Ubukrelekrele bokwenziwa buhambela phambili ngokutsiba kunye nemida kwicandelo lezemozulu, kunye ne-Google DeepMind ibethe itafile kunye nenkqubo yayo entsha, i-GenCast AI, eyenzelwe ukuguqula indlela esiyiqonda ngayo kunye nokuqikelela imozulu. Le modeli ayinconywa nje kuphela njengeyona iphambili kuhlobo lwayo, kodwa ikwathembisa ukuguqula i-meteorology njengoko sisazi, ngenxa yesakhono sayo sokuvelisa uqikelelo ngesantya kunye ngokuchanekileyo ezishiya ngasemva iindlela zemveli.
Yintoni iGenCast kwaye yenza njani umahluko?
I-GenCast yimodeli yobukrelekrele eyenziweyo esekwe kwidatha yembali ukusuka kwiminyaka engama-40 edlulileyo, eqokelelwe ngokukodwa phakathi kwe-1979 kunye ne-2018 yi-European Centre for Medium-Range Weather Forecasts (ECMWF). Ngokungafaniyo neemodeli zesiNtu ezisekelwe kwi-equations zomzimba kwaye zifuna ii-supercomputers ezinamandla ukuba zisebenze, i-GenCast ivelele ngendlela yayo enokwenzeka. Oku kuthetha ukuba ayiqikelele nje imeko enye, kodwa ibonelela ngoluhlu lwezinto ezinokwenzeka, inika okunokwenzeka kwiziphumo ezahlukeneyo zemozulu.
Ukuchaneka kweGenCast kuyamangalisa. Kwiimvavanyo ezenziwe ngedatha esuka kwi-2019, le modeli iphumelele inkqubo ye-ECMWF ye-ENS kwi-97.2% yamatyala, ifikelele kwi-99.8% yokuchaneka kweengqikelelo kwiiyure ezingama-36. La manani ayenza ibe sisixhobo esibalulekileyo kungekuphela nje kuqikelelo lwemihla ngemihla, kodwa nakwiziganeko ezigqithisileyo ezinjengezaqhwithi, amaza obushushu kunye nezitshingitshane zetropikhi.
Iinzuzo zobugcisa kwiindlela zemveli
Enye yezona zinto zinomdla kakhulu ngeGenCast kukukwazi ukuvelisa i-15-day forecast in imizuzu nje esibhozo usebenzisa iyunithi Google Cloud TPU v5. Oku kuyangqubana neeyure ezifunwa ziinkqubo zemveli, ezifana ne-ENS, kwiikhompyutha ezinkulu ezingamashumi amawaka eprosesa. Oku kongiwa kwemithombo akubonisi ukuqhubela phambili kobugcisa kuphela, kodwa kukwayibeka njengesixhobo esifikelelekayo kumacandelo amaninzi kunye namazwe anemida yobuchwepheshe.
I-GenCast isebenzisa i-diffusion modeling algorithms, iteknoloji ekwanika amandla izixhobo ezivelisa imifanekiso kunye neetekisi. Ukulungelelaniswa kwayo ukuze isebenze kunye nejiyometri engqukuva yoMhlaba ivumela ukuba iqonde intsebenziswano entsonkothileyo phakathi kwezinto eziguquguqukayo zasemoyeni ezifana noxinzelelo, ubushushu, umoya kunye nokufuma. Ukongeza, amandla ayo anokwenzeka anceda ukunciphisa ukungaqiniseki, ukubonelela ngeengqikelelo ezithembekileyo nakwiimeko ezinzima.
Ukusetyenziswa okusebenzayo kunye nekamva loqikelelo lwemozulu
Ukongeza kokuchaneka kwayo kwiimeko ezinzima, i-GenCast inezicelo ezicacileyo ezicacileyo. Amacandelo afana nolawulo olungxamisekileyo, ezolimo kunye nokucwangciswa kwamandla angazuza kakhulu kwiinkcazo ezicacileyo kunye nezikhawulezayo. Ngokomzekelo, iinkampani zamandla zinokulindela utshintsho kwimveliso yamandla omoya, ngelixa iinkonzo zongxamiseko zinokulungiselela kangcono iinkanyamba kunye nezivunguvungu ezinzima.
Kwixesha elizayo, le modeli kulindeleke ukuba ivele ngakumbi. Nangona okwangoku ixhomekeke kwidatha yembali ukuqeqesha izibikezelo zayo, izazinzulu emva kweGenCast ziphanda ithuba lokusebenzisa idatha yokujonga yamva nje, njengomswakama wexesha langempela kunye nokufundwa komoya, ukuqhubela phambili ukuphucula ukuchaneka kwayo.
Umzekelo ovulelekileyo kuluntu
Enye into entsha yeGenCast kukuvuleka kwayo. I-Google inqume ukwenza ikhowudi yemodeli kunye nedatha ekhoyo, evumela abaphandi kunye namaziko ukuba ayisebenzise kwaye ayilungelelanise neemfuno zabo ezithile. Oku akukhuthazi kuphela ukusebenzisana kwehlabathi, kodwa kukwakhuthaza uphuhliso lwezicelo ezitsha kunye nokuphuculwa kwesi siseko siluqilima.
Nangona kunjalo, iingcali zithi kusekho indlela yokuhamba phambi kokuba iimodeli ezisekwe kwi-AI zitshintshe ngokupheleleyo iindlela zemveli. Nangona i-GenCast ibonisa amandla amakhulu, isajongene nemingeni efana nokubamba izinto ezithile ezinzima eziphathekayo, eziyimfuneko kwiziganeko zexesha elide.
I-GenCast sele iphawula ngaphambili nasemva kwecandelo lezemozulu, ebonisa indlela ubukrelekrele bokwenziwa bunokoyisa ngayo imida yeenkqubo zemveli, inika uqikelelo olukhawulezayo, oluchanekileyo nolufikelelekayo. Ngokukwazi kwayo ukuphatha imicimbi egqithileyo kunye nendlela yayo evulelekileyo kuluntu lwezenzululwazi, le modeli ithembisa ukuba sisixhobo esibalulekileyo kumlo ochasene nemingeni yemozulu yehlabathi.
Ndingumntu othanda itekhnoloji ojike umdla wakhe we "geek" waba ngumsebenzi. Ndichithe ngaphezulu kweminyaka eli-10 yobomi bam ndisebenzisa itekhnoloji yokusika kwaye ndikhenkceza ngazo zonke iintlobo zeenkqubo ngenxa yokufuna ukwazi okumsulwa. Ngoku ndiqeqeshelwe ubugcisa bekhompyutha nakwimidlalo yevidiyo. Oku kungenxa yokuba ngaphezu kweminyaka emi-5 ndibhalela iiwebhusayithi ezahlukeneyo kwitekhnoloji kunye nemidlalo yevidiyo, ndisenza amanqaku afuna ukukunika ulwazi oludingayo ngolwimi oluqondakalayo kuye wonke umntu.
Ukuba unayo nayiphi na imibuzo, ulwazi lwam lusuka kuyo yonke into enxulumene nenkqubo yokusebenza yeWindows kunye ne-Android yeefowuni eziphathwayo. Kwaye ukuzinikela kwam kukuwe, ndihlala ndikulungele ukuchitha imizuzu embalwa kwaye ndikuncede usombulule nayiphi na imibuzo onokuba nayo kweli lizwe le-intanethi.