{"id":4803,"date":"2016-03-08T22:45:33","date_gmt":"2016-03-08T15:45:33","guid":{"rendered":"http:\/\/ttud.com.vn\/?p=4803"},"modified":"2017-10-02T11:53:58","modified_gmt":"2017-10-02T04:53:58","slug":"tri-tue-nhan-tao-deep-learning-machine-learning-la-gi-chung-hoat-dong-ra-sao-va-co-han-che-nao","status":"publish","type":"post","link":"https:\/\/casta.ttud.com.vn\/index.php\/tri-tue-nhan-tao-deep-learning-machine-learning-la-gi-chung-hoat-dong-ra-sao-va-co-han-che-nao\/","title":{"rendered":"Tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o, deep learning, machine learning l\u00e0 g\u00ec, ch\u00fang ho\u1ea1t \u0111\u1ed9ng ra sao v\u00e0 c\u00f3 h\u1ea1n ch\u1ebf n\u00e0o?"},"content":{"rendered":"

Tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o ( artificial intelligence – AI) gi\u1edd xu\u1ea5t hi\u1ec7n \u1edf kh\u1eafp m\u1ecdi n\u01a1i. N\u00f3 l\u00e0 th\u1ee9 \u0111\u01b0\u1ee3c s\u1eed d\u1ee5ng \u0111\u1ec3 tr\u1ea3 l\u1eddi email t\u1ef1 \u0111\u1ed9ng tr\u00ean Gmail, h\u1ecdc c\u00e1ch l\u00e1i xe cho ch\u00fang ta ng\u1ed3i ch\u01a1i, s\u1eafp x\u1ebfp l\u1ea1i \u1ea3nh c\u1ee7a nh\u1eefng chuy\u1ebfn \u0111i ch\u01a1i th\u00e0nh t\u1eebng album ri\u00eang bi\u1ec7t, th\u1eadm ch\u00ed c\u00f2n gi\u00fap qu\u1ea3n l\u00fd ng\u00f4i nh\u00e0 hay \u0111i mua s\u1eafm n\u1eefa. Nh\u01b0ng b\u1ea1n c\u00f3 bi\u1ebft l\u00e0 tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o kh\u00f4ng ch\u1ec9 \u0111\u01a1n gi\u1ea3n l\u00e0 m\u1ed9t th\u1ef1c th\u1ec3 m\u00e0 n\u00f3 c\u00f2n \u0111\u01b0\u1ee3c chia th\u00e0nh nhi\u1ec1u lo\u1ea1i nh\u1ecf h\u01a1n? Nh\u1eefng h\u1ea1n ch\u1ebf hi\u1ec7n t\u1ea1i c\u1ee7a s\u1ea3n ph\u1ea9m tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o l\u00e0 g\u00ec? V\u00e0 v\u00ec sao ch\u00fang ta kh\u00f4ng c\u1ea7n (hay ch\u01b0a c\u1ea7n) ph\u1ea3i lo l\u1eafng v\u1ec1 vi\u1ec7c tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o b\u00f9ng l\u00ean chi\u1ebfm l\u1ea5y th\u1ebf gi\u1edbi?<\/strong><\/p>\n

\"\"<\/p>\n

Neural network, machine learning, deep learning l\u00e0 g\u00ec?<\/p>\n

\u0110\u00e2y l\u00e0 nh\u1eefng c\u1ee5m t\u1eeb m\u00e0 b\u1ea1n r\u1ea5t th\u01b0\u1eddng hay th\u1ea5y trong nh\u1eefng th\u00f4ng tin li\u00ean quan \u0111\u1ebfn tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o. V\u1ec1 c\u01a1 b\u1ea3n, b\u1ea1n c\u00f3 th\u1ec3 ngh\u0129 v\u1ec1 nh\u1eefng th\u1ee9 n\u00e0y nh\u01b0 c\u00e1c l\u1edbp c\u1ea5u th\u00e0nh tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o.<\/p>\n

Neural network, t\u1ea1m g\u1ecdi l\u00e0 m\u1ea1ng l\u01b0\u1edbi th\u1ea7n kinh nh\u00e2n t\u1ea1o, s\u1ebd n\u1eb1m d\u01b0\u1edbi c\u00f9ng. \u0110\u00e2y l\u00e0 h\u1ec7 th\u1ed1ng c\u00e1c m\u00e1y t\u00ednh v\u00e0 thi\u1ebft b\u1ecb h\u1ec7 th\u1ed1ng \u0111i\u1ec7n to\u00e1n n\u00f3i chung \u0111\u01b0\u1ee3c k\u1ebft n\u1ed1i theo m\u1ed9t c\u00e1ch n\u00e0o \u0111\u00f3 \u0111\u1ec3 m\u00f4 ph\u1ecfng l\u1ea1i m\u1ed9t ph\u1ea7n c\u00e1ch ho\u1ea1t \u0111\u1ed9ng c\u1ee7a c\u00e1c n\u01a1-ron th\u1ea7n kinh trong n\u00e3o ng\u01b0\u1eddi. Nh\u1eefng chi\u1ebfc m\u00e1y t\u00ednh trong neural network c\u00f3 th\u1ec3 n\u1eb1m g\u1ea7n nhau trong c\u00f9ng m\u1ed9t ph\u00f2ng hay c\u00e1ch xa nhau c\u1ea3 ngh\u00ecn c\u00e2y s\u1ed1, m\u1ed7i m\u1ed9t chi\u1ebfc m\u00e1y t\u00ednh trong \u0111\u00f3 c\u00f3 th\u1ec3 \u0111\u01b0\u1ee3c xem nh\u01b0 m\u1ed9t \u0111\u01a1n v\u1ecb th\u1ea7n kinh, g\u1ecdi l\u00e0 node.<\/p>\n

C\u1eadp nh\u1eadt: Neural network kh\u00f4ng nh\u1ea5t thi\u1ebft ph\u1ea3i d\u1ef1a tr\u00ean ph\u1ea7n c\u1ee9ng, n\u00f3 v\u1eabn c\u00f3 th\u1ec3 l\u00e0 ph\u1ea7n m\u1ec1m v\u00e0 c\u00e1c gi\u1ea3i thu\u1eadt.<\/p>\n

Kh\u00e1i ni\u1ec7m v\u1ec1 neural network \u0111\u00e3 c\u00f3 t\u1eeb nh\u1eefng n\u0103m 1950 v\u1edbi s\u1ef1 ra \u0111\u1eddi c\u1ee7a ng\u00e0nh nghi\u00ean c\u1ee9u v\u1ec1 tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o. Ng\u01b0\u1eddi ta n\u00f3i r\u1eb1ng khi n\u1eb1m ri\u00eang bi\u1ec7t, nh\u1eefng node m\u00e1y t\u00ednh n\u00e0y ch\u1ec9 ch\u1ea1y nh\u1eefng g\u00ec \u0111\u01b0\u1ee3c l\u1eadp tr\u00ecnh s\u1eb5n v\u00e0 ch\u1ec9 c\u00f3 th\u1ec3 tr\u1ea3 l\u1eddi nh\u1eefng c\u00e2u h\u1ecfi \u0111\u01a1n gi\u1ea3n, hay n\u00f3i c\u00e1ch l\u00e0 n\u00f3 “kh\u00f4ng th\u00f4ng minh”. C\u0169ng gi\u1ed1ng nh\u01b0 trong c\u01a1 th\u1ec3 ng\u01b0\u1eddi, m\u1ed9t n\u01a1-ron kh\u00f4ng ch\u01b0a l\u00e0m n\u00ean chuy\u1ec7n, nh\u01b0ng khi k\u1ebft n\u1ed1i ch\u00fang th\u00e0nh m\u1ed9t m\u1ea1ng l\u01b0\u1edbi d\u00e0y \u0111\u1eb7t th\u00ec m\u1ecdi chuy\u1ec7n s\u1ebd kh\u00e1c \u0111i r\u1ea5t nhi\u1ec1u. Khi c\u00e1c h\u1ec7 th\u1ed1ng m\u00e1y t\u00ednh \u0111\u01b0\u1ee3c n\u1ed1i l\u1ea1i v\u1edbi nhau, ch\u00fang c\u00f3 th\u1ec3 gi\u1ea3i quy\u1ebft nh\u1eefng v\u1ea5n \u0111\u1ec1 kh\u00f3 kh\u0103n h\u01a1n. V\u00e0 quan tr\u1ecdng nh\u1ea5t, khi \u00e1p d\u1ee5ng \u0111\u00fang thu\u1eadt to\u00e1n, ng\u01b0\u1eddi ta c\u00f3 th\u1ec3 “d\u1ea1y h\u1ecdc” cho m\u00e1y t\u00ednh.<\/p>\n

L\u1edbp k\u1ebf ti\u1ebfp l\u00e0 machine learning. T\u1eeb n\u00e0y m\u00ecnh kh\u00f4ng bi\u1ebft n\u00ean d\u1ecbch sang ti\u1ebfng Vi\u1ec7t nh\u01b0 th\u1ebf n\u00e0o cho ch\u00ednh x\u00e1c n\u00ean t\u1ea1m th\u1eddi c\u1ee9 \u0111\u1ec3 nguy\u00ean ti\u1ebfng Anh. \u0110\u00e2y l\u00e0 m\u1ed9t ch\u01b0\u01a1ng tr\u00ecnh ch\u1ea1y tr\u00ean neural network, n\u00f3 s\u1ebd l\u00e0m nhi\u1ec7m v\u1ee5 hu\u1ea5n luy\u1ec7n m\u00e1y t\u00ednh \u0111\u1ec3 “h\u1ecdc h\u1ecfi” m\u1ed9t v\u1ea5n \u0111\u1ec1 g\u00ec \u0111\u00f3, v\u00ed d\u1ee5 nh\u01b0 h\u1ecdc h\u1ecfi t\u1eeb h\u00e0ng nh\u00ecn m\u1eabu ch\u1eef vi\u1ebft tay c\u1ee7a ng\u01b0\u1eddi d\u00f9ng \u0111\u1ec3 \u0111o\u00e1n xem \u0111\u00f3 l\u00e0 k\u00ed t\u1ef1 g\u00ec, ho\u1eb7c h\u1ecdc t\u1eeb h\u00e0ng ngh\u00ecn b\u1ee9c \u1ea3nh ch\u1ee5p b\u00e3i bi\u1ec3n \u0111\u1ec3 t\u00ecm ra \u0111i\u1ec3m chung v\u00e0 sau \u0111\u00f3 nh\u00ecn ph\u00e1t l\u00e0 bi\u1ebft ngay t\u1ea5m n\u00e0y ch\u1ee5p bi\u1ec3n ch\u1ee9 kh\u00f4ng ph\u1ea3i ch\u1ee5p n\u00fai.<\/p>\n

Deep learning n\u1eb1m \u1edf tr\u00ean c\u00f9ng, c\u0169ng kh\u00f4ng bi\u1ebft d\u1ecbch ra ti\u1ebfng Vi\u1ec7t nh\u01b0 th\u1ebf n\u00e0o. \u0110\u00e2y l\u00e0 m\u1ed9t nh\u00e1nh \u0111\u1eb7c bi\u1ec7t c\u1ee7a ng\u00e0nh khoa h\u1ecdc machine learning. Deep learning tr\u1edf n\u00ean ph\u1ed5 bi\u1ebfn trong th\u1eadp k\u1ec9 g\u1ea7n \u0111\u00e2y nh\u1edd v\u00e0o s\u1ef1 gia t\u0103ng nhanh ch\u00f3ng c\u1ee7a l\u01b0\u1ee3ng d\u1eef li\u1ec7u s\u1ed1 m\u00e0 lo\u00e0i ng\u01b0\u1eddi t\u1ea1o ra, ngo\u00e0i c\u00f2n nh\u1edd s\u1ee9c m\u1ea1nh x\u1eed l\u00fd c\u1ee7a m\u00e1y t\u00ednh gia t\u0103ng trong khi gi\u00e1 th\u00e0nh gi\u1ea3m xu\u1ed1ng. S\u1ebd n\u00f3i k\u0129 h\u01a1n v\u1ec1 deep learning \u1edf b\u00ean d\u01b0\u1edbi.<\/p>\n

Tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o ho\u1ea1t \u0111\u1ed9ng ra sao?<\/p>\n

Gi\u1ea3 s\u1eed nh\u01b0 b\u1ea1n mu\u1ed1n m\u1ed9t c\u00e1i m\u00e1y t\u00ednh bi\u1ebft c\u00e1ch b\u0103ng qua \u0111\u01b0\u1eddng. Theo c\u00e1ch truy\u1ec1n th\u1ed1ng, ch\u00fang ta s\u1ebd l\u1eadp tr\u00ecnh cho n\u00f3 c\u00e1ch nh\u00ecn tr\u00e1i, nh\u00ecn ph\u1ea3i, c\u00e1ch \u0111\u1ee3i xe ch\u1ea1y qua h\u1ebft, c\u00e1ch \u0111i \u0111\u00fang v\u1ea1ch b\u0103ng \u0111\u01b0\u1eddng theo lu\u1eadt v\u00e0 nhi\u1ec1u th\u1ee9 kh\u00e1c, sau \u0111\u00f3 \u0111\u1ec3 cho m\u00e1y t\u1ef1 \u0111i.<\/p>\n

C\u00f2n v\u1edbi k\u0129 thu\u1eadt tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o, m\u00e0 c\u1ee5 th\u1ec3 l\u00e0 m\u1ed9t ch\u01b0\u01a1ng tr\u00ecnh machine learning, b\u1ea1n s\u1ebd cho m\u00e1y t\u00ednh xem 10.000 \u0111o\u1ea1n video v\u1ec1 c\u00e1ch b\u0103ng \u0111\u01b0\u1eddng an to\u00e0n. K\u1ebf ti\u1ebfp b\u1ea1n l\u1ea1i cho n\u00f3 xem th\u00eam 10.000 \u0111o\u1ea1n video n\u1eefa nh\u01b0ng l\u1ea7n n\u00e0y chi\u1ebfu c\u1ea3nh ng\u01b0\u1eddi ta b\u1ecb xe t\u00f4ng khi b\u0103ng \u0111\u01b0\u1eddng \u1ea9u. L\u00fac n\u00e0y b\u1ea1n m\u1edbi th\u1ea3 cho n\u00f3 t\u1ef1 b\u0103ng \u0111\u01b0\u1eddng.<\/p>\n

google_self_driving_car.0. \u200b<\/p>\n

Ph\u1ea7n kh\u00f3 nh\u1ea5t \u0111\u00f3 l\u00e0 b\u1ea1n ph\u1ea3i l\u00e0m sao cho m\u00e1y t\u00ednh hi\u1ec3u v\u00e0 ti\u1ebfp thu \u0111\u01b0\u1ee3c th\u00f4ng tin t\u1eeb c\u00e1c video n\u00e0y, c\u0169ng gi\u1ed1ng nh\u01b0 ph\u1ea7n kh\u00f3 nh\u1ea5t khi \u0111i d\u1ea1y h\u1ecdc l\u00e0 ph\u1ea3i l\u00e0m cho h\u1ecdc sinh hi\u1ec3u \u0111\u01b0\u1ee3c b\u1ea1n n\u00f3i g\u00ec v\u00e0 ghi nh\u1edb \u0111i\u1ec1u \u0111\u00f3. Trong nhi\u1ec1u th\u1eadp k\u1ec9 qua, ng\u01b0\u1eddi ta \u0111\u00e3 th\u1ee7 nhi\u1ec1u ph\u01b0\u01a1ng ph\u00e1p kh\u00e1c nhau \u0111\u1ec3 d\u1ea1y cho m\u00e1y t\u00ednh h\u1ecdc. M\u1ed9t trong nh\u1eefng c\u00e1ch \u0111\u00e3 \u0111\u01b0\u1ee3c x\u00e0i \u0111\u00f3 l\u00e0 “reinforcement learning”, t\u1ee9c l\u00e0 b\u1ea1n s\u1ebd “th\u01b0\u1edfng” cho m\u00e1y t\u00ednh khi n\u00f3 l\u00e0m \u0111\u00fang th\u1ee9 b\u1ea1n mu\u1ed1n r\u1ed3i t\u1eeb t\u1eeb t\u1ed1i \u01b0u \u0111\u1ec3 \u0111\u1ea1t k\u1ebft qu\u1ea3 t\u1ed1t nh\u1ea5t. Ng\u01b0\u1eddi ta v\u1eabn th\u01b0\u1eddng hay hu\u1ea5n luy\u1ec7n th\u00fa v\u1eadt theo c\u00e1ch n\u00e0y. M\u1ed9t c\u00e1ch kh\u00e1c n\u1eefa l\u00e0 “ch\u1ecdn l\u1ecdc t\u1ef1 nhi\u00ean”, t\u1ee9c l\u00e0 nhi\u1ec1u c\u00e1ch gi\u1ea3i quy\u1ebft c\u00f9ng m\u1ed9t v\u1ea5n \u0111\u1ec1 s\u1ebd \u0111\u01b0\u1ee3c \u00e1p d\u1ee5ng cho ch\u1ea1y song song, c\u00e1i n\u00e0y gi\u1ea3i \u0111\u01b0\u1ee3c nhanh v\u00e0 ch\u00ednh x\u00e1c nh\u1ea5t s\u1ebd th\u1eafng nh\u1eefng c\u00e1i c\u00f2n l\u1ea1i.<\/p>\n

C\u00f2n trong th\u1eddi \u0111\u1ea1i ng\u00e0y nay, ng\u01b0\u1eddi ta x\u00e0i m\u1ed9t ph\u01b0\u01a1ng ph\u00e1p g\u1ecdi l\u00e0 deep learning. Deep learning s\u1eed d\u1ee5ng nhi\u1ec1u l\u1edbp trong m\u1ed9t neural network \u0111\u1ec3 ph\u00e2n t\u00edch d\u1eef li\u1ec7u theo nhi\u1ec1u kh\u00eda c\u1ea1nh kh\u00e1c nhau. V\u00ed d\u1ee5, n\u1ebfu b\u1ea1n \u0111\u01b0a cho m\u00e1y t\u00ednh h\u1ecdc m\u1ed9t t\u1ea5m \u1ea3nh theo k\u0129 thu\u1eadt deep learning, m\u1ed7i l\u1edbp trong m\u1ea1ng l\u01b0\u1edbi th\u1ea7n kinh nh\u00e2n t\u1ea1o n\u00e0y s\u1ebd nh\u00ecn nh\u1eadn v\u1ea5n \u0111\u1ec1 theo m\u1ed9t c\u00e1ch ri\u00eang. L\u1edbp d\u01b0\u1edbi c\u00f9ng s\u1ebd ch\u1ec9 \u0111\u01a1n gi\u1ea3n l\u00e0 v\u1ebd ra m\u1ed9t c\u00e1i l\u01b0\u1edbi 5×5 l\u00ean t\u00e1m \u1ea3nh v\u00e0 \u0111\u00e1nh d\u1ea5u “c\u00f3” ho\u1eb7c “kh\u00f4ng” khi c\u00f3 m\u1ed9t \u0111\u1ed1i t\u01b0\u1ee3ng xu\u1ea5t hi\u1ec7n trong \u00f4. N\u1ebfu “c\u00f3”, l\u1edbp b\u00ean tr\u00ean s\u1ebd b\u1eaft \u0111\u1ea7u nh\u00ecn v\u00e0o t\u1eebng \u00f4 n\u00e0y m\u1ed9t c\u00e1ch k\u0129 c\u00e0ng h\u01a1n, n\u00f3 ph\u00e2n t\u00edch xem \u0111\u00e2y c\u00f3 ph\u1ea3i l\u00e0 \u0111i\u1ec3m \u0111\u1ea7u c\u1ee7a m\u1ed9t \u0111\u01b0\u1eddng th\u1eb3ng kh\u00f4ng, hay \u0111\u00e2y l\u00e0 m\u1ed9t c\u00e1i g\u00f3c nghi\u00eang? Th\u1eadt nhi\u1ec1u l\u1edbp nh\u01b0 th\u1ebf s\u1ebd gi\u00fap ph\u1ea7n m\u1ec1m hi\u1ec3u \u0111\u01b0\u1ee3c nh\u1eefng v\u1ea5n \u0111\u1ec1 ph\u1ee9c t\u1ea1p, t\u1ea5t c\u1ea3 \u0111\u1ec1u d\u1ef1a tr\u00ean c\u00e1c b\u1ebb nh\u1ecf n\u00f3 ra r\u1ed3i “\u0111i\u1ec1u tra” t\u1eeb t\u1eeb. C\u0169ng ch\u00ednh v\u00ec l\u00fd do n\u00e0y m\u00e0 ng\u01b0\u1eddi ta g\u1ecdi \u0111\u00e2y l\u00e0 “deep”, t\u1ee9c l\u00e0 s\u00e2u v\u00e0 c\u00f3 nhi\u1ec1u l\u1edbp.<\/p>\n

DeepFace.
\nK\u0129 thu\u1eadt \u0111\u01b0\u1ee3c Facebook \u00e1p d\u1ee5ng \u0111\u1ec3 nh\u1eadn di\u1ec7n g\u01b0\u01a1ng m\u1eb7t, trong \u0111\u00f3 chia b\u1ee9c \u1ea3nh th\u00e0nh l\u1edbp kh\u00e1c nhau \u0111\u1ec3 h\u1ecdc h\u1ecfi\u200b<\/p>\n

Yann LeCun, tr\u01b0\u1edfng b\u1ed9 ph\u1eadn tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o c\u1ee7a Facebook, cho bi\u1ebft: “Khi b\u1ea1n \u0111i l\u00ean c\u00e1c l\u1edbp cao h\u01a1n th\u00ec nh\u1eefng th\u1ee9 \u0111\u01b0\u1ee3c ph\u00e1t hi\u1ec7n s\u1ebd c\u00e0ng l\u00fac c\u00e0ng r\u1ed9ng h\u01a1n. C\u00e0ng l\u00fac c\u00e0ng c\u00f3 nhi\u1ec1u kh\u00eda c\u1ea1nh h\u01a1n \u0111\u01b0\u1ee3c ph\u00e2n t\u00edch. V\u00e0 khi b\u1ea1n leo l\u00ean \u0111\u1ebfn l\u1edbp cao nh\u1ea5t, b\u1ea1n s\u1ebd c\u00f3 nh\u1eefng c\u00f4ng c\u1ee5 \u0111\u1ec3 cho b\u1ea1n bi\u1ebft b\u1ee9c h\u00ecnh \u0111\u00f3 \u0111ang ch\u1ee5p ng\u01b0\u1eddi hay m\u1ed9t ch\u00fat ch\u00f3 hay m\u1ed9t chi\u1ebfc m\u00e1y bay”.<\/p>\n

N\u00e3y gi\u1edd ch\u00fang ta ch\u1ec9 m\u1edbi n\u00f3i \u0111\u1ebfn chuy\u1ec7n nh\u1eadn bi\u1ebft, gi\u1edd th\u00ec \u0111\u1ebfn l\u00fac d\u1ea1y cho m\u00e1y t\u00ednh bi\u1ebft th\u1ee9 m\u00e0 n\u00f3 v\u1eeba nh\u1eadn ra l\u00e0 g\u00ec. M\u1ed9t h\u1ec7 th\u1ed1ng neural network l\u1ea1i \u0111\u01b0\u1ee3c s\u1eed d\u1ee5ng, nh\u01b0ng l\u1ea7n n\u00e0y n\u00f3 s\u1ebd xem x\u00e9t nhi\u1ec1u \u0111\u1eb7c \u0111i\u1ec3m c\u1ee7a m\u1ed9t con m\u00e8o. R\u1ea5t nhi\u1ec1u b\u1ee9c h\u00ecnh ch\u1ee5p m\u00e8o c\u0169ng s\u1ebd \u0111\u01b0\u1ee3c \u0111\u01b0a cho h\u1ec7 th\u1ed1ng xem k\u00e8m theo l\u1eddi d\u1eb7n: \u0111\u00e2y l\u00e0 \u1ea3nh con m\u00e8o \u0111\u00f3 nha. R\u1ed3i ng\u01b0\u1eddi ta l\u1ea1i cho m\u00e1y xem th\u00eam m\u1ed9t lo\u1ea1t \u1ea3nh kh\u00e1c n\u1eefa ch\u1ee5p ch\u00f3, heo, g\u1ea5u, v\u1ecbt v\u00e0 d\u1eb7n: \u0111\u00e2y kh\u00f4ng ph\u1ea3i l\u00e0 con m\u00e8o. Th\u00f4ng qua h\u00e0ng lo\u1ea1t d\u1eef li\u1ec7u nh\u01b0 v\u1eady, ph\u1ea7n m\u1ec1m s\u1ebd bi\u1ebft \u0111\u01b0\u1ee3c nh\u1eefng con m\u00e8o th\u01b0\u1eddng c\u00f3 \u0111i\u1ec3m chung g\u00ec, m\u00f3ng, l\u00f4ng, tay ch\u00e2n, \u0111\u1ea7u, \u0111u\u00f4i c\u1ee7a ch\u00fang ra sao th\u00ec m\u1edbi \u0111\u01b0\u1ee3c g\u1ecdi l\u00e0 m\u00e8o…<\/p>\n

Theo th\u1eddi gian, m\u00e1y s\u1ebd ghi nh\u1edb nh\u1eefng d\u1eef li\u1ec7u n\u00e0y v\u00e0 s\u1eafp x\u1ebfp theo th\u1ee9 t\u1ef1 quan tr\u1ecdng. V\u00ed d\u1ee5, m\u00f3ng vu\u1ed1t kh\u00f4ng ch\u1ec9 m\u00e8o m\u1edbi c\u00f3, nh\u01b0ng n\u1ebfu m\u00f3ng \u0111i chung v\u1edbi b\u00e0n ch\u00e2n to v\u00e0 ria m\u00e9p th\u00ec \u0111\u00e2y \u0111\u00edch th\u1ecb l\u00e0 con m\u00e8o. C\u00e1c m\u1ed1i li\u00ean h\u1ec7 nh\u01b0 th\u1ebf n\u00e0y c\u0169ng s\u1ebd \u0111\u01b0\u1ee3c cung c\u1ea5p theo th\u1eddi gian trong qu\u00e1 tr\u00ecnh ph\u1ea7n m\u1ec1m machine learning h\u1ecdc h\u1ecfi c\u00e1c \u1ea3nh. Quy tr\u00ecnh n\u00e0y di\u1ec5n ra trong th\u1eddi gian d\u00e0i v\u00e0 l\u1eb7p \u0111i l\u1eb7p l\u1ea1i nhi\u1ec1u l\u1ea7n. C\u1ee9 l\u1ea7n sau n\u00f3 s\u1ebd t\u1ed1t h\u01a1n l\u1ea7n tr\u01b0\u1edbc v\u00ec \u0111\u01b0\u1ee3c g\u00f3p \u00fd t\u1eeb con ng\u01b0\u1eddi ho\u1eb7c th\u1eadm ch\u00ed l\u00e0 t\u1eeb nh\u1eefng h\u1ec7 th\u1ed1ng tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o kh\u00e1c.<\/p>\n

B\u1ea1n c\u00f3 th\u1ec3 th\u1ea5y r\u1eb1ng ch\u1ec9 \u0111\u1ec3 nh\u1eadn bi\u1ebft \u0111\u01b0\u1ee3c m\u1ed9t con m\u00e8o th\u00f4i \u0111\u00e3 ph\u1ee9c t\u1ea1p qu\u00e1 ch\u1eebng, trong khi c\u00e1c h\u1ec7 th\u1ed1ng machine learning c\u1ee7a Facebook, Google hay Microsoft ph\u1ea3i nh\u1eadn bi\u1ebft nhi\u1ec1u th\u1ee9 kh\u00e1c n\u1eefa trong \u0111\u1eddi s\u1ed1ng. Th\u1ebf n\u00ean, vi\u1ec7c Microsoft t\u1ef1 h\u00e0o khi ph\u00e1t h\u00e0nh m\u1ed9t \u1ee9ng d\u1ee5ng c\u00f3 kh\u1ea3 n\u0103ng nh\u1eadn bi\u1ebft c\u00e1c gi\u1ed1ng ch\u00f3 kh\u00f3 nhanh nghe c\u00f3 v\u1ebb \u0111\u01a1n gi\u1ea3n nh\u01b0ng ph\u00eda sau n\u00f3 l\u00e0 c\u1ea3 m\u1ed9t m\u1ea1ng l\u01b0\u1edbi th\u1ea7n kinh nh\u00e2n t\u1ea1o ph\u1ee9c t\u1ea1p v\u00e0 \u0111\u00e3 b\u1eaft \u0111\u1ea7u ch\u1ea1y h\u1ecdc h\u1ecfi trong th\u1eddi gian r\u1ea5t d\u00e0i r\u1ed3i.<\/p>\n

dogknight_3346130k.0.
\nH\u00ecnh \u1ea3nh \u0111\u01b0\u1ee3c t\u1ea1o ra t\u1eeb d\u1ef1 \u00e1n tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o Google Drea\u200b<\/p>\n

\u0110\u00e2y c\u00f3 ph\u1ea3i l\u00e0 th\u1ee9 m\u00e0 Google, Facebook v\u00e0 nhi\u1ec1u c\u00f4ng ty kh\u00e1c \u0111ang s\u1eed d\u1ee5ng?<\/p>\n

C\u00e2u tr\u1ea3 l\u1eddi chung s\u1ebd l\u00e0 “\u0110\u00fang r\u1ed3i”.<\/p>\n

Deep learning hi\u1ec7n t\u1ea1i \u0111ang \u0111\u01b0\u1ee3c x\u00e0i cho nhi\u1ec1u t\u00e1c v\u1ee5 kh\u00e1c nhau. C\u00e1c c\u00f4ng ty c\u00f4ng ngh\u1ec7 l\u1edbn th\u01b0\u1eddng l\u1eadp ri\u00eang m\u1ed9t b\u1ed9 ph\u1eadn tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o. Google, Facebook c\u00f2n m\u1edf nh\u1eefng h\u1ec7 th\u1ed1ng ph\u1ea7n m\u1ec1m machine learning c\u1ee7a m\u00ecnh ra cho m\u1ecdi ng\u01b0\u1eddi s\u1eed d\u1ee5ng. Google h\u1ed3i th\u00e1ng tr\u01b0\u1edbc c\u0169ng m\u1edbi m\u1edf m\u1ed9t kh\u00f3a h\u1ecdc k\u00e9o d\u00e0i 3 th\u00e1ng v\u00e8 machine learning v\u00e0 deep learning.<\/p>\n

M\u1ed9t v\u00e0i v\u00ed d\u1ee5 v\u1ec1 machine learning \u0111\u00f3 l\u00e0 c\u00f4ng c\u1ee5 Google Photos. N\u00f3 c\u00f3 kh\u1ea3 n\u0103ng nh\u1eadn bi\u1ebft v\u00e0 ph\u00e2n lo\u1ea1i h\u00ecnh \u1ea3nh m\u00e0 b\u1ea1n ch\u1ee5p theo t\u1eebng ch\u1ee7 \u0111\u1ec1 kh\u00e1c nhau, th\u1eadm ch\u00ed l\u00e0 theo t\u1eebng khu\u00f4n m\u1eb7t kh\u00e1c nhau m\u1ed9t c\u00e1ch t\u1ef1 \u0111\u1ed9ng. Hay Facebook M, m\u1ed9t tr\u1ee3 l\u00fd \u1ea3o c\u00e1 nh\u00e2n n\u1eeda ng\u01b0\u1eddi n\u1eeda m\u00e1y c\u00f3 th\u1ec3 gi\u00fap b\u1ea1n \u0111\u1eb7t m\u1ed9t s\u1ed1 m\u00f3n h\u00e0ng m\u00e0 b\u1ea1n mu\u1ed1n. Microsoft v\u1edbi Cortana, Google v\u1edbi Google Now v\u00e0 Apple v\u1edbi Siri \u0111\u1ec1u l\u00e0 nh\u1eefng v\u00ed d\u1ee5 r\u1ea5t th\u1ef1c c\u1ee7a tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o.<\/p>\n

\u0110\u00e2y c\u0169ng l\u00e0 l\u00fd do v\u00ec sao m\u00e0 trong th\u1eddi gian g\u1ea7n \u0111\u00e2y ch\u00fang ta b\u1eaft \u0111\u1ea7u nghe n\u00f3i nhi\u1ec1u v\u1ec1 machine learning, deep learning v\u00e0 tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o. \u0110\u00f3 l\u00e0 v\u00ec c\u00e1c h\u00e3ng ti\u00eau d\u00f9ng l\u1edbn \u0111\u00e3 b\u1eaft \u0111\u1ea7u nh\u1ea3y v\u00e0o cu\u1ed9c ch\u01a1i v\u00e0 \u0111\u01b0a ra nh\u1eefng s\u1ea3n ph\u1ea9m th\u1ef1c t\u1ebf m\u00e0 ng\u01b0\u1eddi ta c\u00f3 th\u1ec3 c\u1ea7m n\u1eafm v\u00e0 tr\u1ea3i nghi\u1ec7m. Tr\u01b0\u1edbc \u0111\u00f3, deep learning v\u00e0 machine learning ch\u1ec9 n\u1eb1m trong c\u00e1c ph\u00f2ng th\u00ed nghi\u1ec7m \u1edf nh\u1eefng vi\u1ec7n nghi\u00ean c\u1ee9u v\u00e0 tr\u01b0\u1eddng \u0111\u1ea1i h\u1ecdc m\u00e0 th\u00f4i.<\/p>\n

H\u1ea1n ch\u1ebf c\u1ee7a tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o hi\u1ec7n nay<\/p>\n

Deep learning \u0111ang \u0111\u01b0\u1ee3c s\u1eed d\u1ee5ng cho nh\u1eefng th\u1ee9 nh\u01b0 nh\u1eadn d\u1ea1ng gi\u1ecdng n\u00f3i v\u00e0 nh\u1eadn d\u1ea1ng h\u00ecnh \u1ea3nh, nh\u1eefng th\u1ee9 c\u00f3 nhi\u1ec1u ti\u1ec1m n\u0103ng th\u01b0\u01a1ng m\u1ea1i h\u00f3a. Nh\u01b0ng song song \u0111\u00f3, n\u00f3 c\u0169ng c\u00f2n nhi\u1ec1u h\u1ea1n ch\u1ebf.<\/p>\n

\u0110\u1ea7u ti\u00ean, deep learning c\u1ea7n m\u1ed9t l\u01b0\u1ee3ng d\u1eef li\u1ec7u \u0111\u1ea7u v\u00e0o kh\u1ed5ng l\u1ed3 \u0111\u1ec3 m\u00e1y t\u00ednh c\u00f3 th\u1ec3 h\u1ecdc h\u1ecfi. Quy tr\u00ecnh n\u00e0y m\u1ea5t nhi\u1ec1u th\u1eddi gian, nhi\u1ec1u s\u1ee9c m\u1ea1nh x\u1eed l\u00fd m\u00e0 ch\u1ec9 c\u00f3 c\u00e1c server c\u1ee1 l\u1edbn m\u1edbi l\u00e0m \u0111\u01b0\u1ee3c. N\u1ebfu kh\u00f4ng c\u00f3 \u0111\u1ee7 d\u1eef li\u1ec7u \u0111\u1ea7u v\u00e0o, hay c\u00f3 \u0111\u1ee7 d\u1eef li\u1ec7u nh\u01b0ng kh\u00f4ng \u0111\u1ee7 s\u1ee9c m\u1ea1nh \u0111\u1ec3 x\u1eed l\u00fd, th\u00ec m\u1ecdi th\u1ee9 kh\u00f4ng th\u1ec3 di\u1ec5n ra \u0111\u00fang nh\u01b0 \u00fd \u0111\u1ecbnh, k\u1ebft qu\u1ea3 do m\u00e1y h\u1ecdc s\u1ebd kh\u00f4ng ch\u00ednh x\u00e1c.<\/p>\n

Th\u1ee9 hai, deep learning v\u1eabn ch\u01b0a th\u1ec3 nh\u1eadn bi\u1ebft \u0111\u01b0\u1ee3c nh\u1eefng th\u1ee9 ph\u1ee9c t\u1ea1p, v\u00ed d\u1ee5 nh\u01b0 c\u00e1c m\u1ed1i li\u00ean h\u1ec7 th\u00f4ng th\u01b0\u1eddng. Ch\u00fang c\u0169ng s\u1ebd g\u1eb7p kh\u00f3 kh\u0103n khi c\u1ea7n nh\u1eadn bi\u1ebft nh\u1eefng th\u1ee9 t\u01b0\u01a1ng t\u1ef1 nhau. L\u00fd do l\u00e0 v\u00ec hi\u1ec7n ch\u01b0a c\u00f3 k\u0129 thu\u1eadt n\u00e0o \u0111\u1ee7 t\u1ed1t \u0111\u1ec3 tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o c\u00f3 th\u1ec3 r\u00fat ra nh\u1eefng k\u1ebft lu\u1eadn \u0111\u00f3 m\u1ed9t c\u00e1ch logic. B\u00ean c\u1ea1nh \u0111\u00f3, v\u1eabn c\u00f2n nhi\u1ec1u th\u00e1ch th\u1ee9c trong vi\u1ec7c t\u00edch h\u1ee3p ki\u1ebfn th\u1ee9c tr\u1eebu t\u01b0\u1ee3ng v\u00e0o c\u00e1c h\u1ec7 th\u1ed1ng machine learning, v\u00ed d\u1ee5 nh\u01b0 th\u00f4ng tin v\u1ec1 v\u1eadt \u0111\u00f3 l\u00e0 g\u00ec, n\u00f3 d\u00f9ng \u0111\u1ec3 l\u00e0m g\u00ec, ng\u01b0\u1eddi ta hay x\u00e0i n\u00f3 l\u00e0m sao… N\u00f3i c\u00e1ch kh\u00e1c, machine learning ch\u01b0a c\u00f3 \u0111\u01b0\u1ee3c nh\u1eefng ki\u1ebfn th\u1ee9c th\u00f4ng th\u01b0\u1eddng nh\u01b0 con ng\u01b0\u1eddi.<\/p>\n

M\u1ed9t v\u00ed d\u1ee5 r\u1ea5t c\u1ee5 th\u1ec3 cho c\u00e1c b\u1ea1n d\u1ec5 hi\u1ec3u: trong m\u1ed9t d\u1ef1 \u00e1n c\u1ee7a Google, m\u1ed9t m\u1ea1ng l\u01b0\u1edbi th\u1ea7n kinh \u0111\u01b0\u1ee3c s\u1eed d\u1ee5ng \u0111\u1ec3 t\u1ea1o ra b\u1ee9c \u1ea3nh c\u1ee7a m\u1ed9t qu\u1ea3 t\u1ea1 m\u00e0 ng\u01b0\u1eddi ta hay c\u1ea7m t\u1eadp trong ph\u00f2ng gym. K\u1ebft qu\u1ea3 \u0111\u01b0a ra kh\u00e1 \u1ea5n t\u01b0\u1ee3ng: hai v\u00f2ng tr\u00f2n m\u00e0u x\u00e1m \u0111\u01b0\u1ee3c k\u1ebft n\u1ed1i b\u1eb1ng m\u1ed9t c\u00e1i \u1ed1ng ngang. Nh\u01b0ng \u1edf gi\u1eefa c\u00e1i \u1ed1ng n\u00e0y c\u00f2n c\u00f3 c\u00e1nh tay c\u1ee7a con ng\u01b0\u1eddi, v\u00e0 \u0111\u00e2y kh\u00f4ng ph\u1ea3i l\u00e0 th\u1ee9 n\u1eb1m trong “\u0111\u1ec1 b\u00e0i”. L\u00fd do th\u00ec kh\u00e1 l\u00e0 d\u1ec5 \u0111o\u00e1n: h\u1ec7 th\u1ed1ng \u0111\u01b0\u1ee3c d\u1ea1y v\u1ec1 qu\u1ea3 t\u1ea1 b\u1eb1ng nh\u1eefng t\u1ea5m h\u00ecnh ch\u1ee5p ng\u01b0\u1eddi ta \u0111ang c\u1eed t\u1ea1, v\u1eady n\u00ean d\u00ednh tay ng\u01b0\u1eddi v\u00e0o l\u00e0 chuy\u1ec7n hi\u1ec3n nhi\u00ean. H\u1ec7 th\u1ed1ng c\u00f3 th\u1ec3 bi\u1ebft \u0111\u01b0\u1ee3c qu\u1ea3 t\u1ea1 ra sao, nh\u01b0ng n\u00f3 kh\u00f4ng bi\u1ebft l\u00e0 qu\u1ea3 t\u1ea1o s\u1ebd kh\u00f4ng bao gi\u1edd c\u00f3 c\u00e1nh tay c\u1ea3.<\/p>\n

dumbbells.0. \u200b<\/p>\n

V\u1edbi m\u1ed9t s\u1ed1 b\u1ee9c h\u00ecnh \u0111\u01a1n gi\u1ea3n h\u01a1n nh\u01b0ng mahcine v\u1eabn nh\u1ea7m l\u1eabn. Th\u1eed nghi\u1ec7m c\u1ee7a m\u1ed9t nh\u00f3m nh\u00e0 nghi\u00ean c\u1ee9u cho th\u1ea5y r\u1eb1ng khi h\u1ecd \u0111\u01b0a m\u00e1y t\u00ednh xem m\u1ed9t lo\u1ea1t \u1ea3nh ch\u1ec9 c\u00f3 c\u00e1c pixel ng\u1eabu nhi\u00ean nh\u01b0ng m\u00e1y l\u1ea1i ch\u1eafc ch\u1eafn \u0111\u00e2y 95% l\u00e0 m\u1ed9t c\u00e1i… xe t\u1ea3i, hay l\u00e0 m\u1ed9t con sao bi\u1ec3n…<\/p>\n

\u0110\u00f3 ch\u01b0a ph\u1ea3i l\u00e0 t\u1ea5t c\u1ea3. Theo l\u1eddi c\u1ee7a nh\u00e0 khoa h\u1ecdc m\u00e1y t\u00ednh Hector Levesque, nh\u1eefng c\u00f4ng c\u1ee5 tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o hi\u1ec7n nay s\u1eed d\u1ee5ng nhi\u1ec1u “m\u00e1nh” \u0111\u1ec3 x\u00f3a \u0111i nh\u1eefng l\u1ed7 h\u1ed5ng th\u1eadt s\u1ef1 trong ki\u1ebfn th\u1ee9c c\u1ee7a ch\u00fang. C\u00e1c tr\u1ee3 l\u00fd \u1ea3o nh\u01b0 Siri hay Cortana th\u01b0\u1eddng l\u00e0m cho b\u1ea1n c\u00f3 c\u1ea3m gi\u00e1c b\u1ea1n \u0111ang n\u00f3i chuy\u1ec7n v\u1edbi ng\u01b0\u1eddi th\u1eadt v\u00ec ch\u00fang s\u1eed d\u1ee5ng c\u00e1c c\u00e2u n\u00f3i \u0111\u00f9a, c\u00e2u tr\u00edch d\u1eabn, c\u00e1c bi\u1ec3u hi\u1ec7n c\u1ea3m x\u00fac v\u00e0 nhi\u1ec1u th\u1ee9 kh\u00e1c, c\u1ed1t ch\u1ec9 \u0111\u1ec3 b\u1ea1n ph\u00e2n t\u00e2m.<\/p>\n

B\u1ea1n h\u00e3y th\u1eed h\u1ecfi nh\u1eefng th\u1ee9 c\u1ea7n \u0111\u1ebfn suy ngh\u0129 th\u00f4ng th\u01b0\u1eddng, v\u00ed d\u1ee5 nh\u01b0 “con c\u00e1 x\u1ea5u c\u00f3 l\u00e1i xe \u0111\u01b0\u1ee3c kh\u00f4ng” hay “m\u1ed9t c\u1ea7u th\u1ee7 \u0111\u00e1 banh c\u00f3 \u0111\u01b0\u1ee3c ph\u00e9p g\u1eafn c\u00e1nh \u0111\u1ec3 bay hay kh\u00f4ng”. C\u00e1c c\u00e2u h\u1ecfi d\u1ea1ng n\u00e0y qu\u00e1 ph\u1ee9c t\u1ea1p v\u1edbi nh\u1eefng h\u1ec7 th\u1ed1ng tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o hi\u1ec7n nay n\u00ean th\u01b0\u1eddng s\u1ebd kh\u00f4ng c\u00f3 k\u1ebft qu\u1ea3 n\u00e0o \u0111\u01b0\u1ee3c tr\u1ea3 v\u1ec1 cho b\u1ea1n, n\u1ebfu c\u00f3 c\u0169ng l\u00e0 nh\u1eefng th\u1ee9 kh\u00f4ng li\u00ean quan hay ch\u1ec9 \u0111\u01a1n gi\u1ea3n l\u00e0 search c\u00e2u n\u00f3i c\u1ee7a b\u1ea1n tr\u00ean Internet.<\/p>\n

M\u00f9a \u0111\u00f4ng AI<\/p>\n

Ng\u00e0nh khoa h\u1ecdc tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o l\u00e0 m\u1ed9t ng\u00e0nh d\u1ec5 l\u00ean nh\u01b0ng c\u0169ng d\u1ec5 xu\u1ed1ng. N\u0103m 1958, t\u1edd New York Times t\u1eebng n\u00f3i v\u1ec1 m\u1ed9t chi\u1ebfc m\u00e1y c\u00f3 kh\u1ea3 n\u0103ng ph\u00e2n bi\u1ec7t b\u00ean tr\u00e1i v\u00e0 b\u00ean ph\u1ea3i nh\u01b0 l\u00e0 m\u1ed9t d\u1ea1ng m\u1ea7m s\u1ed1ng c\u1ee7a robot th\u00f4ng minh. Nh\u01b0ng \u0111\u1ebfn nay, ch\u00fang ta v\u1eabn ch\u01b0a th\u1ec3 t\u1ea1o ra \u0111\u01b0\u1ee3c m\u1ed9t con robot n\u00e0o c\u00f3 tr\u00ed th\u00f4ng minh \u0111\u1ebfn v\u1eady. V\u00e0 khi nh\u1eefng l\u1eddi h\u1ee9a h\u1eb9n \u0111\u00f3 kh\u00f4ng \u0111\u01b0\u1ee3c th\u1ef1c hi\u1ec7n, ng\u01b0\u1eddi ta d\u00f9ng t\u1eeb “m\u00f9a \u0111\u00f4ng AI”. \u0110\u00f3 l\u00e0 giai \u0111o\u1ea1n m\u00e0 l\u01b0\u1ee3ng ti\u1ec1n \u0111\u1ea7u t\u01b0 cho AI gi\u1ea3m m\u1ea1nh, \u00edt c\u00f2n ai nh\u1eafc v\u1ec1 n\u00f3, v\u00e0 ng\u01b0\u1eddi ta c\u0169ng t\u1ecf ra ho\u00e0i nghi v\u1ec1 nh\u1eefng k\u1ebft qu\u1ea3 c\u00f3 th\u1ec3 \u0111\u1ea1t \u0111\u01b0\u1ee3c. T\u00ednh \u0111\u1ebfn nay \u0111\u00e3 c\u00f3 kho\u1ea3ng 6 “m\u00f9a \u0111\u00f4ng AI” nh\u1ecf v\u00e0 2 m\u00f9a l\u1edbn t\u1eebng xu\u1ea5t hi\u1ec7n, v\u00e0o cu\u1ed1i nh\u1eefng n\u0103m 70 v\u00e0 \u0111\u1ea7u c\u00e1c n\u0103m 90.<\/p>\n

Tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o c\u00f3 tri gi\u00e1c?<\/p>\n

Nhi\u1ec1u ng\u01b0\u1eddi \u0111ang l\u00e0m vi\u1ec7c trong ng\u00e0nh tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o cho r\u1eb1ng ch\u00fang ta s\u1ebd r\u1ea5t kh\u00f3 \u0111\u1ec3 t\u1ea1o ra m\u1ed9t tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o c\u00f3 tri gi\u00e1c. “C\u00f3 r\u1ea5t \u00edt b\u1eb1ng ch\u1ee9ng \u1edf th\u1eddi \u0111i\u1ec3m hi\u1ec7n t\u1ea1i cho th\u1ea5y hi v\u1ecdng v\u1ec1 vi\u1ec7c t\u1ea1o ra m\u1ed9t tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o c\u00f3 t\u00ednh linh ho\u1ea1t cao v\u00e0 l\u00e0m \u0111\u01b0\u1ee3c nh\u1eefng th\u1ee9 m\u00e0 ch\u00fang kh\u00f4ng \u0111\u01b0\u1ee3c t\u1ea1o ra \u0111\u1ec3 th\u1ef1c hi\u1ec7n”, theo l\u1eddi gi\u00e1o s\u01b0 Andrei Barbu t\u1eeb \u0111\u1ea1i h\u1ecdc MIT. \u00d4ng nh\u1ea5n m\u1ea1nh vi\u1ec7c nghi\u00ean c\u1ee9u tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o hi\u1ec7n nay ch\u1ec9 t\u1ea1o ra \u0111\u01b0\u1ee3c nh\u1eefng h\u1ec7 th\u1ed1ng \u0111\u01b0\u1ee3c t\u1ed1i \u01b0u \u0111\u1ec3 gi\u1ea3i quy\u1ebft m\u1ed9t v\u1ea5n \u0111\u1ec1 c\u1ee5 th\u1ec3 m\u00e0 th\u00f4i.<\/p>\n

C\u0169ng \u0111\u00e3 c\u00f3 m\u1ed9t s\u1ed1 c\u00f4ng tr\u00ecnh nghi\u00ean c\u1ee9u v\u1ec1 vi\u1ec7c cho m\u00e1y t\u00ednh h\u1ecdc nh\u01b0ng kh\u00f4ng c\u1ea7n s\u1ef1 gi\u00e1m s\u00e1t, t\u1ee9c l\u00e0 c\u1ee9 \u0111\u01b0a d\u1eef li\u1ec7u cho m\u00e1y h\u1ecdc m\u00e0 kh\u00f4ng d\u00e1n nh\u00e3n \u0111\u00fang sai hay gi\u1ea3i th\u00edch g\u00ec c\u1ea3. Tuy nhi\u00ean, Andrei Barbu nh\u1eadn x\u00e9t nh\u1eefng d\u1ef1 \u00e1n nh\u01b0 th\u1ebf n\u00e0y v\u1eabn ch\u01b0a c\u00f3 ti\u1ebfn tri\u1ec3n n\u00e0o v\u00e0 v\u1eabn c\u00f2n r\u1ea5t xa x\u1eddi \u0111\u1ec3 \u0111\u1ebfn \u0111\u01b0\u1ee3c ng\u00e0y c\u00f3 k\u1ebft qu\u1ea3. M\u1ed9t v\u00ed d\u1ee5 \u0111\u00e3 t\u1eebng xu\u1ea5t hi\u1ec7n \u0111\u00f3 l\u00e0 m\u1ed9t h\u1ec7 th\u1ed1ng neural network c\u1ee7a Google \u0111\u00e3 t\u1eebng l\u1ea5y ng\u1eabu nhi\u00ean \u1ea3nh thumbnail c\u1ee7a 10 tri\u1ec7u video tr\u00ean YouTube \u0111\u1ec3 t\u1ef1 d\u1ea1y m\u00ecnh con m\u00e8o tr\u00f4ng gi\u1ed1ng nh\u01b0 th\u1ebf n\u00e0o. Th\u1ebf nh\u01b0ng, Google n\u00f3i \u0111\u00e2y ch\u1ec9 l\u00e0 m\u1ed9t th\u1eed nghi\u1ec7m v\u00e0 kh\u00f4ng n\u00f3i g\u00ec v\u1ec1 \u0111\u1ed9 ch\u00ednh x\u00e1c c\u1ee7a n\u00f3 c\u1ea3.<\/p>\n

N\u00f3i c\u00e1ch kh\u00e1c, ch\u00fang ta v\u1eabn ch\u01b0a bi\u1ebft c\u00e1ch l\u00e0m cho m\u00e1y t\u00ednh t\u1ef1 h\u1ecdc m\u00e0 kh\u00f4ng c\u1ea7n gi\u00e1m s\u00e1t. \u0110\u00f3 l\u00e0 r\u00e0o c\u1ea3n l\u1edbn nh\u1ea5t. T\u1ee9c l\u00e0 v\u1eabn c\u00f2n r\u1ea5t xa m\u1edbi t\u1edbi c\u00e1i ng\u00e0y m\u00e0 robot c\u00f3 tri gi\u00e1c v\u00e0 \u0111\u00e1nh tr\u1ea3 l\u1ea1i con ng\u01b0\u1eddi \ud83d\ude00<\/p>\n

Nh\u01b0 Elon Musk \u0111\u00e3 n\u00f3i, c\u00f4ng ty \u00f4ng c\u00f3 t\u1ea1o ra m\u1ed9t tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o cho t\u00ednh n\u0103ng t\u1ef1 l\u00e1i xe tr\u00ean nh\u1eefng chi\u1ebfc Tesla. Nh\u01b0ng \u00f4ng ch\u01b0a bao gi\u1edd n\u00f3i r\u1eb1ng n\u00f3 s\u1ebd bi\u1ebft h\u1ebft m\u1ecdi th\u1ee9. \u0110\u00e2y ch\u1ec9 \u0111\u01a1n gi\u1ea3n l\u00e0 m\u1ed9t m\u1ea1ng l\u01b0\u1edbi gi\u00fap c\u00e1c xe h\u1ecdc h\u1ecfi l\u1eabn nhau. Khi m\u1ed9t chi\u1ebfc n\u00e0y h\u1ecdc \u0111\u01b0\u1ee3c \u0111i\u1ec1u g\u00ec th\u00ec nh\u1eefng chi\u1ebfc kh\u00e1c c\u0169ng bi\u1ebft \u0111i\u1ec1u t\u01b0\u01a1ng t\u1ef1. K\u1ebft qu\u1ea3 cu\u1ed1i c\u00f9ng kh\u00f4ng ph\u1ea3i l\u00e0 nh\u1eefng chi\u1ebfc xe c\u00f3 th\u1ec3 l\u00e0m m\u1ecdi th\u1ee9 tr\u00ean \u0111\u1eddi, n\u00f3 ch\u1ec9 \u0111\u1ec3 gi\u1ea3i quy\u1ebft m\u1ed9t v\u1ea5n \u0111\u1ec1 r\u1ea5t c\u1ee5 th\u1ec3 m\u00e0 th\u00f4i.<\/p>\n

Ngu\u1ed3n: The Verge, Evolvingai, Wikipedia, Wikipedia (2), Toptal<\/strong> \u200b<\/em><\/p>\n

Ngu\u1ed3n tinhte.vn<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"

Tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o ( artificial intelligence – AI) gi\u1edd xu\u1ea5t hi\u1ec7n \u1edf kh\u1eafp m\u1ecdi n\u01a1i. N\u00f3 l\u00e0 th\u1ee9 \u0111\u01b0\u1ee3c s\u1eed d\u1ee5ng \u0111\u1ec3 tr\u1ea3 l\u1eddi email t\u1ef1 \u0111\u1ed9ng tr\u00ean Gmail, h\u1ecdc c\u00e1ch l\u00e1i xe cho ch\u00fang ta ng\u1ed3i ch\u01a1i, s\u1eafp x\u1ebfp l\u1ea1i \u1ea3nh c\u1ee7a nh\u1eefng chuy\u1ebfn \u0111i ch\u01a1i th\u00e0nh t\u1eebng album ri\u00eang bi\u1ec7t, th\u1eadm ch\u00ed […]<\/p>\n","protected":false},"author":17,"featured_media":5337,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[47],"tags":[],"class_list":["post-4803","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-tin-tuc"],"_links":{"self":[{"href":"https:\/\/casta.ttud.com.vn\/index.php\/wp-json\/wp\/v2\/posts\/4803","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/casta.ttud.com.vn\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/casta.ttud.com.vn\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/casta.ttud.com.vn\/index.php\/wp-json\/wp\/v2\/users\/17"}],"replies":[{"embeddable":true,"href":"https:\/\/casta.ttud.com.vn\/index.php\/wp-json\/wp\/v2\/comments?post=4803"}],"version-history":[{"count":2,"href":"https:\/\/casta.ttud.com.vn\/index.php\/wp-json\/wp\/v2\/posts\/4803\/revisions"}],"predecessor-version":[{"id":5338,"href":"https:\/\/casta.ttud.com.vn\/index.php\/wp-json\/wp\/v2\/posts\/4803\/revisions\/5338"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/casta.ttud.com.vn\/index.php\/wp-json\/wp\/v2\/media\/5337"}],"wp:attachment":[{"href":"https:\/\/casta.ttud.com.vn\/index.php\/wp-json\/wp\/v2\/media?parent=4803"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/casta.ttud.com.vn\/index.php\/wp-json\/wp\/v2\/categories?post=4803"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/casta.ttud.com.vn\/index.php\/wp-json\/wp\/v2\/tags?post=4803"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}