Iphayinti elandelayo oyithelayo ingase ikukitaze ama-tastebuds wakho ngenxa yokuncenceza kobuchwepheshe ngobuhlakani bokwenziwa. Kungase kube yinto enkulu elandelayo endaweni yokudla nesiphuzo-impela ukuthengisa okulula kuno ubhiya wangasese futhi i-steak yokwenziwa.
Kakade ukunambitheka okuzuziwe kolwanga oluthile, ukunambitheka kukabhiya kungenziwa kube ngcono ngosizo lwe-AI. Iqembu lososayensi lidlulele ngalé kokubuza i-chatbot okuthile okungahleliwe ukunambitheka kwe-soda esethulweni esimnandi. Ngokuhlanganisa ukuhlaziya okubanzi kwamakhemikhali namamodeli e-AI aqeqeshwe kudatha yezinzwa, abacwaningi bahlonze izinhlanganisela ezithile ezingathuthukisa ukunambitheka kweziphuzo zakho ozithandayo.
"Ucwaningo lwethu luveza ukuthi idatha enkulu kanye nokufunda komshini kungambula kanjani izixhumanisi eziyinkimbinkimbi phakathi kwamakhemikhali okudla, ukunambitheka, kanye nombono wabathengi," the ucwaningo iphepha eshicilelwe namuhla kujenali Nature nokuxhumana uthi. "Lokhu kuvula indlela yokwakhiwa kwemikhiqizo yokudla eyenziwe yanoveli enephunga elimnandi, elijabulisa isixuku kusetshenziswa indlela eqhutshwa idatha esikhundleni sokuzama ukwenza iphutha."
Ucwaningo, olwenziwa ithimba lamazwe ngamazwe lochwepheshe abangaphezu kweshumi nambili, lahlaziya obhiya baseBelgium abangaphezu kuka-250 abasebenzisa izitayela ezihlukahlukene. Ubhiya ngamunye wenze iphrofayili yamakhemikhali eqinile ukukala izakhiwo ezihlukene ezingaphezu kuka-200 njenge-ester, utshwala, ama-asidi nezinhlanganisela zephunga.
Lezi zigxivizo zeminwe zekhemikhali zabe sezifaniswa nokuhlolwa kwezinzwa eziningi ezivela kuphaneli yabanambithayo abathole izimfanelo zokunambitheka ezihlukene ezingu-50 kanye nezibuyekezo eziku-inthanethi ezingaphezu kuka-180,000 ezivela kwabathanda ubhiya abalinganisela iphunga elimnandi, ukunambitheka kanye nokwazisa kukonke.
Ngokusebenzisa le dathasethi enkulu, abacwaningi ngokulandelayo baqeqeshe futhi bahlola ama-algorithms okufunda omshini ukuze bahlukanise ubudlelwano obuyinkimbinkimbi phakathi kokwakheka kwamakhemikhali kabhiya kanye nama-flavour abonwayo kanye nokukhanga komthengi. I-algorithm eyodwa evamile yavela njengomdlali ophezulu, esebenza kahle kakhulu kunezindlela ezivamile zezibalo.
"I-algorithm esebenza kahle kakhulu, i-gradient booster, ikhiqiza amamodeli adlula ngokuphawulekayo ukubikezela okusekelwe kwizibalo ezivamile futhi ibikezela ngokunembile izici zokudla eziyinkimbinkimbi kanye nokwazisa kwabathengi kumaphrofayili wamakhemikhali," kufunda ucwaningo. Abacwaningi bathi le ndlela “ivumela ukuhlonza izinhlanganisela ezithile nezingalindelekile njengezisusa ukunambitheka nokwazisa kukabhiya.”
Ukuthuthukisa i-gradient kuyindlela yokufunda yomshini ehlanganisa amamodeli amaningi amancane, anganembile ibe enkulu ekwaziyo ukubikezela okuhle. Isetshenziswa kakhulu emabhange nasekunakekelweni kwezempilo kanye nasemikhankasweni yokumaketha ezama ukubikezela imiphumela yemizamo ethile.
Ukwenza ubhiya omuhle kwakungeyona injongo yakhona eyinhloko, kodwa ukukhuphula i-gradient kwasebenza kakhulu kunamanye amamodeli okuqagela ahlolwe ithimba locwaningo.
Abacwaningi bathola ukuthi “ukugxiliswa kwenhlanganisela ye-flavour akuhlangani ngaso sonke isikhathi nombono,” okusho ukuthi abathengi bathambekele ekwaziseni ama-nuances amaningi ngaphezu kokunambitheka nje lapho bejabulela ubhiya omnandi. Lokhu kunzima ukubikezela ngamamodeli ajwayelekile kanye nokuhlaziya okulula, futhi kungesinye sezizathu ezenza ithimba licabangele inani elikhulu kangaka lokuguquguquka.
Ngokusebenzisa amasu okufunda ngomshini atolikayo, ososayensi bathola izinhlanganisela ezingalindelekile ezithonya kakhulu ikhwalithi kabhiya. Isibonelo, i-methanethiol ne-ethyl phenyl acetate ngokuvamile zihlotshaniswa nokunganyakazi, kodwa zingenza ubhiya unambitheke kahle ngemithamo emincane.
Abashayeli be-flavour abajwayelekile njenge-ethyl acetate ne-lactic acid baphinde bahlolwa.
Ukuqinisekisa amamodeli abo, abacwaningi benze izivivinyo zokunambitha lapho baphuza khona obhiya abalinganiselwe kabi ngamakhompiyutha akhonjwe ngomshini akhulisa ukwazisa. Lokhu kulungiswa okulula kuholele ekwenyukeni okuphawulekayo kuzikolo ze-flavour zabaphaneli abaqeqeshiwe kanye nezilinganiso ezincanyelwayo eziphelele zamabhiya ashintshiwe uma kuqhathaniswa nowangempela.
Abaphisi botshwala nabathandi bobhiya bomhlaba maduzane kungase kudingeke ukuthi babonge ubuhlakani bokwenziwa ngokuthuthukisa ukujabulela kwethu esinye seziphuzo ezidala nesithandwa kakhulu ebantwini. Kodwa izicelo zidlulela ngalé kukabhiya ongcono. Iphrofayili efanayo yamakhemikhali enikwe amandla yi-AI ingase ithuthukise ukunambitheka kwayo yonke into kusukela kwezinye izinhlobo zenyama esekelwe ezitshalweni kuya kumasoda anoshukela omncane kanye nokudla okulula.
"Umsebenzi wamanje ubonisa ukuthi naphezu kwezithiyo ezithile ezibalulekile ezisele, ukuhlanganisa intuthuko yakamuva ekuhlaziyweni kwamakhemikhali, ukuhlaziya izinzwa, kanye nezindlela zesimanje zokufunda umshini kunikeza izindlela ezithokozisayo ze-chemistry yokudla kanye nobunjiniyela," kusho abacwaningi. "Maduze, lawa mathuluzi angase anikeze izixazululo ekulawuleni ikhwalithi nokuthuthukiswa kweresiphi, kanye nezindlela ezintsha zokucwaninga kwesayensi yezinzwa nokunambitheka."
Ngokuzayo uma uvula ubhiya olinganiselayo, omnandi, kungase kuphikisane nesisho sakudala sokuthi 'okwenziwayo' kuphansi futhi kunikeze isikhashana ukuphakamisa ingilazi ku-AI.
eyahlelwa Ryan Ozawa.
Hlala unolwazi lwezindaba ze-crypto, thola izibuyekezo zansuku zonke ebhokisini lakho lokungenayo.
- I-SEO Powered Content & PR Distribution. Khuliswa Namuhla.
- I-PlatoData.Network Vertical Generative Ai. Zinike Amandla. Finyelela Lapha.
- I-PlatoAiStream. I-Web3 Intelligence. Ulwazi Lukhulisiwe. Finyelela Lapha.
- I-PlatoESG. Ikhabhoni, I-CleanTech, Amandla, Environment, Ilanga, Ukuphathwa Kwemfucuza. Finyelela Lapha.
- I-PlatoHealth. I-Biotech kanye ne-Clinical Trials Intelligence. Finyelela Lapha.
- Source: https://decrypt.co/223564/science-taps-ai-to-make-beer-taste-better
