"Kasuwa tana hannun 'yan tsiraru"
A halin yanzu, kamfanonin da ke tuka motocin fasinja ta atomatik za a iya raba su da kusan kashi uku. Rukunin farko shine tsarin rufaffiyar madauki mai kama da Apple (NASDAQ: AAPL). Mabuɗin abubuwan kamar su guntu da algorithms an yi su da kansu. Tesla (NASDAQ: TSLA) yayi wannan. Wasu sabbin kamfanonin samar da makamashi kuma suna fatan za su fara aiki a hankali. wannan hanya. Kashi na biyu shi ne tsarin budaddiyar kwamfuta irin na Android. Wasu masana'antun suna yin dandamali masu wayo, wasu kuma suna yin motoci. Misali, Huawei da Baidu (NASDAQ: BIDU) suna da niyya a wannan fanni. Nau'i na uku shine robotics (taksi maras direba), kamar kamfanoni irin su Waymo.
Wannan labarin zai fi nazartar yuwuwar waɗannan hanyoyi guda uku ta fuskar fasaha da bunƙasa kasuwanci, tare da tattauna makomar wasu sabbin masana'antun motocin lantarki ko kamfanonin tuƙi masu cin gashin kansu. Kada ku raina fasaha. Don tuki mai cin gashin kansa, fasaha ita ce rayuwa, kuma babbar hanyar fasaha ita ce hanya mai dabara. Don haka wannan labarin kuma tattaunawa ce kan hanyoyi daban-daban na dabarun tuki masu cin gashin kansu.
A fagen ƙwararrun motoci, musamman a fannin tuƙi mai cin gashin kai, ɗaukar tsarin rufaffiyar madauki na Apple na iya sauƙaƙawa masana'antun su haɓaka aiki da haɓaka aiki. Amsa da sauri ga bukatun mabukaci.
Bari in fara magana game da wasan kwaikwayo. Aiki yana da mahimmanci don tuƙi mai cin gashin kansa. Seymour Cray, mahaifin manyan kwamfutoci, ya taɓa faɗi kalma mai ban sha'awa, "Kowa zai iya gina CPU mai sauri. Dabarar ita ce gina tsarin sauri".
Tare da gazawar sannu a hankali na Dokar Moore, ba zai yuwu a ƙara aikin kawai ta hanyar ƙara adadin transistor kowane yanki ɗaya ba. Kuma saboda iyakancewar yanki da amfani da makamashi, ma'aunin guntu shima yana iyakance. Tabbas, Tesla FSD HW3.0 na yanzu (FSD ana kiransa Cikakkiyar Tuƙi) tsari ne kawai na 14nm, kuma akwai sarari don haɓakawa.
A halin yanzu, yawancin kwakwalwan kwamfuta na dijital an ƙirƙira su ne bisa ga Von Neumann Architecture tare da rabuwar ƙwaƙwalwar ajiya da kalkuleta, wanda ke haifar da dukkan tsarin kwamfutoci (ciki har da wayoyi masu wayo). Daga software zuwa tsarin aiki zuwa kwakwalwan kwamfuta, yana da matukar tasiri. Koyaya, Von Neumann Architecture bai dace da zurfin koyo wanda tuƙi mai cin gashin kansa ya dogara da shi ba, kuma yana buƙatar haɓakawa ko ma ci gaba.
Misali, akwai “bangon ƙwaƙwalwar ajiya” inda na’ura mai ƙididdigewa ke aiki da sauri fiye da ƙwaƙwalwar ajiya, wanda ke haifar da matsalolin aiki. Zane-zanen kwakwalwan kwamfuta-kamar kwakwalwa yana da ci gaba a cikin gine-gine, amma tsallen da ya yi nisa ba za a yi amfani da shi nan da nan ba. Bugu da ƙari, hanyar sadarwar juzu'in hoton za a iya jujjuya shi zuwa ayyukan matrix, wanda ƙila ba zai dace da kwakwalwan kwamfuta-kamar kwakwalwa ba.
Don haka, kamar yadda Dokar Moore da tsarin gine-ginen Von Neumann suka haɗu da ƙugiya, haɓaka aikin gaba ya fi buƙatu a samu ta hanyar Tsarin Tsare-tsare na Domain Specific (DSA, wanda zai iya komawa ga masu sarrafawa). Wadanda suka ci kyautar Turing Award John Hennessy da David Patterson ne suka gabatar da DSA. Bidi'a ce da ba ta da nisa sosai, kuma ra'ayi ne da za a iya aiwatar da shi nan take.
Zamu iya fahimtar ra'ayin DSA daga hangen nesa. Gabaɗaya, manyan kwakwalwan kwamfuta na yanzu suna da biliyoyin zuwa dubun biliyoyin transistor. Yadda waɗannan ɗimbin lambobi na transistor ke rarrabawa, haɗin kai, da haɗuwa suna da babban tasiri akan aikin takamaiman aikace-aikacen. A nan gaba, ya zama dole a gina "tsari mai sauri" daga mahangar software da hardware gabaɗaya, kuma a dogara ga ingantawa da daidaita tsarin.
"Yanayin Android" ba shine mafita mai kyau ba a fagen manyan motoci.
Mutane da yawa sun yi imanin cewa a zamanin tuƙi mai cin gashin kai, akwai kuma Apple (closed loop) da Android (buɗe) a fannin wayowin komai da ruwan ka, haka nan za a sami manyan masu samar da software kamar Google. Amsata mai sauki ce. Hanyar Android ba za ta yi aiki akan tuƙi mai cin gashin kai ba saboda bai dace da alkiblar haɓaka fasahar mota mai kaifin baki ba.
"Yanayin Android" ba shine mafita mai kyau ba a fagen manyan motoci.
Mutane da yawa sun yi imanin cewa a zamanin tuƙi mai cin gashin kai, akwai kuma Apple (closed loop) da Android (buɗe) a fannin wayowin komai da ruwan ka, haka nan za a sami manyan masu samar da software kamar Google. Amsata mai sauki ce. Hanyar Android ba za ta yi aiki akan tuƙi mai cin gashin kanta ba saboda bai dace da tsarin gine-ginen wayoyin hannu da na motoci masu wayo ya bambanta ba. Mayar da hankali ga wayoyin hannu shine ilimin halittu. Ecosystem yana nufin samar da aikace-aikace daban-daban dangane da tsarin ARM da IOS ko Android. Don haka, ana iya fahimtar wayoyi masu wayo na Android azaman haɗin gungun ma'auni na gama gari. Ma'auni na guntu shine ARM, a saman guntu akwai tsarin Android, sannan akwai apps daban-daban akan Intanet. Saboda daidaitawar sa, ko guntu ne, tsarin Android, ko App, yana iya zama kasuwanci cikin sauƙi.
Mayar da hankali na motoci masu wayo shine algorithm da bayanai da kayan aikin da ke tallafawa algorithm. Algorithm ɗin yana buƙatar babban aiki sosai ko an horar da shi a cikin gajimare ko kuma an ƙirƙira shi akan tasha. Kayan aikin mota mai wayo yana buƙatar haɓaka aiki da yawa don ƙayyadaddun aikace-aikace na musamman da algorithms. Don haka, algorithms ko guntu kawai ko tsarin aiki kawai za su fuskanci matsalolin haɓaka aiki a cikin dogon lokaci. Lokacin da kowane bangare ya haɓaka da kansa kawai za'a iya inganta shi cikin sauƙi. Rarraba software da hardware zai haifar da aikin da ba za a iya inganta shi ba.
Za mu iya kwatanta shi ta wannan hanya, NVIDIA Xavier yana da transistor biliyan 9, Tesla FSD HW 3.0 yana da transistor biliyan 6, amma ma'aunin wutar lantarki na Xavier bai kai HW3.0 ba. Kuma an ce FSD HW na gaba yana da ingantaccen aikin sau 7 idan aka kwatanta da na yanzu. Don haka, saboda mai zanen guntu na Tesla Peter Bannon da tawagarsa sun fi masu zanen NVIDIA ƙarfi, ko kuma saboda hanyar Tesla na haɗa software da hardware ya fi kyau. Muna tsammanin hanyoyin haɗa software da hardware dole ne su zama muhimmin dalili na haɓaka aikin guntu. Rarraba algorithms da bayanai ba kyakkyawan ra'ayi bane. Ba shi da amfani ga saurin amsawa kan buƙatun mabukaci da saurin haɓakawa.
Saboda haka, a fagen tuƙi mai cin gashin kansa, haɗa algorithms ko chips da sayar da su daban ba kasuwanci mai kyau ba ne a cikin dogon lokaci.