ʻO ka hopena hoʻomāinoino o ka hoʻokele autonomous: Tesla, Huawei, Apple, Weilai Xiaopeng, Baidu, Didi, ʻo wai ka mea hiki ke lilo i footnote o ka mōʻaukala?

"Aia ka mākeke i ka lima o ka poʻe liʻiliʻi"

I kēia manawa, hiki ke hoʻokaʻawale ʻia nā hui e hoʻokele kaʻa kaʻa i ʻekolu mau ʻāpana. ʻO ka māhele mua he ʻōnaehana pani pani e like me Apple (NASDAQ: AAPL). ʻO nā mea nui e like me nā chips a me nā algorithm i hana ʻia e lākou iho. ʻO Tesla (NASDAQ: TSLA) e hana i kēia. Manaʻo kekahi mau hui kaʻa ikehu hou e hoʻomaka mālie. keia alanui. ʻO ka ʻāpana ʻelua he ʻōnaehana wehe like me ka Android. Hana kekahi mau mea hana i nā paepae akamai, a hana kekahi i nā kaʻa. No ka laʻana, he manaʻo ko Huawei a me Baidu (NASDAQ: BIDU) i kēia mea. ʻO ke kolu o ka ʻāpana he robotics (kaʻa kaʻa ʻole), e like me nā hui e like me Waymo.

kiʻi-mai-PEXELS1

E kālele ana kēia ʻatikala i ka hiki ʻana o kēia mau ala ʻekolu mai ka ʻike o ka ʻenehana a me ka hoʻomohala ʻana i ka ʻoihana, a kūkākūkā i ka wā e hiki mai ana o kekahi mau mea hana kaʻa mana hou a i ʻole nā ​​​​hui hoʻokele autonomous. Mai hoʻohaʻahaʻa i ka ʻenehana. No ka hoʻokele kaʻa kaʻawale, ʻo ka ʻenehana ke ola, a ʻo ke ala ʻenehana nui ke ala hoʻolālā. No laila, he kūkākūkā pū kēia ʻatikala ma nā ala like ʻole o nā hoʻolālā hoʻokele autonomous.

Ua hiki mai ka wā o ka hoʻohui ʻana i nā lako polokalamu a me nā lako. ʻO ka "Apple model" i hōʻike ʻia e Tesla ke ala maikaʻi loa.

Ma ke kahua o nā kaʻa akamai, ʻoi aku hoʻi ma ke kahua o ka hoʻokele autonomous, ʻo ka hoʻohana ʻana i ke kumu hoʻohālike pani pani a Apple hiki ke maʻalahi i nā mea hana e hoʻomaikaʻi i ka hana. E pane koke i nā mea kūʻai aku.
E kamaʻilio mua wau e pili ana i ka hana. Pono ka hana no ka hoʻokele kaʻawale. ʻO Seymour Cray, ka makuakāne o nā supercomputers, i kekahi manawa i'ōlelo i kahi hua'ōlelo hoihoi loa, "E hiki i kekahi ke kūkulu i ka CPU wikiwiki.
Me ka hāʻule ʻana o ke kānāwai o Moore, ʻaʻole hiki ke hoʻonui wale i ka hana ma ka hoʻonui ʻana i ka helu o nā transistors no kēlā me kēia ʻāpana. A ma muli o ka palena o ka ʻāpana a me ka hoʻohana ʻana i ka ikehu, ua kaupalena ʻia ka nui o ka chip. ʻOiaʻiʻo, ʻo ka Tesla FSD HW3.0 o kēia manawa (FSD i kapa ʻia ʻo Full Self-Driving) he hana 14nm wale nō, a aia kahi wahi no ka hoʻomaikaʻi.
I kēia manawa, ua hoʻolālā ʻia ka hapa nui o nā pahu kikohoʻe ma muli o ka Von Neumann Architecture me ka hoʻokaʻawale ʻana o ka hoʻomanaʻo a me ka helu helu, e hana ana i ka ʻōnaehana holoʻokoʻa o nā kamepiula (me nā kelepona akamai). Mai ka polokalamu a hiki i nā ʻōnaehana hana a hiki i nā chips, pili nui ia. Eia naʻe, ʻaʻole kūpono loa ka Von Neumann Architecture no ke aʻo hohonu e hilinaʻi nei ka hoʻokele autonomous, a pono ka hoʻomaikaʻi ʻana a i ʻole ka holomua.
No ka laʻana, aia kahi "pā hoʻomanaʻo" kahi e holo wikiwiki ai ka helu helu ma mua o ka hoʻomanaʻo, hiki ke hoʻopilikia i ka hana. Loaʻa i ka hoʻolālā ʻana o nā ʻāpana lolo e like me ka lolo i ka hoʻolālā, akā ʻaʻole hiki ke hoʻohana koke ʻia ka lele mamao loa. Eia kekahi, hiki ke hoʻololi ʻia ke kiʻi convolutional network i loko o nā hana matrix, ʻaʻole kūpono paha ia no nā ʻāpana lolo.
No laila, ʻoiai ʻo Moore's Law a me ka hoʻolālā Von Neumann e hui pū ana i nā bottlenecks, pono e hoʻokō ʻia ka hoʻonui ʻana i ka wā e hiki mai ana ma o Domain Specific Architecture (DSA, hiki ke kuhikuhi i nā mea hana hoʻolaʻa). Ua noi ʻia ʻo DSA e nā mea lanakila o Turing Award ʻo John Hennessy lāua ʻo David Patterson. He hana hou ʻaʻole i mamao loa i mua, a he manaʻo hiki ke hoʻokō koke ʻia.
Hiki iā mākou ke hoʻomaopopo i ka manaʻo o DSA mai kahi hiʻohiʻona macro. ʻO ka mea maʻamau, ʻo nā ʻāpana kiʻekiʻe o kēia manawa he mau piliona a ʻumi mau piliona o nā transistors. Pehea ka puʻunaue ʻia o kēia mau helu nui o nā transistors, pili, a hui pū ʻia i ka hopena nui i ka hana o kahi noi kikoʻī. I ka wā e hiki mai ana, pono e kūkulu i kahi "ʻōnaehana wikiwiki" mai ka manaʻo holoʻokoʻa o ka lako polokalamu a me ka lako, a hilinaʻi i ka optimization a me ka hoʻoponopono ʻana o ka hale.

acasv (3)

ʻAʻole maikaʻi ka "mode Android" ma ke kahua o nā kaʻa akamai.

Manaʻo ka poʻe he nui i ke au o ka hoʻokele autonomous, aia nō ʻo Apple (paʻa loop) a me Android (open) ma ke kahua o nā kelepona akamai, a e loaʻa nō hoʻi nā mea lako polokalamu kaumaha e like me Google. He maʻalahi kaʻu pane. ʻAʻole e hana ka ala Android ma ke kaʻa kaʻa ʻokoʻa no ka mea ʻaʻole ia e hoʻokō i ke kuhikuhi o ka hoʻomohala ʻenehana kaʻa akamai e hiki mai ana.

ʻAʻole maikaʻi ka "mode Android" ma ke kahua o nā kaʻa akamai.

Manaʻo ka poʻe he nui i ke au o ka hoʻokele autonomous, aia nō ʻo Apple (paʻa loop) a me Android (open) ma ke kahua o nā kelepona akamai, a e loaʻa nō hoʻi nā mea lako polokalamu kaumaha e like me Google. He maʻalahi kaʻu pane. ʻAʻole e hana ka ala Android ma ka hoʻokele kaʻawale no ka mea ʻaʻole ia e hoʻokō i ka ʻO ka hoʻolālā o nā kelepona akamai a me nā kaʻa akamai he ʻokoʻa. ʻO ka manaʻo o nā smartphones ʻo ka ékology. ʻO ka Ecosystem ka hāʻawi ʻana i nā noi like ʻole e pili ana i ka ARM a me IOS a i ʻole nā ​​ʻōnaehana hana Android. No laila, hiki ke hoʻomaopopo ʻia nā kelepona akamai Android ma ke ʻano he hui pū ʻana o nā ʻāpana maʻamau maʻamau. ʻO ka maʻamau chip ka ARM, ma luna o ka chip ka ʻōnaehana hana Android, a laila aia nā polokalamu like ʻole ma ka Pūnaewele. Ma muli o kona hoʻohālikelike ʻana, inā he chip, ʻōnaehana Android, a i ʻole App, hiki iā ia ke lilo i ʻoihana kūʻokoʻa.

cx
acasv (1)

ʻO ka manaʻo o nā kaʻa akamai ka algorithm a me ka ʻikepili a me nā lako e kākoʻo ana i ka algorithm. Pono ka algorithm i ka hana kiʻekiʻe loa inā i aʻo ʻia i ke ao a i ʻole i hoʻopaʻa ʻia ma ka terminal. Pono ka lako o ke kaʻa akamai i ka hoʻonui ʻana i ka hana no nā noi kūikawā a me nā algorithms. No laila, ʻo nā algorithms a i ʻole nā ​​ʻāpana wale nō a i ʻole nā ​​ʻōnaehana hana wale nō e alo i nā dilemmas hoʻokō hana i ka wā lōʻihi. Ke hoʻomohala wale ʻia kēlā me kēia ʻāpana iā ia iho hiki ke maʻalahi i ka optimized. ʻO ka hoʻokaʻawale ʻana o nā polokalamu a me nā lako e hopena i ka hana ʻaʻole hiki ke hoʻonui ʻia.

Hiki iā mākou ke hoʻohālikelike i kēia ʻano, aia ʻo NVIDIA Xavier i 9 billion transistors, Tesla FSD HW 3.0 he 6 billion transistors, akā ʻaʻole maikaʻi ka helu helu mana o Xavier e like me HW3.0. A ua ʻōlelo ʻia ʻo ka hanauna hou FSD HW ka hoʻomaikaʻi ʻana i ka hana o 7 mau manawa i hoʻohālikelike ʻia me ka mea i kēia manawa. No laila, no ka mea ʻoi aku ka ikaika o ka mea hoʻolālā chip Tesla ʻo Peter Bannon a me kāna hui ma mua o nā mea hoʻolālā NVIDIA, a i ʻole no ka mea ʻoi aku ka maikaʻi o ke ʻano o Tesla no ka hoʻohui ʻana i nā polokalamu a me nā lako. Manaʻo mākou he kumu nui ke ʻano o ka hoʻohui ʻana i nā polokalamu a me nā lako hana no ka hoʻomaikaʻi ʻana i ka hana chip. ʻAʻole maikaʻi ka hoʻokaʻawale ʻana i nā algorithm a me nā ʻikepili. ʻAʻole kūpono ia i ka manaʻo wikiwiki e pili ana i nā pono o nā mea kūʻai aku a me ka wikiwiki wikiwiki.

No laila, ma ke kahua o ka hoʻokele autonomous, ʻo ka wehe ʻana i nā algorithm a i ʻole nā ​​ʻāpana a kūʻai aku iā lākou ma kahi kaʻawale ʻaʻole ia he ʻoihana maikaʻi i ka wā lōʻihi.

Dec-10-2020