va\u0159en\u00edm. Takov\u00e9 \u00fakoly jsou mimo mo\u017enosti specializovan\u00e9ho AI. Je tu tak\u00e9 zaj\u00edmav\u00e1 praktick\u00e1 ot\u00e1zka: lze pomoc\u00ed specializovan\u00e9ho AI vytvo\u0159it bezpe\u010dn\u00e9 autonomn\u00ed vozidlo? Zku\u0161enosti ukazuj\u00ed, \u017ee takov\u00e9 AI m\u00e1 st\u00e1le mnoho probl\u00e9m\u016f s chov\u00e1n\u00edm v abnorm\u00e1ln\u00edch situac\u00edch, i kdy\u017e jde o \u0159\u00edzen\u00ed, co\u017e situaci zna\u010dn\u011b komplikuje.<\/p>\n

Mysl\u00edm, \u017ee bychom v\u0161ichni cht\u011bli m\u00edt AI, kter\u00e1 by n\u00e1m mohla pomoci u\u010dinit nov\u00e9 z\u00e1sadn\u00ed objevy v medic\u00edn\u011b. Nen\u00ed jasn\u00e9, zda jsou sou\u010dasn\u00e9 technologie pro to vhodn\u00e9, proto\u017ee biologie je slo\u017eit\u00e1 oblast. Je t\u0159eba b\u00fdt p\u0159ipraven p\u0159e\u010d\u00edst \u0159adu knih. V\u011bdci rozum\u00ed p\u0159\u00ed\u010dinn\u00fdm souvislostem ve vz\u00e1jemn\u00e9m p\u016fsoben\u00ed s\u00edt\u00ed a molekul, mohou vyv\u00edjet teorie o planet\u00e1ch a tak d\u00e1le. Av\u0161ak se specializovan\u00fdm AI nem\u016f\u017eeme vytvo\u0159it stroje schopn\u00e9 takov\u00fdch objev\u016f. A s obecnou AI bychom mohli prov\u00e9st revoluci ve v\u011bd\u011b, technologi\u00edch a medic\u00edn\u011b. Podle m\u00e9ho n\u00e1zoru je velmi d\u016fle\u017eit\u00e9 pokra\u010dovat v pr\u00e1ci na vytvo\u0159en\u00ed obecn\u00e9 AI.<\/p>\n

Zd\u00e1 se, \u017ee pod \u201eobecnou\u201c m\u00e1te na mysli silnou AI?<\/b><\/p>\n

Kdy\u017e \u0159\u00edk\u00e1m \u201eobecn\u00fd\u201c, m\u00e1m na mysli, \u017ee AI bude schopna na m\u00edst\u011b myslet a samostatn\u011b \u0159e\u0161it nov\u00e9 \u00fakoly. Na rozd\u00edl od nap\u0159\u00edklad Go, kde se probl\u00e9m v posledn\u00edch 2000 letech nezm\u011bnil.<\/p>\n

Obecn\u00e1 AI mus\u00ed b\u00fdt schopna rozhodovat jak v politice, tak v medic\u00edn\u011b. Je to analogick\u00e9 lidsk\u00e9 schopnosti; ka\u017ed\u00fd rozumn\u011b sm\u00fd\u0161lej\u00edc\u00ed \u010dlov\u011bk je schopen d\u011blat velmi mnoho. Vezm\u011bte nezku\u0161en\u00e9 studenty a b\u011bhem n\u011bkolika dn\u00ed je p\u0159im\u011bjte pracovat prakticky na \u010demkoliv, od pr\u00e1vn\u00edho \u00fakolu po medic\u00ednu. To je d\u00edky tomu, \u017ee maj\u00ed obecn\u00e9 porozum\u011bn\u00ed sv\u011btu a um\u011bj\u00ed \u010d\u00edst, a proto mohou p\u0159isp\u011bt k velmi \u0161irok\u00e9mu spektru \u010dinnost\u00ed.<\/p>\n

Vztah mezi takov\u00fdm intelektem a siln\u00fdm spo\u010d\u00edv\u00e1 v tom, \u017ee slab\u00fd intelekt pravd\u011bpodobn\u011b nedok\u00e1\u017ee \u0159e\u0161it obecn\u00e9 \u00fakoly. K vytvo\u0159en\u00ed n\u011b\u010deho dostate\u010dn\u011b spolehliv\u00e9ho, schopn\u00e9ho pracov\u00e1n\u00edm s neust\u00e1le se m\u011bn\u00edc\u00edm sv\u011btem, budete pravd\u011bpodobn\u011b pot\u0159ebovat alespo\u0148 p\u0159ibl\u00ed\u017eit se k obecn\u00e9mu intelektu.<\/p>\n

Ale te\u010f jsme od toho velmi daleko. AlphaGo m\u016f\u017ee skv\u011ble hr\u00e1t na desce 19\u00d719, ale mus\u00ed se p\u0159e\u0161kolit, aby mohl hr\u00e1t na obd\u00e9ln\u00edkov\u00e9 desce. Nebo vezm\u011bte pr\u016fm\u011brn\u00fd syst\u00e9m hlubok\u00e9ho u\u010den\u00ed: je schopen rozpoznat slona, pokud je dob\u0159e osv\u011btlen a viditeln\u00e1 textura jeho k\u016f\u017ee. A pokud je vid\u011bt jen silueta slona, syst\u00e9m ho jist\u011b nebude moci rozpoznat.<\/p>\n

Ve sv\u00e9 knize zmi\u0148ujete, \u017ee hlubok\u00e9 u\u010den\u00ed nen\u00ed schopno dos\u00e1hnout schopnost\u00ed obecn\u00e9 AI, proto\u017ee nen\u00ed schopno hlubok\u00e9ho pochopen\u00ed.<\/b><\/p>\n

V kognitivistice se hovo\u0159\u00ed o vytv\u00e1\u0159en\u00ed r\u016fzn\u00fdch kognitivn\u00edch model\u016f. Sed\u00edm v hotelov\u00e9m pokoji a ch\u00e1pu, \u017ee tam je sk\u0159\u00ed\u0148, tam je postel, tam je televizor, kter\u00fd je neobvykle zav\u011b\u0161en\u00fd. Zn\u00e1m v\u0161echny tyto p\u0159edm\u011bty, nejen je identifikuji. Tak\u00e9 rozum\u00edm, jak si navz\u00e1jem souvisej\u00ed. M\u00e1m n\u00e1pady o fungov\u00e1n\u00ed okoln\u00edho sv\u011bta. Nejsou dokonal\u00e9. Mohou b\u00fdt chybn\u00e9, ale jsou velmi dobr\u00e9. A na jejich z\u00e1klad\u011b \u010din\u00edm spoustu z\u00e1v\u011br\u016f, kter\u00e9 se st\u00e1vaj\u00ed vod\u00edtkem pro m\u00e9 ka\u017edodenn\u00ed akce.<\/p>\n

Druh\u00e1 krajnost je n\u011bco jako hern\u00ed syst\u00e9m Atari vytvo\u0159en\u00fd DeepMind, ve kter\u00e9m si pamatoval, co m\u00e1 d\u011blat, kdy\u017e vid\u011bl pixely na ur\u010dit\u00e9m m\u00edst\u011b na obrazovce. Pokud z\u00edsk\u00e1te dostatek dat, m\u016f\u017ee se zd\u00e1t, \u017ee m\u00e1te porozum\u011bn\u00ed, ale ve skute\u010dnosti je velmi povrchn\u00ed. D\u016fkazem toho je, \u017ee pokud posunete objekty o t\u0159i pixely, AI hraje mnohem h\u016f\u0159e. Zm\u011bny ho zm\u00e1tou. To je opak hlubok\u00e9ho porozum\u011bn\u00ed.<\/p>\n

Abychom tento probl\u00e9m vy\u0159e\u0161ili, navrhujete vr\u00e1tit se k klasick\u00e9 AI. Jak\u00e9 jej\u00ed v\u00fdhody bychom m\u011bli usilovat vyu\u017e\u00edt?<\/b><\/p>\n

Existuje n\u011bkolik v\u00fdhod. <\/p>\n

Za prv\u00e9, klasick\u00e1 AI je vlastn\u011b r\u00e1mec pro vytv\u00e1\u0159en\u00ed kognitivn\u00edch model\u016f sv\u011bta, na jejich\u017e z\u00e1klad\u011b lze potom d\u011blat z\u00e1v\u011bry.<\/p>\n

Za druh\u00e9, klasick\u00e1 AI je ide\u00e1ln\u011b kompatibiln\u00ed s pravidly. V sou\u010dasnosti se v oblasti hlubok\u00e9ho u\u010den\u00ed objevuje podivn\u00fd trend, kdy se specialist\u00e9 sna\u017e\u00ed vyh\u00fdbat pravidl\u016fm. Cht\u011bj\u00ed d\u011blat v\u0161echno na neuronov\u00fdch s\u00edt\u00edch a ned\u011blat nic, co vypad\u00e1 jako klasick\u00e9 programov\u00e1n\u00ed. Ale existuj\u00ed \u00fakoly, kter\u00e9 byly klidn\u011b vy\u0159e\u0161eny t\u00edmto zp\u016fsobem, a nikdo si toho nev\u0161iml. Nap\u0159\u00edklad pl\u00e1nov\u00e1n\u00ed tras v Google Maps.<\/p>\n

Ve skute\u010dnosti pot\u0159ebujeme oba p\u0159\u00edstupy. Strojov\u00e9 u\u010den\u00ed umo\u017e\u0148uje dob\u0159e se u\u010dit z dat, ale velmi \u0161patn\u011b pom\u00e1h\u00e1 p\u0159i zobrazen\u00ed abstrakce, kterou p\u0159edstavuje po\u010d\u00edta\u010dov\u00fd program. Klasick\u00e1 AI dob\u0159e pracuje s abstrakcemi, ale mus\u00ed b\u00fdt zcela programov\u00e1na ru\u010dn\u011b, a ve sv\u011bt\u011b bylo nasb\u00edr\u00e1no p\u0159\u00edli\u0161 mnoho znalost\u00ed, aby bylo mo\u017en\u00e9 je v\u0161echny naprogramovat. Je jasn\u00e9, \u017ee mus\u00edme oba p\u0159\u00edstupy kombinovat.<\/p>\n

To souvis\u00ed s kapitolou, ve kter\u00e9 mluv\u00edte o tom, co se m\u016f\u017eeme nau\u010dit od lidsk\u00e9ho rozumu. A to p\u0159edev\u0161\u00edm o konceptu, kter\u00fd vych\u00e1z\u00ed z ji\u017e zm\u00edn\u011bn\u00e9 my\u0161lenky, \u017ee na\u0161e v\u011bdom\u00ed se skl\u00e1d\u00e1 z mnoha r\u016fzn\u00fdch syst\u00e9m\u016f, kter\u00e9 funguj\u00ed r\u016fzn\u00fdmi zp\u016fsoby.<\/b><\/p>\n

Mysl\u00edm, \u017ee existuje jin\u00fd zp\u016fsob, jak to vysv\u011btlit: ka\u017ed\u00fd kognitivn\u00ed syst\u00e9m, kter\u00fd m\u00e1me, skute\u010dn\u011b \u0159e\u0161\u00ed r\u016fzn\u00e9 \u00fakoly. Podobn\u00e9 \u010d\u00e1sti AI by m\u011bly b\u00fdt navr\u017eeny k \u0159e\u0161en\u00ed r\u016fzn\u00fdch \u00fakol\u016f, kter\u00e9 maj\u00ed r\u016fzn\u00e9 charakteristiky.<\/p>\n

Nyn\u00ed se sna\u017e\u00edme pou\u017e\u00edvat jak\u00e9si all-in-one technologie k \u0159e\u0161en\u00ed \u00fakol\u016f, kter\u00e9 se radik\u00e1ln\u011b li\u0161\u00ed. Pochopit v\u011btu rozhodn\u011b nen\u00ed tot\u00e9\u017e, co rozpoznat objekt. Ale lid\u00e9 se sna\u017e\u00ed v obou p\u0159\u00edpadech pou\u017e\u00edvat hlubok\u00e9 u\u010den\u00ed. Z kognitivn\u00edho hlediska jsou to kvalitativn\u011b odli\u0161n\u00e9 \u00fakoly. Jsem ohromen t\u00edm, jak m\u00e1lo odborn\u00edk\u016f na hlubok\u00e9 u\u010den\u00ed si cen\u00ed klasick\u00e9 AI. Pro\u010d \u010dekat na z\u00e1zra\u010dn\u00e9 \u0159e\u0161en\u00ed? To je nedosa\u017eiteln\u00e9, a bezv\u00fdsledn\u00e9 hled\u00e1n\u00ed ned\u00e1v\u00e1 mo\u017enost pochopit slo\u017eitost \u00fakolu vytvo\u0159en\u00ed AI.<\/p>\n

Tak\u00e9 zmi\u0148ujete, \u017ee AI syst\u00e9my jsou nezbytn\u00e9 pro porozum\u011bn\u00ed p\u0159\u00ed\u010dinn\u00fdm souvislostem. Mysl\u00edte, \u017ee n\u00e1m v tom pom\u016f\u017ee hlubok\u00e9 u\u010den\u00ed, klasick\u00e1 AI nebo n\u011bco zcela nov\u00e9ho?<\/b><\/p>\n

To je dal\u0161\u00ed oblast, pro kterou hlubok\u00e9 u\u010den\u00ed p\u0159\u00edli\u0161 dob\u0159e nefunguje. Nevysv\u011btluje p\u0159\u00ed\u010diny n\u011bjak\u00fdch ud\u00e1lost\u00ed, ale po\u010d\u00edt\u00e1 pravd\u011bpodobnost ud\u00e1losti za dan\u00fdch okolnost\u00ed.<\/p>\n

O \u010dem mluv\u00edme? Sledujete ur\u010dit\u00e9 sc\u00e9n\u00e1\u0159e a rozum\u00edte, pro\u010d se to d\u011bje a co se m\u016f\u017ee st\u00e1t, pokud se n\u011bjak\u00e9 okolnosti zm\u011bn\u00ed. M\u016f\u017eu se pod\u00edvat na stojan, na kter\u00e9m stoj\u00ed televize, a p\u0159edstavit si, \u017ee pokud od\u0159\u00edznu jednu nohu, stojan se p\u0159evr\u00e1t\u00ed a televize spadne. To je p\u0159\u00ed\u010dinn\u00e1 souvislost.<\/p>\n

Klasick\u00fd AI n\u00e1m pro to d\u00e1v\u00e1 ur\u010dit\u00e9 n\u00e1stroje. Nap\u0159\u00edklad m\u016f\u017ee p\u0159edstavit, co je podpora a co je selh\u00e1n\u00ed. Ale nebudu p\u0159eh\u00e1n\u011bt. Probl\u00e9m je v tom, \u017ee klasick\u00fd AI do zna\u010dn\u00e9 m\u00edry z\u00e1vis\u00ed na \u00faplnosti informac\u00ed o tom, co se d\u011bje, a j\u00e1 jsem dosp\u011bl k z\u00e1v\u011bru, pouze jsem se pod\u00edval na podstavec. Jak n\u011bjak\u00fdm zp\u016fsobem mohu generalizovat, p\u0159edstavovat \u010d\u00e1sti podstavce, kter\u00e9 nevid\u00edm. Zat\u00edm nem\u00e1me n\u00e1stroje pro proveden\u00ed t\u00e9to vlastnosti.<\/p>\n

Je\u0161t\u011b \u0159\u00edk\u00e1te, \u017ee lid\u00e9 maj\u00ed vrozen\u00e9 znalosti. Jak to lze realizovat v AI?<\/b><\/p>\n

V okam\u017eiku narozen\u00ed je n\u00e1\u0161 mozek ji\u017e velmi pe\u010dliv\u011b promy\u0161len\u00fd syst\u00e9m. Nen\u00ed fixn\u00ed, p\u0159\u00edroda vytvo\u0159ila prvn\u00ed hrub\u00fd n\u00e1vrh. A pot\u00e9 u\u010den\u00ed pom\u00e1h\u00e1 tento n\u00e1vrh revidovat po cel\u00fd n\u00e1\u0161 \u017eivot.<\/p>\n

Hrub\u00fd n\u00e1vrh mozku ji\u017e m\u00e1 ur\u010dit\u00e9 schopnosti. Novorozen\u00e1 horsk\u00e1 koza je schopn\u00e1 b\u011bhem n\u011bkolika hodin chybn\u011b sestoupit po svahu hory. Je z\u0159ejm\u00e9, \u017ee u\u017e m\u00e1 pochopen\u00ed trojrozm\u011brn\u00e9ho prostoru, sv\u00e9ho t\u011bla a vz\u00e1jemn\u00fdch vztah\u016f mezi nimi. Velmi slo\u017eit\u00fd syst\u00e9m.<\/p>\n

\u010c\u00e1ste\u010dn\u011b proto pova\u017euji, \u017ee pot\u0159ebujeme hybridy. Je t\u011b\u017ek\u00e9 si p\u0159edstavit, jak by bylo mo\u017en\u00e9 vytvo\u0159it robota, kter\u00fd dob\u0159e funguje ve sv\u011bt\u011b bez podobn\u00fdch znalost\u00ed, z \u010deho by m\u011bl za\u010d\u00edt, m\u00edsto toho, aby za\u010d\u00ednal s pr\u00e1zdn\u00fdm listem a u\u010dil se z dlouh\u00e9, obrovsk\u00e9 zku\u0161enosti.<\/p>\n

Pokud jde o lidi, na\u0161e vrozen\u00e9 znalosti poch\u00e1zej\u00ed z na\u0161eho genomu, kter\u00fd se dlouho vyv\u00edjel. A u AI syst\u00e9m\u016f budeme muset j\u00edt jinou cestou. \u010c\u00e1ste\u010dn\u011b by to mohla b\u00fdt pravidla pro konstrukci na\u0161ich algoritm\u016f. \u010c\u00e1ste\u010dn\u011b by to mohla b\u00fdt pravidla pro vytv\u00e1\u0159en\u00ed datov\u00fdch struktur, kter\u00fdmi tyto algoritmy manipuluj\u00ed. A \u010d\u00e1ste\u010dn\u011b to mohou b\u00fdt znalosti, kter\u00e9 p\u0159\u00edmo vlo\u017e\u00edme do stroj\u016f.<\/p>\n

Zaj\u00edmav\u00e9 je, \u017ee v knize doch\u00e1z\u00edte k my\u0161lence d\u016fv\u011bry a tvorby d\u016fv\u011bryhodn\u00fdch syst\u00e9m\u016f. Pro\u010d jste si vybral pr\u00e1v\u011b tento kriterium?<\/b><\/p>\n

Mysl\u00edm si, \u017ee dnes to p\u0159edstavuje hru s m\u00ed\u010dem. P\u0159ijde mi, \u017ee pro\u017e\u00edv\u00e1me podivn\u00fd okam\u017eik v historii, kdy do zna\u010dn\u00e9 m\u00edry d\u016fv\u011b\u0159ujeme softwaru, kter\u00fd si d\u016fv\u011bru nezaslou\u017e\u00ed. Mysl\u00edm, \u017ee na\u0161e dne\u0161n\u00ed obavy nebudou v\u011b\u010dn\u00e9. Za sto let AI spln\u00ed na\u0161i d\u016fv\u011bru, mo\u017en\u00e1 i d\u0159\u00edve.<\/p>\n

Ale dnes je AI nebezpe\u010dn\u00e9. Ne v tom smyslu, jak se ob\u00e1v\u00e1 Elon Musk, ale v tom, \u017ee syst\u00e9my pohovor\u016f p\u0159i p\u0159ij\u00edm\u00e1n\u00ed diskriminuj\u00ed \u017eeny, a to bez ohledu na to, co d\u011blaj\u00ed program\u00e1to\u0159i, proto\u017ee jejich n\u00e1stroje jsou p\u0159\u00edli\u0161 jednoduch\u00e9.<\/p>\n

Cht\u011bl bych, aby n\u00e1\u0161 AI byl kvalitn\u011bj\u0161\u00ed. Nechci, aby za\u010dala \"zima um\u011bl\u00e9 inteligence\", kdy si lid\u00e9 uv\u011bdom\u00ed, \u017ee AI nefunguje a je prost\u011b nebezpe\u010dn\u00e9, a nebudou cht\u00edt to napravit.<\/p>\n

V jist\u00e9m smyslu va\u0161e kniha skute\u010dn\u011b vypad\u00e1 velmi optimisticky. P\u0159edpokl\u00e1d\u00e1te, \u017ee je mo\u017en\u00e9 vybudovat d\u016fv\u011bryhodn\u00e9 AI. Jen se mus\u00edme pod\u00edvat jin\u00fdm sm\u011brem.<\/b><\/p>\n

Spr\u00e1vn\u011b, kniha je velmi pesimistick\u00e1 z kr\u00e1tkodob\u00e9ho hlediska a velmi optimistick\u00e1 z dlouhodob\u00e9ho. Domn\u00edv\u00e1me se, \u017ee v\u0161echny probl\u00e9my, kter\u00e9 jsme popsali, lze vy\u0159e\u0161it, pokud se pod\u00edv\u00e1me \u0161\u00ed\u0159eji na to, jak\u00e9 by m\u011bly b\u00fdt spr\u00e1vn\u00e9 odpov\u011bdi. A mysl\u00edme si, \u017ee pokud se to stane, sv\u011bt se zlep\u0161\u00ed.
\n
Zdroj: habr.com<\/a><\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"

\u042d\u0442\u043e\u0442 \u0442\u0435\u043a\u0441\u0442 \u2014 \u043d\u0435 \u0440\u0435\u0437\u0443\u043b\u044c\u0442\u0430\u0442 \u043d\u0430\u0443\u0447\u043d\u043e\u0433\u043e \u0438\u0441\u0441\u043b\u0435\u0434\u043e\u0432\u0430\u043d\u0438\u044f, \u0430 \u043e\u0434\u043d\u043e \u0438\u0437 \u043c\u043d\u043e\u0433\u0438\u0445 \u043c\u043d\u0435\u043d\u0438\u0439 \u043e\u0442\u043d\u043e\u0441\u0438\u0442\u0435\u043b\u044c\u043d\u043e \u043d\u0430\u0448\u0435\u0433\u043e \u0431\u043b\u0438\u0436\u0430\u0439\u0448\u0435\u0433\u043e \u0442\u0435\u0445\u043d\u043e\u043b\u043e\u0433\u0438\u0447\u0435\u0441\u043a\u043e\u0433\u043e \u0440\u0430\u0437\u0432\u0438\u0442\u0438\u044f. \u0418 \u0437\u0430\u043e\u0434\u043d\u043e \u043f\u0440\u0438\u0433\u043b\u0430\u0448\u0435\u043d\u0438\u0435 \u043a \u0434\u0438\u0441\u043a\u0443\u0441\u0441\u0438\u0438. \u0413\u0430\u0440\u0438 \u041c\u0430\u0440\u043a\u0443\u0441, \u043f\u0440\u043e\u0444\u0435\u0441\u0441\u043e\u0440 \u041d\u044c\u044e-\u0419\u043e\u0440\u043a\u0441\u043a\u043e\u0433\u043e \u0443\u043d\u0438\u0432\u0435\u0440\u0441\u0438\u0442\u0435\u0442\u0430, \u0443\u0432\u0435\u0440\u0435\u043d, \u0447\u0442\u043e \u0433\u043b\u0443\u0431\u043e\u043a\u043e\u0435 \u043e\u0431\u0443\u0447\u0435\u043d\u0438\u0435 \u0438\u0433\u0440\u0430\u0435\u0442 \u0432\u0430\u0436\u043d\u0443\u044e \u0440\u043e\u043b\u044c \u0432 \u0440\u0430\u0437\u0432\u0438\u0442\u0438\u0438 \u0418\u0418. \u041d\u043e \u043e\u043d \u0442\u0430\u043a\u0436\u0435 \u0441\u0447\u0438\u0442\u0430\u0435\u0442, \u0447\u0442\u043e \u0438\u0437\u0431\u044b\u0442\u043e\u0447\u043d\u043e\u0435 \u0443\u0432\u043b\u0435\u0447\u0435\u043d\u0438\u0435 \u044d\u0442\u043e\u0439 \u043c\u0435\u0442\u043e\u0434\u0438\u043a\u043e\u0439 \u043c\u043e\u0436\u0435\u0442 \u043f\u0440\u0438\u0432\u0435\u0441\u0442\u0438 \u043a \u0435\u0451 \u0434\u0438\u0441\u043a\u0440\u0435\u0434\u0438\u0442\u0430\u0446\u0438\u0438. \u0412 \u0441\u0432\u043e\u0435\u0439 \u043a\u043d\u0438\u0433\u0435 Rebooting […]<\/p>\n","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"author":1,"featured_media":29302,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[702],"tags":[],"class_list":["post-39055","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\n","aioseo_head_json":{"title":"\ud83e\udd47Nem\u016f\u017eeme d\u016fv\u011b\u0159ovat AI syst\u00e9m\u016fm zalo\u017een\u00fdm pouze na hlubok\u00e9m u\u010den\u00ed | ProHoster","description":"Tento text.","canonical_url":"https:\/\/prohoster.info\/cs\/blog\/news\/my-ne-mozhem-doveryat-ii-sistemam-postroennym-na-odnom-lish-glubokom-obuchenii","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"cs_CZ","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\u041c\u044b \u043d\u0435 \u043c\u043e\u0436\u0435\u043c \u0434\u043e\u0432\u0435\u0440\u044f\u0442\u044c \u0418\u0418-\u0441\u0438\u0441\u0442\u0435\u043c\u0430\u043c, \u043f\u043e\u0441\u0442\u0440\u043e\u0435\u043d\u043d\u044b\u043c \u043d\u0430 \u043e\u0434\u043d\u043e\u043c \u043b\u0438\u0448\u044c \u0433\u043b\u0443\u0431\u043e\u043a\u043e\u043c \u043e\u0431\u0443\u0447\u0435\u043d\u0438\u0438 | ProHoster","og:description":"\u042d\u0442\u043e\u0442 \u0442\u0435\u043a\u0441\u0442.","og:url":"https:\/\/prohoster.info\/cs\/blog\/news\/my-ne-mozhem-doveryat-ii-sistemam-postroennym-na-odnom-lish-glubokom-obuchenii","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":"2019-10-31T19:27:36+00:00","article:modified_time":"2019-10-31T19:27:36+00:00","article:publisher":"https:\/\/www.facebook.com\/prohoster","article:author":"https:\/\/www.facebook.com\/prohoster"},"aioseo_meta_data":{"post_id":"39055","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":"2026-01-24 00:33:19","breadcrumb_settings":null,"limit_modified_date":false,"reviewed_by":null,"ai":null,"created":"2021-03-01 00:57:31","updated":"2026-01-24 00:33:19","focus_keyword":null,"additional_keywords":null,"truseo_locale":null},"gt_translate_keys":[{"key":"link","format":"url"}],"_links":{"self":[{"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/posts\/39055","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/comments?post=39055"}],"version-history":[{"count":0,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/posts\/39055\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/media\/29302"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/media?parent=39055"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/categories?post=39055"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/tags?post=39055"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}