Impi yama-algorithms

Uma kuziwa ekusetshenzisweni kobuhlakani bokwenziwa kwezempi, i-nightmare yezinganekwane zesayensi ivuka ngokushesha, i-AI ehlubukayo nebulalayo evukela isintu ukuyibhubhisa. Ngeshwa, ukwesaba kwamasosha nabaholi ukuthi "isitha sizosifica" kunamandla ngendlela efanayo ekuthuthukisweni kwama-algorithms zempi.

I-Algorithmic Warfarelokho, ngokusho kwabaningi, okungashintsha ngokuyisisekelo isimo senkundla yempi njengoba sazi, ngokuyinhloko ngenxa yokuthi impi izoshesha, idlula kude ikhono labantu lokwenza izinqumo. Ujenene waseMelika Jack Shanahan (1), inhloko ye-US Joint Center for Artificial Intelligence, igcizelela, nokho, ukuthi ngaphambi kokwethula ubuhlakani bokwenziwa ezindaweni zezikhali, kufanele siqinisekise ukuthi lezi zinhlelo zisalawulwa abantu futhi aziqali izimpi ngokwazo.

"Uma isitha sinemishini nama-algorithms, sizolahlekelwa yile ngxabano"

Ikhono lokushayela impi ye-algorithmic isekelwe ekusetshenzisweni kwentuthuko kubuchwepheshe bamakhompiyutha ezindaweni ezintathu ezibalulekile. Okokuqala amashumi eminyaka okukhula okubonakalayo kwamandla ekhompyuthalokhu kuthuthukise kakhulu ukusebenza komshini wokufunda. Okwesibili ukukhula ngokushesha kwezinsiza "Idatha enkulu", okungukuthi, enkulu, ngokuvamile ezenzakalelayo, ephethwe futhi edalwe ngokuqhubekayo amasethi edatha afanele ukufundwa komshini. Owesithathu ukhathazekile ukuthuthukiswa okusheshayo kobuchwepheshe be-cloud computing, lapho amakhompyutha angakwazi ukufinyelela kalula izinsiza zedatha futhi azicubungule ukuze axazulule izinkinga.

I-War Algorithmnjengoba kuchazwe ochwepheshe, kufanele kuqala ivezwe nge ikhodi yekhompyutha. Okwesibili, kufanele kube umphumela weplatifomu ekwazi ukuqoqa ulwazi nokwenza izinqumo, ithathe izinqumo, okungenani ngokombono, ezingadingi ukungenelela kwabantu. Okwesithathu, okubonakala kusobala, kodwa akunjalo, ngoba kusesenzweni kuphela lapho kucaca khona ukuthi inqubo ehloselwe enye into ingaba wusizo empini futhi ngokuphambene nalokho, kufanele ikwazi ukusebenza ngaphansi kwezimo. ukungqubuzana kwezikhali.

Ukuhlaziywa kwezikhombisi-ndlela ezingenhla nokusebenzisana kwazo kubonisa lokho impi ye-algorithmic akuwona ubuchwepheshe obuhlukile njengokuthi, isibonelo. izikhali zamandla noma imicibisholo ye-hypersonic. Imiphumela yawo ibanzi futhi kancane kancane iba yonke indawo ezimpini. Okokuqala izimoto zempi baba nobuhlakani, okungenzeka benze ibutho lezokuvikela elilisebenzisayo lisebenze kahle futhi liphumelele. Imishini enjalo ehlakaniphile inemingcele ecacile okudingeka iqondwe kahle.

"" Kusho uShanahan ekwindla edlule engxoxweni nowayeyi-CEO yakwa-Google u-Eric Schmidt kanye nephini likamongameli we-Google wezindaba zamazwe ngamazwe u-Kent Walker. "".

Umbiko osalungiswa woMkhandlu Wezokuphepha Kazwelonke wase-US nge-AI ubhekisela eChina izikhathi ezingaphezu kuka-50, ugqamisa umgomo osemthethweni waseChina wokuba umholi womhlaba ku-AI ngo-2030 (bona futhi: ).

Lawa mazwi ashiwo eWashington engqungqutheleni ekhethekile eyenzeka ngemva kokuba isikhungo saseShanakhan esishiwo ngenhla sethule umbiko walo wokuqala kuCongress, esilungiswe ngokubambisana nochwepheshe abadumile emkhakheni wezobunhloli bokwenziwa, okuhlanganisa noMqondisi Wocwaningo weMicrosoft u-Eric Horwitz, i-AWS CEO u-Andy Jassa kanye Umcwaningi Oyinhloko We-Google Cloud u-Andrew Moore. Umbiko wokugcina uzoshicilelwa ngo-Okthoba 2020.

Abasebenzi bakwa-Google bayabhikisha

Eminyakeni embalwa edlule, iPentagon yangenela. impi ye-algorithmic kanye nenani lamaphrojekthi ahlobene ne-AI ngaphansi kwephrojekthi ye-Maven, esekelwe ekusebenzisaneni nezinkampani zobuchwepheshe, okuhlanganisa i-Google kanye neziqalo ezifana ne-Clarifai. Ngokuyinhloko kwakumayelana nokusebenza ukuhlakanipha okungekhona okwangempelaukwenza lula ukuhlonza izinto.

Lapho kwaziwa ngokubamba iqhaza kwe-Google kulo msebenzi entwasahlobo ka-2018, izinkulungwane zabasebenzi be-Mountain View giant basayina incwadi evulekile bebhikishela ukubamba iqhaza kwenkampani ezimpini. Ngemva kwezinyanga zezinxushunxushu zabasebenzi I-Google yamukele isethi yayo yemithetho ye-AIokuhlanganisa ukuvinjelwa kokuhlanganyela emicimbini.

I-Google futhi izibophezele ekuqedeni inkontileka ye-Project Maven ekupheleni kuka-2019. Ukuphuma kwe-Google akuqedanga i-Project Maven. Yathengwa nguPeter Thiel's Palantir. I-Air Force kanye ne-US Marine Corps bahlela ukusebenzisa izimoto zasemoyeni ezikhethekile ezinganamuntu, njenge-Global Hawk, njengengxenye yephrojekthi ye-Maven, ngayinye okufanele iqaphe ngeso lengqondo amakhilomitha-skwele ayi-100.

Ngesikhathi kwenzekani ngaseProject Maven, kwacaca ukuthi amasosha ase-US alidinga ngokuphuthumayo ifu lawo. Yilokhu uShanahan akusho ngesikhathi kuqhubeka ingqungquthela. Lokhu kubonakale lapho izithombe zevidiyo nezibuyekezo zesistimu kufanele zithuthwe ngamaloli ziyiswe ezindaweni zamasosha ezisabalele endaweni yonke. Esakhiweni ikhompuyutha yamafu ehlanganisiwe, okuzosiza ukuxazulula izinkinga zalolu hlobo, njengengxenye yephrojekthi yengqalasizinda ye-IT ehlanganisiwe yebutho le-Jedi, i-Microsoft, i-Amazon, i-Oracle ne-IBM. I-Google ayingenxa yamakhodi abo okuziphatha.

Kusobala esitatimendeni sikaShanahan ukuthi uguquko olukhulu lwe-AI kwezempi lusanda kuqala. Futhi indima yesikhungo sayo embuthweni wamasosha ase-US iyakhula. Lokhu kubonakala ngokucacile esilinganisweni sesabelomali se-JAIC. Ngo-2019, ifinyelele ngaphansi kwama- $ 90 wezigidi. Ngo-2020, kufanele ngabe isivele iyizigidi ezingama- $ 414, noma cishe amaphesenti ayishumi esabelomali se-AI se-Pentagon sika-$10 billion.

Umshini ubona isosha elizinikele

Amasosha ase-US asehlome ngezinhlelo ezifana ne-Phalanx (2), okuwuhlobo lwesikhali esizimele esisetshenziswa emikhunjini ye-US Navy ukuhlasela imicibisholo engenayo. Uma kutholwa umcibisholo, uzivulela ngokuzenzakalelayo futhi ucekele phansi yonke into esendleleni yawo. Ngokusho kwabakwaFord, angakwazi ukuhlasela ngemicibisholo emine noma emihlanu engxenyeni yomzuzwana ngaphandle kokuthi adlule abheke ngakunye okuhlosiwe.

Esinye isibonelo i-semi-autonomous Harpy (3), uhlelo lwezohwebo olungenamuntu. I-harpy isetshenziselwa ukucekela phansi ama-radar ezitha. Isibonelo, ngo-2003, ngenkathi i-US yethula isiteleka e-Iraq eyayinezinhlelo zokunqamula i-radar esemoyeni, ama-drones enziwe e-Israel asiza ekutholeni nasekuwacekeleni phansi ukuze amaMelika andize ngokuphepha endaweni yomoya yase-Iraq.

3. Ukwethulwa kwe-drone yohlelo lwe-IAI Harpy

Esinye isibonelo esaziwayo sezikhali ezizimele Uhlelo lwaseKorea Samsung SGR-1, etholakala endaweni engavunyelwe ukulwa phakathi kweNyakatho neNingizimu Korea, eklanyelwe ukukhomba nokudubula abahlaseli ebangeni elingafika kumakhilomitha amane. Ngokusho kwencazelo, uhlelo "lungakwazi ukuhlukanisa phakathi komuntu ozinikelayo nomuntu ongazinikeli" ngokusekelwe endaweni yezandla zakhe noma ukuqaphela isikhundla sesikhali ezandleni zakhe.

4. Ukuboniswa kokutholwa kwesosha elizinikele ngohlelo lwe-Samsung SGR-1

Abantu baseMelika besaba ukusala ngemuva

Njengamanje, okungenani amazwe angama-30 emhlabeni jikelele asebenzisa izikhali ezizenzakalelayo ezinamazinga ahlukene okuthuthukiswa nokusetshenziswa kwe-AI. I-China, i-Russia kanye ne-United States babona ubuhlakani bokwenziwa njengento ebaluleke kakhulu ekwakheni isikhundla sabo sekusasa emhlabeni. "Noma ngubani ophumelele umjaho we-AI uzobusa umhlaba," uMongameli waseRussia uVladimir Putin etshela abafundi ngo-Agasti 2017. UMongameli wePeople's Republic of China, u-Xi Jinping, akakaze enze izitatimende eziphakeme kangaka kwabezindaba, kodwa ungumshayeli omkhulu walo myalelo othi i-China ibe namandla amakhulu emkhakheni we-AI ngo-2030.

Kunokukhathazeka okukhulayo e-US mayelana "nomphumela wesathelayithi", oye wabonisa ukuthi i-United States ayikuhlomele ngokwedlulele ukuhlangabezana nezinselele ezintsha ezibangelwa ubuhlakani bokwenziwa. Futhi lokhu kungaba yingozi ekuthuleni, uma kuphela ngenxa yokuthi izwe elisongelwa ukubusa lingase lifune ukuqeda inzuzo yamasu yesitha ngenye indlela, okungukuthi, ngempi.

Yize inhloso yasekuqaleni yephrojekthi ye-Maven bekuwukusiza ukuthola ama-Islamic ISIS fighters, ukubaluleka kwayo ekuthuthukisweni okuqhubekayo kwezinhlelo zezobunhloli bokwenziwa kwezempi kukhulu. Impi ye-elekthronikhi esekelwe kumarekhodi, iziqapha nezinzwa (okuhlanganisa iselula, endizayo) ihlotshaniswa nenani elikhulu lokugeleza kwedatha okungafani, engasetshenziswa kuphela ngempumelelo ngosizo lwama-algorithms e-AI.

Inkundla yempi eyingxube isiphenduke inguqulo yezempi ye-IoT, ulwazi olucebile olubalulekile lokuhlola izinsongo namathuba amaqhinga namasu. Ukwazi ukuphatha le datha ngesikhathi sangempela kunezinzuzo ezinkulu, kodwa ukwehluleka ukufunda kulolu lwazi kungaba yinhlekelele. Ikhono lokucubungula ngokushesha ukugeleza kolwazi kusuka ezinkundleni ezihlukahlukene ezisebenza ezindaweni eziningi linikeza izinzuzo ezimbili ezinkulu zempi: скорость i ukufinyeleleka. I-Artificial intelligence ikuvumela ukuthi uhlaziye izimo eziguqukayo zenkundla yempi ngesikhathi sangempela futhi ushaye ngokushesha nangendlela efanele, kuyilapho unciphisa ubungozi bamabutho akho.

Le nkundla yempi entsha nayo yonke indawo futhi. I-AI isenhliziyweni yalokho okubizwa ngokuthi ama-drone swarms, athole ukunakwa okukhulu eminyakeni yamuva. Ngosizo lwezinzwa ezitholakala yonke indawo, akuvumeli nje kuphela ukuthi i-drones ihambe endaweni enobutha, kodwa ekugcineni ingase ivumele ukwakheka kokwakheka okuyinkimbinkimbi kwezinhlobo ezihlukahlukene zezimoto zasemoyeni ezingenamuntu ezisebenza ezindaweni eziningi, ezinezikhali ezengeziwe ezivumela amaqhinga okulwa ayinkimbinkimbi, avumelane ngokushesha ne isitha. amasu ukuze asebenzise ithuba lenkundla yempi futhi abike izimo ezishintshayo.

Intuthuko ekuqokweni okuhlosiwe okusizwa yi-AI kanye nokuzulazula kuthuthukisa amathuba okusebenza ngempumelelo ezinhlelweni eziningi zokuvikela ezinamasu namasu, ikakhulukazi ukuvikela imicibisholo, ngokwenza ngcono izindlela zokuthola, ukulandelela nokuhlonza okuqondiwe.

kwandisa njalo amandla okulingisa kanye namathuluzi okudlala asetshenziselwa ukucwaninga izikhali zenuzi nezivamile. Ukumodela ngobuningi nokulingisa kuzobaluleka ukuthuthukisa uhlelo olubanzi lwezizinda eziningi lwezinhlelo eziqondiwe zokulawula ukulwa nomsebenzi onzima. I-AI iphinde inothise ukusebenzisana kwamaqembu amaningi (5). I-AI ivumela abadlali ukuthi bengeze futhi balungise okuhlukile kwegeyimu ukuze bahlole ukuthi izimo eziguqukayo (izikhali, ukubandakanyeka okuhlangene, amasosha engeziwe, njll.) zingathinta kanjani ukusebenza nokwenza izinqumo.

Kwamasosha, ukuhlonza into kuyisiqalo semvelo se-AI. Okokuqala, ukuhlaziya okuphelele nokusheshayo kwenani elikhulayo lezithombe nolwazi oluqoqwe kumasathelayithi nama-drones luyadingeka ukuze kutholwe izinto ezibalulekile zempi, njengemicibisholo, ukunyakaza kwamasosha kanye neminye imininingwane ehlobene nobuhlakani. Namuhla, inkundla yempi ihlanganisa yonke indawo—ulwandle, izwe, umoya, umkhathi nendawo yama-computer—emhlabeni wonke.

I-Cyberspacenjengesizinda sedijithali ngokwemvelo, ifaneleka ngokwemvelo izinhlelo zokusebenza ze-AI. Ngasohlangothini oluhlaselayo, i-AI ingasiza ekutholeni nasekuqondiseni amanodi enethiwekhi ngayinye noma ama-akhawunti angawodwana ukuze iwaqongelele, iwaphazamise, noma iwazise kabi. Ukuhlaselwa kwe-Cyber ​​kungqalasizinda yangaphakathi namanethiwekhi womyalo kungaba yinhlekelele. Mayelana nokuzivikela, i-AI ingasiza ekutholeni lokho kungena nokuthola okudidayo okulimazayo ezinhlelweni zokusebenza kwezempi nezempi.

Isivinini esilindelekile nesiyingozi

Nokho, ukwenza izinqumo ngokushesha nokwenza ngokushesha kungase kungakusizi. ukuze kulawulwe ngempumelelo inkinga. Izinzuzo ze-intelligence yokwenziwa kanye nezinhlelo ezizimele enkundleni yempi zingase zingavumeli isikhathi sokuxoxisana, njengoba sazi emlandweni, okuvame ukuphumelela njengendlela yokuvimbela noma yokulawula inhlekelele. Empeleni, ukwehlisa ijubane, ukumisa isikhashana, kanye nesikhathi sokuxoxisana kungaba ukhiye wokunqoba, noma okungenani ukugwema inhlekelele, ikakhulukazi lapho izikhali zenuzi zisengozini.

Izinqumo mayelana nempi nokuthula azikwazi ukushiywa ekuhlaziyeni okubikezelwayo. Kunomehluko obalulekile endleleni idatha esetshenziselwa ngayo izinjongo zesayensi, ezomnotho, ezokuthutha kanye nezibikezelo. ukuziphatha komuntu.

Abanye bangase babone i-AI njengamandla enza buthaka ukuzwelana kwamasu futhi ngaleyo ndlela andise ubungozi bempi. Idatha eyonakele ngengozi noma ngamabomu ingabangela amasistimu e-AI enze izenzo ezingahlosiwe, njengokukhomba ngokungeyikho nokukhomba okuqondiwe okungalungile. Isivinini sesenzo esibekwe endabeni yokuthuthukiswa kwama-algorithms empi singasho ukukhula ngaphambi kwesikhathi noma okungadingekile okuvimbela ukuphathwa okunengqondo kwenhlekelele. Ngakolunye uhlangothi, ama-algorithms nawo ngeke alinde futhi achaze, ngoba nawo kulindeleke ukuthi asheshe.

Isici esiphazamisayo ukusebenza kwe-algorithms yobuhlakani bokwenziwa futhi eyethulwe yithi muva nje kwaMT. Ngisho nochwepheshe abazi kahle ukuthi i-AI iholela kanjani emiphumeleni esiyibona ekuphumeni.

Endabeni yama-algorithms empi, asikwazi ukukhokhela ukungazi okunjalo ngemvelo nokuthi "bacabanga" kanjani ngayo. Asifuni ukuvuka phakathi kwamabili ngenxa yokuqhuma kwezikhali zenuzi ngoba ubuhlakani bokwenziwa "obethu" noma "babo" bunqume ukuthi sekuyisikhathi sokuthi ekugcineni silungise igeyimu.