{"id":92946,"date":"2025-11-24T05:54:51","date_gmt":"2025-11-24T04:54:51","guid":{"rendered":"https:\/\/www.infokuryr.cz\/n\/?p=92946"},"modified":"2025-11-24T05:54:51","modified_gmt":"2025-11-24T04:54:51","slug":"prelom-umele-inteligence-ve-cteni-mysli-vedci-dekodovali-myslenky-bez-pouziti-mozkovych-implantatu","status":"publish","type":"post","link":"https:\/\/www.infokuryr.cz\/n\/2025\/11\/24\/prelom-umele-inteligence-ve-cteni-mysli-vedci-dekodovali-myslenky-bez-pouziti-mozkovych-implantatu\/","title":{"rendered":"P\u0159elom um\u011bl\u00e9 inteligence ve \u010dten\u00ed mysli: V\u011bdci dek\u00f3dovali my\u0161lenky bez pou\u017eit\u00ed mozkov\u00fdch implant\u00e1t\u016f"},"content":{"rendered":"
<\/div>\n

V\u00fdzkumn\u00edci z\u00a0<\/span>Kalifornsk\u00e9 univerzity v Berkeley<\/span><\/em>\u00a0a japonsk\u00fdch laborato\u0159\u00ed\u00a0<\/span>NTT Communication Science<\/span><\/em>\u00a0Laboratories vyvinuli um\u011blou inteligenci (AI) schopnou p\u0159elo\u017eit mozkovou aktivitu do \u010diteln\u00e9ho textu \u2013 bez invazivn\u00edch implant\u00e1t\u016f, co\u017e je ohromuj\u00edc\u00ed skok sm\u011brem.<\/span><\/strong><\/p>\n

Technologie s n\u00e1zvem \u201etitulkov\u00e1n\u00ed my\u0161lenek\u201c vyu\u017e\u00edv\u00e1 funk\u010dn\u00ed magnetickou rezonanci a um\u011blou inteligenci k rekonstrukci my\u0161lenek s p\u0159ekvapuj\u00edc\u00ed p\u0159esnost\u00ed, co\u017e vyvol\u00e1v\u00e1 nad\u011bje na nov\u00e9 medic\u00ednsk\u00e9 objevy a z\u00e1rove\u0148 znepokojen\u00ed nad bezprecedentn\u00edm naru\u0161en\u00edm soukrom\u00ed.<\/span><\/p>\n

Funk\u010dn\u00ed magnetick\u00e1 rezonance (fMRI) je v\u00fdkonn\u00e1 neurozobrazovac\u00ed technika, kter\u00e1 umo\u017e\u0148uje v\u00fdzkumn\u00edk\u016fm a l\u00e9ka\u0159\u016fm mapovat mozkovou aktivitu detekc\u00ed souvisej\u00edc\u00edch zm\u011bn v pr\u016ftoku krve.<\/span><\/p>\n

fMRI je cenn\u00fdm n\u00e1strojem pro zkoum\u00e1n\u00ed mozkov\u00fdch funkc\u00ed a m\u00e1 \u0159adu aplikac\u00ed ve v\u00fdzkumu a klinick\u00e9m prost\u0159ed\u00ed. Je v\u0161ak nezbytn\u00e9 p\u0159istupovat k \u00fadaj\u016fm a v\u00fdsledk\u016fm fMRI kriticky, p\u0159i\u010dem\u017e je t\u0159eba myslet na jejich omezen\u00ed a v\u00fdzvy p\u0159i interpretaci v\u00fdstup\u016f.<\/span><\/p>\n

Syst\u00e9m se spol\u00e9h\u00e1 na modely hlubok\u00e9ho u\u010den\u00ed, kter\u00e9 jsou natr\u00e9nov\u00e1ny na interpretaci neuronov\u00fdch vzorc\u016f spojen\u00fdch s vizu\u00e1ln\u00edm a s\u00e9mantick\u00fdm zpracov\u00e1n\u00edm.<\/span><\/p>\n

V experimentech \u00fa\u010dastn\u00edci sledovali tis\u00edce kr\u00e1tk\u00fdch videoklip\u016f b\u011bhem fMRI vy\u0161et\u0159en\u00ed. Model um\u011bl\u00e9 inteligence analyzoval tyto skeny spolu s p\u00edsemn\u00fdmi titulky k vide\u00edm a u\u010dil se spojovat mozkovou aktivitu s konkr\u00e9tn\u00edmi v\u00fdznamy.<\/span><\/p>\n

P\u0159i testov\u00e1n\u00ed um\u011bl\u00e1 inteligence dek\u00f3dovala mozkovou aktivitu do popisn\u00fdch v\u011bt.<\/span><\/p>\n

Nap\u0159\u00edklad pot\u00e9, co si \u00fa\u010dastn\u00edk prohl\u00e9dl video n\u011bkoho, kdo sk\u00e1\u010de z vodop\u00e1du, syst\u00e9m nejprve uhodl \u201epramenov\u00fd tok\u201c a pot\u00e9 up\u0159esnil sv\u016fj v\u00fdstup na \u201e\u010dlov\u011bk sk\u00e1\u010de p\u0159es hlubok\u00fd vodop\u00e1d na horsk\u00e9m h\u0159ebenu\u201c. P\u0159esto\u017ee to nebylo doslovn\u011b dokonal\u00e9, s\u00e9mantick\u00e1 podobnost byla pozoruhodn\u00e1.<\/span><\/p>\n

Tomoyasu Horikawa, vedouc\u00ed v\u00fdzkumn\u00edk v\u00a0<\/span>NTT Communication Science Laboratories<\/span><\/em>\u00a0, vysv\u011btlil, \u017ee\u00a0<\/span>um\u011bl\u00e1 inteligence<\/span><\/a>\u00a0generuje text porovn\u00e1v\u00e1n\u00edm vzorc\u016f mozkov\u00e9 aktivity s nau\u010den\u00fdmi sekvencemi \u010d\u00edsel odvozen\u00fdmi z video titulk\u016f.<\/span><\/p>\n

Horikawa \u0159\u00edk\u00e1, \u017ee tato metoda dok\u00e1\u017ee \u201evytv\u00e1\u0159et komplexn\u00ed popisy vizu\u00e1ln\u00edho obsahu, a to i bez spol\u00e9h\u00e1n\u00ed na oblasti mozku souvisej\u00edc\u00ed s jazykem\u201c, co\u017e nazna\u010duje potenci\u00e1ln\u00ed vyu\u017eit\u00ed pro pacienty s poruchami \u0159e\u010di.<\/span><\/p>\n

L\u00e9ka\u0159sk\u00fd slib versus riziko pro soukrom\u00ed<\/span><\/h2>\n

Tato technologie by mohla zp\u016fsobit revoluci v komunikaci lid\u00ed s onemocn\u011bn\u00edmi, jako jsou amyotrofick\u00e1 later\u00e1ln\u00ed skler\u00f3za (ALS), syndrom uzam\u010den\u00ed (\u010dlov\u011bk pln\u011b p\u0159i v\u011bdom\u00ed, ale nehybn\u00fd) nebo t\u011b\u017ek\u00e1 af\u00e1zie (\u010dlov\u011bk neschopn\u00fd komunikovat).<\/span><\/p>\n

Psycholog Scott Barry Kaufman, kter\u00fd nen\u00ed spojen se studi\u00ed, to nazval \u201ev\u00fdznamn\u00fdm miln\u00edkem\u201c pro neverb\u00e1ln\u00ed jedince. Etici v\u0161ak varuj\u00ed p\u0159ed hrozn\u00fdmi n\u00e1sledky, pokud se takov\u00e1 moc zneu\u017eije.<\/span><\/p>\n

Marcello Ienca, profesor etiky um\u011bl\u00e9 inteligence a neurov\u011bdy na\u00a0<\/span>Technick\u00e9 univerzit\u011b v Mnichov\u011b<\/span><\/em>\u00a0, varoval:\u00a0<\/span>\u201ePokud se tam dostaneme, mus\u00edme m\u00edt velmi, velmi p\u0159\u00edsn\u00e1 pravidla, pokud jde o ud\u011blov\u00e1n\u00ed p\u0159\u00edstupu k mysl\u00edm a mozk\u016fm lid\u00ed.\u201c<\/span><\/em><\/p>\n

Pouk\u00e1zal na rizika zve\u0159ejn\u011bn\u00ed citliv\u00fdch du\u0161evn\u00edch \u00fadaj\u016f, v\u010detn\u011b \u010dasn\u00fdch p\u0159\u00edznak\u016f demence nebo deprese. Syst\u00e9m v sou\u010dasnosti vy\u017eaduje rozs\u00e1hlou spolupr\u00e1ci. \u00da\u010dastn\u00edci mus\u00ed podstoupit hodiny fMRI vy\u0161et\u0159en\u00ed a z\u00e1rove\u0148 si prohl\u00ed\u017eet kur\u00e1torem vybran\u00fd obsah.<\/span><\/p>\n

Alex Huth z\u00a0\u00a0<\/span>Kalifornsk\u00e9 univerzity v Berkeley<\/span><\/em>\u00a0uklidnil skeptiky slovy:\u00a0<\/span>\u201eNikdo zat\u00edm neuk\u00e1zal, \u017ee je to mo\u017en\u00e9,\u201c<\/span><\/em>\u00a0pokud jde o neopr\u00e1vn\u011bnou extrakci my\u0161lenek. Slovo \u201ezat\u00edm\u201c v\u0161ak zlov\u011bstn\u011b p\u0159etrv\u00e1v\u00e1.<\/span><\/p>\n

Studie uzn\u00e1v\u00e1 etick\u00e1 dilemata, zejm\u00e9na t\u00fdkaj\u00edc\u00ed se \u201edu\u0161evn\u00edho soukrom\u00ed\u201c. \u0141ukasz Szoszkiewicz, expert na neuropr\u00e1va, nal\u00e9h\u00e1 na preventivn\u00ed z\u00e1ruky:\u00a0<\/span>\u201eNeurov\u011bda se rychle rozv\u00edj\u00ed a asisten\u010dn\u00ed potenci\u00e1l je obrovsk\u00fd, ale ochrana du\u0161evn\u00edho soukrom\u00ed a svobody my\u0161len\u00ed nem\u016f\u017ee \u010dekat.\u201c<\/span><\/em><\/p>\n

Navrhovan\u00e1 \u0159e\u0161en\u00ed zahrnuj\u00ed mechanismy \u201eodem\u010den\u00ed\u201c, kde u\u017eivatel\u00e9 v\u011bdom\u011b aktivuj\u00ed dek\u00f3dov\u00e1n\u00ed pomoc\u00ed ment\u00e1ln\u00edho kl\u00ed\u010dov\u00e9ho slova.<\/span><\/p>\n

Horikawa zd\u016fraznil omezen\u00ed \u2013 um\u011bl\u00e1 inteligence nezvl\u00e1d\u00e1 neobvykl\u00e9 nebo nep\u0159edv\u00eddateln\u00e9 situace (nap\u0159\u00edklad \u201emu\u017e kou\u0161e psa\u201c). S rostouc\u00ed sofistikovanost\u00ed model\u016f um\u011bl\u00e9 inteligence se v\u0161ak hranice mezi asisten\u010dn\u00edm n\u00e1strojem a invazivn\u00edm sledov\u00e1n\u00edm st\u00edr\u00e1.<\/span><\/p>\n

Spole\u010dnost\u00a0<\/span>Neuralink<\/span><\/em>\u00a0Elona Muska a dal\u0161\u00ed neurotechnologick\u00e9 firmy se p\u0159edb\u00edhaj\u00ed sm\u011brem k rozhran\u00edm mozek-po\u010d\u00edta\u010d pro spot\u0159ebitele. S rychl\u00fdm rozvojem um\u011bl\u00e9 inteligence se v\u015bak zvy\u0161uje riziko zneu\u017eit\u00ed ze strany spole\u010dnost\u00ed nebo vl\u00e1d.<\/span><\/p>\n

Ienca varuje:\u00a0<\/span>\u201eToto je kone\u010dn\u00e1 v\u00fdzva pro soukrom\u00ed.\u201c<\/span><\/em><\/p>\n

Z\u00e1v\u011br<\/span><\/h2>\n

Zat\u00edm je tato technologie omezena na laborato\u0159e a z\u00e1vis\u00ed na objemn\u00fdch magnetick\u00fdch rezonan\u010dn\u00edch p\u0159\u00edstroj\u00edch a ochotn\u00fdch \u00fa\u010dastn\u00edkech. Ale s klesaj\u00edc\u00edmi v\u00fdpo\u010detn\u00edmi n\u00e1roky a zdokonalov\u00e1n\u00edm um\u011bl\u00e9 inteligence se r\u00fdsuje hrozba sledov\u00e1n\u00ed my\u0161lenek v re\u00e1ln\u00e9m \u010dase.<\/span><\/p>\n

P\u0159esto\u017ee \u201etitulkov\u00e1n\u00ed mysli\u201c nab\u00edz\u00ed lidem s poruchami \u0159e\u010di potenci\u00e1l zm\u011bnit \u017eivot, jeho temn\u011bj\u0161\u00ed d\u016fsledky nelze ignorovat.<\/span><\/p>\n

Stejn\u00e9 n\u00e1stroje, kter\u00e9 odemykaj\u00ed komunikaci, by mohly tak\u00e9 zni\u010dit posledn\u00ed ba\u0161tu soukrom\u00ed: na\u0161e vnit\u0159n\u00ed my\u0161lenky. Jak zd\u016fraznil Szoszkiewicz:\u00a0<\/span>\u201eNeur\u00e1ln\u00ed data bychom m\u011bli automaticky pova\u017eovat za citliv\u00e1.\u201c<\/span><\/em><\/p>\n

Ot\u00e1zkou nen\u00ed, zda se tato technologie bude vyv\u00edjet. Ot\u00e1zkou je, zda ji lidstvo dok\u00e1\u017ee ovl\u00e1dnout d\u0159\u00edve, ne\u017e ona ovl\u00e1dne n\u00e1s.<\/span><\/p>\n

Prohl\u00e9dn\u011bte si n\u00e1sleduj\u00edc\u00ed video (v angli\u010dtin\u011b) o velk\u00fdch technologick\u00fdch spole\u010dnostech, kter\u00e9 ji\u017e maj\u00ed technologii \u010dten\u00ed my\u0161lenek, kterou v\u00e1m je\u0161t\u011b nesd\u011blili:<\/span><\/p>\n