{"id":686,"date":"2017-11-05T07:36:57","date_gmt":"2017-11-05T00:36:57","guid":{"rendered":"https:\/\/rosetta.vn\/nguyenxuanxanh\/?p=686"},"modified":"2024-09-02T22:17:22","modified_gmt":"2024-09-02T15:17:22","slug":"mot-googler-viet-nam-tim-cach-day-may-hieu-cam-xuc-cua-con-nguoi","status":"publish","type":"post","link":"https:\/\/rosetta.vn\/nguyenxuanxanh\/mot-googler-viet-nam-tim-cach-day-may-hieu-cam-xuc-cua-con-nguoi\/","title":{"rendered":"M\u1ed9t Googler Vi\u1ec7t Nam t\u00ecm c\u00e1ch d\u1ea1y m\u00e1y hi\u1ec3u c\u1ea3m x\u00fac c\u1ee7a con ng\u01b0\u1eddi"},"content":{"rendered":"

 <\/p>\n

M\u1ed8T GOOGLER VI\u1ec6T NAM<\/span><\/h1>\n

T\u00ccM C\u00c1CH D\u1ea0Y M\u00c1Y HI\u1ec2U\u00a0<\/span>C\u1ea2M X\u00daC C\u1ee6A CON NG\u01af\u1edcI:<\/span><\/h2>\n

\u2014 TI\u1ebeN S\u0128 L\u00ca VI\u1ebeT QU\u1ed0C \u2014<\/span><\/h2>\n

T\u00e1c gi\u1ea3: Daniela Hernandez<\/strong> (12.02.14)<\/span><\/p>\n

(Ngu\u1ed3n: WIRED)1<\/sup><\/p>\n

Ng\u01b0\u1eddi d\u1ecbch<\/i>: V\u00f5 Th\u00e0nh Minh Tu\u1ec7 & Nguy\u1ec5n Xu\u00e2n Xanh<\/strong><\/span><\/p>\n

L\u1eddi n\u00f3i \u0111\u1ea7u<\/strong><\/span> c\u1ee7a ng\u01b0\u1eddi d\u1ecbch<\/i>. L\u00ea Vi\u1ebft Qu\u1ed1c m\u00e0 b\u00e0i b\u00e1o d\u01b0\u1edbi \u0111\u00e2y n\u00f3i v\u1ec1 anh l\u00e0 m\u1ed9t trong nh\u1eefng chuy\u00ean gia \u0111\u1ea7u ng\u00e0nh tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o tr\u00ean th\u1ebf gi\u1edbi. Anh l\u00e0 m\u1ed9t trong nh\u1eefng th\u00e0nh vi\u00ean then ch\u1ed1t t\u1ea1i trung t\u00e2m nghi\u00ean c\u1ee9u Google Brain v\u00e0 t\u1eebng \u0111\u01b0\u1ee3c t\u1ea1p ch\u00ed c\u00f4ng ngh\u1ec7 c\u1ee7a MIT (MIT Technology Review) b\u1ea7u v\u00e0o 35 Innovators d\u01b0\u1edbi 35 tu\u1ed5i. Anh Qu\u1ed1c sinh ra t\u1ea1i m\u1ed9t l\u00e0ng qu\u00ea \u1edf mi\u1ec1n Trung, l\u00fac nh\u1ecf l\u00e0ng kh\u00f4ng c\u00f3 \u0111i\u1ec7n n\u00ean anh hay ra th\u01b0 vi\u1ec7n \u0111\u1ecdc say s\u01b0a s\u00e1ch v\u1edf. Kho\u1ea3ng 14 tu\u1ed5i, anh nh\u1eadn ra r\u1eb1ng con ng\u01b0\u1eddi c\u1ea7n nh\u1eefng c\u1ed7 m\u00e1y th\u00f4ng minh c\u00f3 th\u1ec3 “s\u00e1ng ch\u1ebf” ra nh\u1eefng \u00fd t\u01b0\u1edfng m\u1edbi – \u0111\u00e2y l\u00e0 gi\u1ea5c m\u01a1 c\u1ee7a anh. T\u1ed1t nghi\u1ec7p Qu\u1ed1c h\u1ecdc Hu\u1ebf n\u0103m 2000, Qu\u1ed1c v\u00e0o \u0111\u1ea1i h\u1ecdc B\u00e1ch Khoa TP.HCM h\u1ecdc m\u1ed9t h\u1ecdc k\u1ef3 th\u00ec nh\u1eadn \u0111\u01b0\u1ee3c h\u1ecdc b\u1ed5ng ph\u00e1t tri\u1ec3n c\u1ee7a \u00dac (AusAID). Qu\u1ed1c ra H\u00e0 N\u1ed9i h\u1ecdc ti\u1ebfng Anh 6 th\u00e1ng th\u00ec sang \u00dac h\u1ecdc khoa h\u1ecdc m\u00e1y t\u00ednh \u1edf ANU (Australian National University) v\u00e0o n\u0103m 2001 \u0111\u1ec3 b\u1eaft \u0111\u1ea7u 4 n\u0103m \u0111\u1ea1i h\u1ecdc. \u0110\u1ebfn cu\u1ed1i n\u0103m th\u1ee9 nh\u1ea5t \u0111\u1ea7u n\u0103m th\u1ee9 hai \u0111\u1ea1i h\u1ecdc, Qu\u1ed1c tham d\u1ef1 ch\u01b0\u01a1ng tr\u00ecnh Distinguished Scholar l\u00e0m c\u00f4ng tr\u00ecnh nghi\u00ean c\u1ee9u c\u1ee7a ANU li\u00ean k\u1ebft v\u1edbi NICTA (National ICT Australia). T\u1eeb gi\u1eefa n\u0103m 2004, Qu\u1ed1c l\u00e0m vi\u1ec7c v\u1ec1 machine learning (m\u1ed9t chuy\u00ean ng\u00e0nh c\u1ee7a b\u1ed9 m\u00f4n tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o) v\u1edbi GS Alex Smola. Giai \u0111o\u1ea1n n\u00e0y Qu\u1ed1c c\u0169ng \u0111\u00e3 c\u00f3 m\u1ed9t s\u1ed1 b\u00e0i b\u00e1o \u0111\u01b0\u1ee3c \u0111\u0103ng trong c\u00e1c k\u1ef7 y\u1ebfu h\u1ed9i th\u1ea3o v\u00e0 m\u1ed9t s\u1ed1 t\u1ea1p ch\u00ed chuy\u00ean ng\u00e0nh. \u0110\u1ea7u n\u0103m 2006, Qu\u1ed1c c\u00f3 sang M\u1ef9 c\u1ed9ng t\u00e1c v\u1edbi m\u1ed9t nh\u00f3m nghi\u00ean c\u1ee9u \u1edf Microsoft. N\u0103m 2007 Qu\u1ed1c sang \u0110\u1ee9c l\u00e0m nghi\u00ean c\u1ee9u v\u1edbi vi\u1ec7n Max Planck Biological Cybernetics. C\u00f9ng th\u1eddi gian \u0111\u00f3, Qu\u1ed1c n\u1ed9p h\u1ed3 s\u01a1 l\u00e0m ti\u1ebfn s\u0129 \u1edf Stanford v\u00e0 \u0111\u01b0\u1ee3c ch\u1ea5p nh\u1eadn. Qu\u1ed1c \u0111\u1ebfn Stanford l\u00e0m vi\u1ec7c v\u1ec1 machine learning d\u01b0\u1edbi s\u1ef1 h\u01b0\u1edbng d\u1eabn c\u1ee7a GS Andrew Ng. Trong th\u1eddi gian nghi\u00ean c\u1ee9u \u1edf Stanfrod, t\u1eeb hai n\u0103m nay, Google c\u00f3 l\u1eddi m\u1eddi c\u1ed9ng t\u00e1c n\u00ean Qu\u1ed1c \u0111\u1ebfn l\u00e0m vi\u1ec7c \u1edf \u0111\u00e2y c\u00f9ng nh\u00f3m v\u1edbi GS Andrew Ng. Sau khi b\u1ea3o v\u1ec7 th\u00e0nh c\u00f4ng lu\u1eadn \u00e1n ti\u1ebfn s\u0129 v\u00e0o \u0111\u1ea7u n\u0103m 2013, Qu\u1ed1c b\u1eaft \u0111\u1ea7u l\u00e0m vi\u1ec7c t\u1ea1i c\u00f4ng ty Google chuy\u00ean nghi\u00ean c\u1ee9u v\u1ec1 tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o v\u00e0 ng\u00f4n ng\u1eef. Nh\u1eefng s\u1ea3n ph\u1ea9m nh\u01b0 Google Translate (c\u00f4ng c\u1ee5 d\u1ecbch c\u1ee7a Google) v\u00e0 Google Search (c\u00f4ng c\u1ee5 t\u00ecm ki\u1ebfm c\u1ee7a Google) \u0111\u1ec1u c\u00f3 c\u00e1c \u0111\u00f3ng g\u00f3p Vi\u1ec7t Qu\u1ed1c.<\/p>\n

N\u1ebfu \u0111\u01b0\u1ee3c c\u1ea3i ti\u1ebfn \u0111\u00e1ng k\u1ec3, c\u00f4ng ngh\u1ec7 d\u1ecbch thu\u1eadt c\u1ee7a Google s\u1ebd c\u00f3 t\u1ea7m quan tr\u1ecdng r\u1ea5t l\u1edbn cho t\u01b0\u01a1ng lai, c\u00f3 \u00fd ngh\u0129a nh\u01b0 m\u1ed9t cu\u1ed9c c\u00e1ch m\u1ea1ng. N\u00f3 kh\u00f4ng ph\u1ea3i ch\u1ec9 nh\u1eb1m ph\u1ee5c v\u1ee5 kh\u00e1ch h\u00e0ng tr\u01b0\u1edbc m\u1eaft, m\u00e0 th\u00fac \u0111\u1ea9y cu\u1ed9c to\u00e0n c\u1ea7u h\u00f3a tri th\u1ee9c<\/em> nh\u00e2n lo\u1ea1i. Tinh hoa th\u1ebf gi\u1edbi c\u00f3 th\u1ec3 ch\u1ed1c l\u00e1t \u0111\u1ebfn t\u1edbi t\u1eadn c\u1eeda m\u1ecdi nh\u00e0, b\u1eb1ng ng\u00f4n ng\u1eef m\u1eb9 \u0111\u1ebb. D\u1ecbch thu\u1eadt g\u00f3p ph\u1ea7n c\u1ea3i thi\u1ec7n v\u00e0 ph\u00e1t tri\u1ec3n v\u0103n h\u00f3a, l\u1ea5p \u0111\u1ea7y l\u1ed7 h\u1ed5ng v\u0103n h\u00f3a \u1edf nh\u1eefng v\u00f9ng tr\u0169ng. B\u1eb1ng s\u1ef1 chan h\u00f2a v\u0103n h\u00f3a, c\u00e1c d\u00e2n t\u1ed9c s\u1ebd hi\u1ec3u bi\u1ebft nhau, g\u1ea7n g\u0169i nhau h\u01a1n, c\u00f3 th\u1ec3 s\u1ebd kh\u00f4ng c\u00f2n th\u1ea5y ph\u1ea3i gi\u1ebft h\u1ea1i l\u1eabn nhau nh\u01b0 th\u1eddi hoang d\u00e3.<\/p>\n

Ch\u00fang ta r\u1ea5t vui m\u1eebng c\u00f3 m\u1ed9t ng\u01b0\u1eddi Vi\u1ec7t Nam t\u00e0i hoa \u0111\u1ee9ng gi\u1eefa d\u00f2ng l\u1ecbch s\u1eed \u0111\u1ec3 t\u00e1c \u0111\u1ed9ng: L\u00ea Vi\u1ebft Qu\u1ed1c.<\/p>\n

\"Quoc<\/p>\n

L\u00ea Vi\u1ebft Qu\u1ed1c (Ariel Zambelich\/WIRED)<\/p>\n

 <\/p>\n

Vi\u1ec7t Qu\u1ed1c nh\u00ecn th\u1ebf gi\u1edbi nh\u01b0 m\u1ed9t chu\u1ed7i s\u1ed1<\/b><\/p>\n

M\u1ed9t b\u1ee9c \u1ea3nh k\u1ef9 thu\u1eadt s\u1ed1 kh\u00f4ng g\u00ec h\u01a1n l\u00e0 c\u00e1c con s\u1ed1, anh n\u00f3i, v\u00e0 n\u1ebfu b\u1ea1n t\u00e1ch \u0111\u01b0\u1ee3c c\u00e1c t\u1eeb n\u00f3i th\u00e0nh c\u00e1c \u00e2m v\u1ecb (phonemes), b\u1ea1n c\u0169ng c\u00f3 th\u1ec3 bi\u1ebfn ch\u00fang th\u00e0nh c\u00e1c con s\u1ed1. Sau \u0111\u00f3, b\u1ea1n c\u00f3 th\u1ec3 \u0111\u01b0a s\u1ed1 li\u1ec7u \u0111\u00f3 v\u00e0o m\u00e1y, v\u00e0 \u0111i\u1ec1u \u0111\u00f3 c\u00f3 ngh\u0129a l\u00e0 m\u00e1y sau c\u00f9ng c\u00f3 th\u1ec3 hi\u1ec3u \u0111\u01b0\u1ee3c c\u00e1c n\u1ed9i dung c\u1ee7a h\u00ecnh \u1ea3nh v\u00e0 \u00fd ngh\u0129a c\u1ee7a c\u00e1c t\u1eeb. Facebook c\u00f3 th\u1ec3 nh\u1eadn ra khu\u00f4n m\u1eb7t c\u1ee7a b\u1ea1n, v\u00e0 Google c\u00f3 th\u1ec3 h\u00e0nh \u0111\u1ed9ng theo nh\u1eefng t\u1eeb c\u1ee5 th\u1ec3 b\u1ea1n n\u00f3i.<\/p>\n

Nh\u01b0ng Qu\u1ed1c mu\u1ed1n \u0111i xa h\u01a1n. Anh mu\u1ed1n t\u1ea1o ra c\u00e1c c\u00f4ng ngh\u1ec7 c\u00f3 th\u1ec3 l\u1ea5y to\u00e0n b\u1ed9 c\u00e1c c\u00e2u (sentences), to\u00e0n b\u1ed9 c\u00e1c \u0111o\u1ea1n v\u0103n (paragraphs) , v\u00e0 c\u00e1c lo\u1ea1i ng\u00f4n ng\u1eef t\u1ef1 nhi\u00ean kh\u00e1c v\u00e0 bi\u1ebfn ch\u00fang th\u00e0nh c\u00e1c con s\u1ed1 \u2212 ho\u1eb7c c\u00e1c vect\u01a1, c\u00e1c c\u1ea5u tr\u00fac to\u00e1n m\u00e0 c\u00e1c nh\u00e0 khoa h\u1ecdc m\u00e1y t\u00ednh s\u1eed d\u1ee5ng \u0111\u1ec3 d\u1ecbch nh\u1eefng \u0111i\u1ec1u ch\u00fang ta nh\u00ecn th\u1ea5y v\u00e0 nghe th\u1ea5y th\u00e0nh th\u00f4ng tin m\u00e0 m\u00e1y c\u00f3 th\u1ec3 hi\u1ec3u. Anh ta th\u1eadm ch\u00ed c\u00f2n th\u0103m d\u00f2, kh\u00e1m ph\u00e1 nh\u1eefng kh\u1ea3 n\u0103ng m\u00e1y c\u00f3 th\u1ec3 hi\u1ec3u nh\u1eefng th\u1ee9 nh\u01b0 \u00fd ki\u1ebfn \u200b\u200bv\u00e0 c\u1ea3m x\u00fac.<\/p>\n

\u0110\u1ed1i v\u1edbi m\u1ed9t s\u1ed1 c\u00f4ng ngh\u1ec7 n\u00e0y, vi\u1ec7c th\u1ef1c hi\u1ec7n ch\u00fang c\u00f2n l\u00e0 m\u1ed9t ch\u1eb7ng \u0111\u01b0\u1eddng d\u00e0i. Nh\u01b0ng Vi\u1ec7t Qu\u1ed1c c\u00f3 nhi\u1ec1u ngu\u1ed3n l\u1ef1c \u0111\u1ec3 s\u1eed d\u1ee5ng h\u01a1n l\u00e0 ph\u1ea7n l\u1edbn nh\u1eefng ng\u01b0\u1eddi kh\u00e1c. Anh l\u00e0m vi\u1ec7c t\u1ea1i trung t\u00e2m Google Brain, n\u01a1i ng\u01b0\u1eddi kh\u1ed5ng l\u1ed3 Google \u0111i s\u00e2u v\u00e0o ng\u00e0nh “h\u1ecdc s\u00e2u” (Deep Learning), m\u1ed9t h\u00ecnh th\u1ee9c tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o, lo\u1ea1i c\u00f4ng ngh\u1ec7 x\u1eed l\u00fd d\u1eef li\u1ec7u theo nh\u1eefng c\u00e1ch th\u1ee9c b\u1eaft ch\u01b0\u1edbc b\u1ed9 n\u00e3o con ng\u01b0\u1eddi, \u00edt ra v\u1ec1 v\u00e0i ph\u01b0\u01a1ng di\u1ec7n n\u00e0o \u0111\u00f3.<\/p>\n

Qu\u1ed1c l\u00e0 m\u1ed9t trong nh\u1eefng l\u1eadp tr\u00ecnh vi\u00ean ch\u00ednh c\u1ee7a c\u00f4ng tr\u00ecnh \u0111\u1ea7u ti\u00ean c\u1ee7a Google Brain, m\u1ed9t h\u1ec7 th\u1ed1ng t\u1ef1 h\u1ecdc c\u00e1ch nh\u1eadn d\u1ea1ng m\u00e8o tr\u00ean c\u00e1c h\u00ecnh \u1ea3nh tr\u00ean YouTube, v\u00e0 k\u1ec3 t\u1eeb \u0111\u00f3, ng\u01b0\u1eddi Vi\u1ec7t 32 tu\u1ed5i n\u00e0y \u0111\u00e3 gi\u00fap \u0111\u1ee1 x\u00e2y d\u1ef1ng c\u00e1c h\u1ec7 th\u1ed1ng Google nh\u1eadn d\u1ea1ng c\u00e1c t\u1eeb n\u00f3i tr\u00ean \u0111i\u1ec7n tho\u1ea1i Android, v\u00e0 t\u1ef1 \u0111\u1ed9ng \u0111\u00e1nh d\u1ea5u (tag) \u1ea3nh c\u1ee7a b\u1ea1n tr\u00ean web, c\u1ea3 hai k\u1ef9 thu\u1eadt \u0111\u1ec1u \u0111\u01b0\u1ee3c h\u1ed7 tr\u1ee3 t\u1eeb c\u00f4ng ngh\u1ec7 h\u1ecdc-s\u00e2u.<\/p>\n

\u201cH\u1ecdc s\u00e2u\u201d \u0111ang th\u00fac \u0111\u1ea9y ph\u00e1t tri\u1ec3n c\u00e1c c\u00f4ng c\u1ee5 t\u01b0\u01a1ng t\u1ef1 t\u1ea1i c\u00e1c c\u00f4ng ty internet kh\u1ed5ng l\u1ed3 kh\u00e1c, bao g\u1ed3m Facebook v\u00e0 Microsoft. Kh\u1ea3 n\u0103ng c\u1ee7a n\u00f3 t\u0103ng t\u1ed1c nh\u1eadn d\u1ea1ng h\u00ecnh \u1ea3nh v\u00e0 gi\u1ecdng n\u00f3i \u0111\u00e3 \u0111\u01b0\u1ee3c ghi nh\u1eadn r\u00f5 r\u1ec7t, v\u00e0 c\u00f4ng ty kh\u1ed5ng l\u1ed3 Trung Qu\u1ed1c Baidu \u0111\u00e3 n\u00f3i c\u00f4ng khai v\u1ec1 c\u00f4ng ngh\u1ec7 n\u00e0y c\u00f3 th\u1ec3 t\u0103ng doanh thu nh\u01b0 th\u1ebf n\u00e0o b\u1eb1ng c\u00e1ch cung c\u1ea5p m\u1ed9t c\u00e1ch t\u1ed1t h\u01a1n cho kh\u00e1ch h\u00e0ng \u0111\u1ec3 nh\u1eafm m\u1ee5c ti\u00eau qu\u1ea3ng c\u00e1o. Nh\u01b0ng Qu\u1ed1c l\u00e0 m\u1ed9t trong nh\u1eefng ng\u01b0\u1eddi mong mu\u1ed1n \u0111\u1ea9y c\u00f4ng ngh\u1ec7 v\u00e0o nhi\u1ec1u l\u0129nh v\u1ef1c h\u01a1n n\u1eefa, bao g\u1ed3m m\u1ecdi th\u1ee9 t\u1eeb s\u1ef1 hi\u1ec3u bi\u1ebft v\u1ec1 ng\u00f4n ng\u1eef t\u1ef1 nhi\u00ean \u0111\u1ebfn ng\u01b0\u1eddi m\u00e1y, \u0111\u1ebfn c\u00e1ch c\u00f4ng c\u1ee5 t\u00ecm ki\u1ebfm hi\u1ec7n gi\u1edd.<\/p>\n

T\u1ea1i Google, anh \u0111\u00e3 gi\u00fap ph\u00e1t tri\u1ec3n m\u1ed9t h\u1ec7 th\u1ed1ng chuy\u1ec3n ho\u00e1 t\u1eeb v\u1ef1ng th\u00e0nh vect\u01a1 to\u00e1n h\u1ecdc. V\u00e0 theo Google, c\u00f4ng tr\u00ecnh n\u00e0y sau \u0111\u00f3 s\u1ebd \u0111\u01b0a v\u00e0o h\u1ec7 th\u1ed1ng \u0111\u01b0\u1ee3c ph\u00e1t tri\u1ec3n ch\u1ee7 y\u1ebfu b\u1edfi m\u1ed9t nh\u00e0 nghi\u00ean c\u1ee9u c\u00f3 t\u00ean l\u00e0 Tomas Mikolov2<\/sup>. \u0110\u01b0\u1ee3c g\u1ecdi l\u00e0 Word2Vec, h\u1ec7 th\u1ed1ng x\u00e1c \u0111\u1ecbnh nh\u1eefng t\u1eeb kh\u00e1c nhau tr\u00ean m\u1ea1ng \u0111\u01b0\u1ee3c li\u00ean quan nh\u01b0 th\u1ebf n\u00e0o, v\u00e0 Google hi\u1ec7n \u0111ang s\u1eed d\u1ee5ng n\u00f3 nh\u01b0 m\u1ed9t ph\u01b0\u01a1ng ti\u1ec7n \u0111\u1ec3 t\u0103ng c\u01b0\u1eddng “\u0111\u1ed3 th\u1ecb ki\u1ebfn \u200b\u200bth\u1ee9c” (knowledge graph) \u2212 \u0111\u00f3 l\u00e0 m\u1ed9t t\u1eadp h\u1ee3p l\u1edbn c\u00e1c k\u1ebft n\u1ed1i gi\u1eefa c\u00e1c kh\u00e1i ni\u1ec7m li\u00ean quan l\u00e0m cho c\u1ed7 m\u00e1y t\u00ecm ki\u1ebfm c\u1ee7a Google ho\u1ea1t \u0111\u1ed9ng r\u1ea5t t\u1ed1t. \u0110\u00f3 l\u00e0 m\u1ed9t c\u00e1ch \u0111\u1ec3 ki\u1ec3m tra c\u00e1c s\u1ef1 ki\u1ec7n m\u1ed9t c\u00e1ch nhanh ch\u00f3ng h\u01a1n \u2212 v\u00e0 \u1edf quy m\u00f4 l\u1edbn h\u01a1n3<\/sup>. V\u00e0 c\u00f2n nhi\u1ec1u th\u1ee9 h\u01a1n n\u1eefa \u0111ang tr\u00ean \u0111\u01b0\u1eddng th\u1ef1c hi\u1ec7n.<\/p>\n

‘Th\u1ef1c l\u00e0 b\u1ef1c b\u1ed9i’<\/b><\/p>\n

Khi Qu\u1ed1c l\u1ea7n \u0111\u1ea7u ti\u00ean b\u1eaft \u0111\u1ea7u nghi\u00ean c\u1ee9u AI, v\u00e0o nh\u1eefng n\u0103m 1990, n\u00f3 “th\u1ef1c s\u1ef1 l\u00e0m phi\u1ec1n” anh. Anh kh\u00f4ng th\u00edch c\u00e1c h\u1ec7 th\u1ed1ng m\u00e1y-h\u1ecdc d\u1ef1a qu\u00e1 nhi\u1ec1u v\u00e0o \u0111\u1ea7u v\u00e0o (input) t\u1eeb c\u00e1c k\u1ef9 s\u01b0 con ng\u01b0\u1eddi. M\u00e1y c\u00f3 th\u1ec3 h\u1ecdc \u2212 \u00edt nh\u1ea5t \u0111\u1ebfn m\u1ed9t m\u1ee9c \u0111\u1ed9 nh\u1ea5t \u0111\u1ecbnh \u2212 nh\u01b0ng ch\u00fang c\u1ea7n \u0111\u1ebfn h\u00e0ng l\u1ed1 h\u01b0\u1edbng d\u1eabn \u0111\u1ec3 l\u00e0m \u0111\u01b0\u1ee3c nh\u01b0 v\u1eady. Ch\u00fang kh\u00f4ng th\u1ec3 h\u1ecdc c\u00e1ch nh\u1eadn ra nh\u1eefng b\u1ee9c \u1ea3nh tr\u1eeb khi nh\u1eefng b\u1ee9c \u1ea3nh \u0111\u00f3 \u0111\u01b0\u1ee3c d\u00e1n nh\u00e3n ch\u1ec9 d\u1eabn t\u1eebng c\u00e1i (v\u00ed d\u1ee5: \u1ea3nh n\u00e0y l\u00e0 con m\u00e8o, \u1ea3nh kia l\u00e0 con c\u00e1, \u2026). \u0110\u1ec3 \u0111\u1ea1t \u0111\u01b0\u1ee3c s\u1ef1 th\u00f4ng minh th\u1ef1c s\u1ef1, Qu\u1ed1c n\u00f3i, m\u00e1y ph\u1ea3i t\u1ef1 h\u1ecdc, kh\u00f4ng c\u1ea7n nh\u00e3n hi\u1ec7u ch\u1ec9 d\u1eabn, gi\u1ed1ng nh\u01b0 con ng\u01b0\u1eddi.<\/p>\n

“Ch\u00fang ta (con ng\u01b0\u1eddi) h\u1ecdc \u0111\u01b0\u1ee3c t\u1eeb r\u1ea5t nhi\u1ec1u d\u1eef li\u1ec7u th\u00f4 kh\u00f4ng c\u1ea7n nh\u00e3n g\u00ec c\u1ea3”, Qu\u1ed1c n\u00f3i, ng\u01b0\u1eddi \u0111\u00e3 nghi\u00ean c\u1ee9u v\u1ec1 tr\u00ed th\u00f4ng minh nh\u00e2n t\u1ea1o t\u1ea1i Stanford, c\u00f9ng v\u1edbi Andrew Ng, hi\u1ec7n l\u00e0 gi\u00e1m \u0111\u1ed1c nghi\u00ean c\u1ee9u c\u1ee7a Baidu v\u00e0 l\u00e0 m\u1ed9t trong nh\u1eefng ng\u01b0\u1eddi s\u00e1ng l\u1eadp ra d\u1ef1 \u00e1n Google Brain. “Th\u1eadt l\u00e0 tuy\u1ec7t v\u1eddi n\u1ebfu ch\u00fang ta c\u00f3 th\u1ec3 c\u00f3 m\u1ed9t thu\u1eadt to\u00e1n c\u00f3 th\u1ec3 h\u1ecdc nh\u01b0 v\u1eady \u2013 h\u1ecdc gi\u1ed1ng nh\u01b0 ng\u01b0\u1eddi – v\u00ec \u0111\u00f3 l\u00e0 m\u1ed9t c\u00e1ch h\u1ecdc th\u1ef1c t\u1ebf h\u01a1n, ch\u00fang ta c\u00f3 nhi\u1ec1u d\u1eef li\u1ec7u kh\u00f4ng d\u00e1n nh\u00e3n h\u01a1n l\u00e0 nh\u1eefng d\u1eef li\u1ec7u \u0111\u01b0\u1ee3c d\u00e1n nh\u00e3n”. Th\u1eadt ra, h\u1ea7u h\u1ebft nh\u1eefng th\u1ee9 ch\u00fang t\u00f4i g\u1eedi \u0111\u1ebfn nh\u1eefng n\u01a1i nh\u01b0 Facebook v\u00e0 Twitter v\u00e0 Google \u0111\u1ec1u kh\u00f4ng \u0111\u01b0\u1ee3c g\u1eafn nh\u00e3n.<\/p>\n

\u0110\u00e2y l\u00e0 nh\u1eefng g\u00ec m\u00e0 h\u1ecdc s\u00e2u t\u00ecm c\u00e1ch \u0111\u1ea1t \u0111\u01b0\u1ee3c. S\u1eed d\u1ee5ng h\u00e0ng tr\u0103m m\u00e1y \u0111\u1ec3 v\u1eadn h\u00e0nh \u201cc\u00e1c m\u1ea1ng th\u1ea7n kinh\u201d ph\u1ee9c t\u1ea1p – c\u00e1c ki\u1ebfn tr\u00fac (construct) ph\u1ea7n m\u1ec1m c\u00f3 nhi\u1ec7m v\u1ee5 b\u1eaft ch\u01b0\u1edbc c\u00e1c m\u1ea1ng n\u01a1-ron trong n\u00e3o – n\u00f3 cho ph\u00e9p c\u00e1c m\u00e1y h\u1ecdc h\u1ecfi. Trong m\u1ed9t s\u1ed1 tr\u01b0\u1eddng h\u1ee3p, vi\u1ec7c h\u1ecdc \u0111\u00f3 x\u1ea3y ra m\u1ed9t m\u00ecnh m\u00e0 kh\u00f4ng c\u00f3 ai d\u00e1n nh\u00e3n t\u1ea5t c\u1ea3 c\u00e1c d\u1eef li\u1ec7u.<\/p>\n

M\u00e1y t\u00ecm m\u00e8o c\u1ee7a Google l\u00e0 m\u1ed9t v\u00ed d\u1ee5 quan tr\u1ecdng. Th\u1eadt kh\u00f4ng may, ba n\u0103m sau, lo\u1ea1i h\u1ecdc kh\u00f4ng gi\u00e1m s\u00e1t n\u00e0y \u0111\u00e3 kh\u00f4ng th\u1ef1c s\u1ef1 \u0111\u01b0\u1ee3c \u01b0a chu\u1ed9ng, v\u00e0 h\u1ea7u h\u1ebft c\u00e1c h\u1ec7 th\u1ed1ng h\u1ecdc-s\u00e2u th\u01b0\u01a1ng m\u1ea1i v\u1eabn d\u1ef1a v\u00e0o vi\u1ec7c h\u1ecdc c\u00f3 gi\u00e1m s\u00e1t. “M\u1eb7c d\u00f9 chuy\u1ec7n m\u00e8o kh\u00f4ng h\u1eefu \u00edch, nh\u01b0ng trong \u0111\u1ea7u t\u00f4i, m\u1ed9t d\u1ea5u hi\u1ec7u s\u00e1ng ra \u0111ang th\u00fac \u0111\u1ea9y qu\u00e1 tr\u00ecnh h\u1ecdc s\u00e2u \u0111\u1ec3 d\u00e0nh nhi\u1ec1u th\u1eddi gian h\u01a1n cho vi\u1ec7c h\u1ecdc kh\u00f4ng gi\u00e1m s\u00e1t nh\u01b0 m\u1ed9t h\u1ee9a h\u1eb9n cho t\u01b0\u01a1ng lai”, Ng n\u00f3i.<\/p>\n

\u0110i\u1ec1u \u0111\u00f3 c\u00f3 th\u1ec3 h\u1eefu \u00edch cho vi\u1ec7c x\u1eed l\u00fd ng\u00f4n ng\u1eef t\u1ef1 nhi\u00ean, hay NLP (Natural Language Processing), m\u1ed9t ng\u00e0nh AI nh\u1eb1m t\u00ecm ki\u1ebfm \u0111\u1ec3 hi\u1ec3u \u00fd ngh\u0129a \u0111\u1eb1ng sau ng\u00f4n ng\u1eef \u2212 m\u1ed9t nhi\u1ec7m v\u1ee5 kh\u00f3 kh\u0103n h\u01a1n vi\u1ec7c l\u00e0m sao \u0111\u1ec3 cho m\u00e1y m\u00f3c hi\u1ec3u h\u00ecnh \u1ea3nh v\u00e0 c\u00e1c l\u1ec7nh ti\u1ebfng n\u00f3i \u0111\u01a1n gi\u1ea3n.<\/p>\n

S\u1ef1 th\u00e1ch th\u1ee9c c\u1ee7a Ng\u00f4n ng\u1eef<\/b><\/p>\n

M\u1ed9t ph\u1ea7n c\u1ee7a s\u1ef1 th\u00e1ch th\u1ee9c l\u00e0 ng\u00f4n ng\u1eef c\u00f3 th\u1ec3 th\u1ef1c s\u1ef1 tinh t\u1ebf, v\u00e0 ch\u00fang ta v\u1eabn kh\u00f4ng c\u00f3 c\u00e1ch n\u00e0o t\u1ed1t \u0111\u1ec3 d\u1ecbch c\u00e1c kh\u00e1i ni\u1ec7m tinh t\u1ebf. Ch\u1eb3ng h\u1ea1n, c\u00f9ng m\u1ed9t m\u1ed9t t\u1eeb c\u00f3 th\u1ec3 mang c\u00e1c \u00fd ngh\u0129a kh\u00e1c nhau t\u00f9y theo b\u1ed1i c\u1ea3nh, v\u00e0 h\u1ea7u h\u1ebft c\u00e1c h\u1ec7 th\u1ed1ng tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o tr\u00ean \u0111\u1eddi v\u1eabn c\u00f2n g\u1eb7p kh\u00f3 kh\u0103n khi ph\u00e2n bi\u1ec7t gi\u1eefa ch\u00fang. Ch\u00e2m bi\u1ebfm ho\u1eb7c h\u00e0i h\u01b0\u1edbc th\u01b0\u1eddng th\u1ea5t b\u1ea1i ho\u00e0n to\u00e0n \u2212 th\u1eadm ch\u00ed tr\u00ean c\u00e1c m\u00e1y t\u00ednh th\u00f4ng minh nh\u1ea5t.<\/p>\n

“M\u00e1y ho\u1ea1t \u0111\u1ed9ng r\u1ea5t t\u1ed1t v\u1edbi s\u1ed1. Nh\u01b0ng ch\u00fang kh\u00f4ng t\u1ed1t v\u1edbi c\u00e1c bi\u1ec3u t\u01b0\u1ee3ng (symbols)”, Qu\u1ed1c n\u00f3i. V\u00e0 ng\u00f4n ng\u1eef l\u00e0 m\u1ed9t \u0111i\u1ec1u g\u00ec c\u00f3 t\u00ednh bi\u1ec3u t\u01b0\u1ee3ng cao.<\/p>\n

B\u00ed quy\u1ebft l\u00e0 t\u00ecm m\u1ed9t c\u00e1ch \u0111\u1ec3 d\u1ecbch c\u00e1c bi\u1ec3u t\u01b0\u1ee3ng th\u00e0nh s\u1ed1. “L\u00e0m th\u1ebf n\u00e0o \u0111\u1ec3 x\u00e2y d\u1ef1ng m\u1ed9t kh\u00e1i ni\u1ec7m th\u00e0nh m\u1ed9t c\u1ea5u tr\u00fac to\u00e1n h\u1ecdc \u0111\u1ec3 m\u00e1y m\u00f3c c\u00f3 th\u1ec3 x\u1eed l\u00fd n\u00f3 \u2013 \u0111i\u1ec1u \u0111\u00f3 v\u1eabn c\u00f2n ch\u01b0a r\u00f5,” anh n\u00f3i. Nh\u01b0ng, v\u1edbi nh\u1eefng th\u1ee9 nh\u01b0 Word2Vec v\u00e0 c\u00e1c c\u00f4ng ngh\u1ec7 x\u1eed l\u00fd ng\u00f4n ng\u1eef t\u1ef1 nhi\u00ean kh\u00e1c, Qu\u1ed1c v\u00e0 nh\u1eefng ng\u01b0\u1eddi kh\u00e1c \u0111ang s\u1eb5n s\u00e0ng \u0111\u1ec3 l\u00e0m m\u1ed9t ti\u1ebfn b\u1ed9 nh\u1ea5t \u0111\u1ecbnh \u1edf \u0111\u00e2y. V\u00e0 \u0111\u00f3 l\u00e0 nh\u1eefng g\u00ec ch\u00fang ta c\u1ea7n \u0111\u1ec3 x\u00e2y d\u1ef1ng c\u00e1c m\u00e1y c\u00f3 th\u1ec3 th\u1ef1c s\u1ef1 hi\u1ec3u \u0111\u01b0\u1ee3c nh\u1eefng l\u01b0\u1ee3ng th\u00f4ng tin kh\u1ed5ng l\u1ed3 \u0111\u01b0\u1ee3c \u0111\u01b0a l\u00ean m\u1ea1ng v\u1edbi m\u1ed7i gi\u00e2y ph\u00fat tr\u00f4i qua.<\/p>\n

\u201cS\u1ebd kh\u00f4ng th\u1ec3 c\u00f3 c\u00e1ch n\u00e0o \u0111\u1ec3 c\u00f3 s\u1ef1 gi\u00e1m s\u00e1t \u0111\u1ea7y \u0111\u1ee7 cho \u0111i\u1ec1u \u0111\u00f3\u201d, Richard Socher n\u00f3i, ng\u01b0\u1eddi \u0111\u00e3 l\u00e0m nghi\u00ean c\u1ee9u ti\u1ebfn s\u0129 v\u1ec1 h\u1ecdc s\u00e2u b\u00ean c\u1ea1nh Qu\u1ed1c t\u1ea1i Stanford, v\u00e0 ti\u1ebfp t\u1ee5c t\u1eadp trung v\u00e0o NLP. “Hy v\u1ecdng l\u00e0 th\u00f4ng qua s\u1ef1 k\u1ebft h\u1ee3p c\u1ee7a vi\u1ec7c h\u1ecdc kh\u00f4ng gi\u00e1m s\u00e1t v\u00e0 c\u00f3 gi\u00e1m s\u00e1t, b\u1ea1n th\u1ef1c s\u1ef1 c\u00f3 th\u1ec3 c\u00f3 \u0111\u01b0\u1ee3c m\u1ed9t h\u1ec7 th\u1ed1ng s\u1ebd l\u00e0m nh\u1eefng vi\u1ec7c l\u00e0m kinh ng\u1ea1c b\u1ea1n.”<\/p>\n

\u0110o\u1ea1n v\u0103n b\u1eb1ng s\u1ed1<\/b><\/p>\n

Theo nh\u1eefng g\u00ec \u0111\u01b0\u1ee3c bi\u1ebft, chuy\u00ean gia m\u1ea1ng n\u01a1-ron Yoshua Bengio \u0111\u00e3 ph\u00e1t tri\u1ec3n m\u1ed9t thu\u1eadt to\u00e1n th\u00f4ng minh \u0111\u00e3 \u00e1nh x\u1ea1 c\u00e1c t\u1eeb (words) th\u00e0nh vect\u01a1, v\u00e0 \u0111i\u1ec1u n\u00e0y \u0111\u1ee7 th\u00f4ng minh \u0111\u1ec3 gom c\u00e1c t\u1eeb bi\u1ec3u th\u1ecb (represent) c\u00e1c kh\u00e1i ni\u1ec7m t\u01b0\u01a1ng t\u1ef1 th\u00e0nh c\u00e1c nh\u00f3m. V\u00ed d\u1ee5, c\u00e1c t\u1eeb th\u1ee9 Hai ho\u1eb7c th\u1ee9 Ba s\u1ebd c\u00f3 c\u00e1c vect\u01a1 t\u01b0\u01a1ng t\u1ef1 v\u00ec ch\u00fang bi\u1ec3u th\u1ecb c\u00e1c ng\u00e0y trong tu\u1ea7n, trong khi m\u1ed9t c\u00e1i g\u00ec \u0111\u00f3 nh\u01b0 b\u00e1nh sandwich s\u1ebd c\u00f3 m\u1ed9t s\u1ef1 bi\u1ec3u th\u1ecb to\u00e1n h\u1ecdc ho\u00e0n to\u00e0n kh\u00e1c.<\/p>\n

S\u1ef1 ti\u1ebfn h\u00f3a t\u1ef1 nhi\u00ean s\u1ebd bi\u1ec3u th\u1ecb nh\u1eefng \u00fd t\u01b0\u1edfng ph\u1ee9c t\u1ea1p h\u01a1n \u2212 gi\u1ed1ng nh\u01b0 c\u00e2u ho\u1eb7c to\u00e0n b\u1ed9 \u0111o\u1ea1n v\u0103n \u2212 v\u1ec1 m\u1eb7t to\u00e1n h\u1ecdc. \u0110\u1ea7u n\u0103m nay, Qu\u1ed1c c\u00f4ng b\u1ed1 m\u1ed9t b\u00e0i b\u00e1o v\u1edbi m\u1ed9t ng\u01b0\u1eddi kh\u00e1c c\u1ee7a Google, Tomas Mikolov, th\u1ef1c hi\u1ec7n \u0111i\u1ec1u \u0111\u00f3. H\u1ecd m\u00f4 t\u1ea3 “c\u00e1c vect\u01a1 \u0111o\u1ea1n v\u0103n,” c\u00e1c c\u00e1ch to\u00e1n h\u1ecdc \u0111\u1ec3 bi\u1ec3u th\u1ecb cho to\u00e0n b\u1ed9 \u0111o\u1ea1n v\u0103n. Google s\u1ebd kh\u00f4ng n\u00f3i li\u1ec7u n\u00f3 hi\u1ec7n \u0111ang c\u00f3 l\u1ee3i \u00edch t\u1eeb c\u00f4ng vi\u1ec7c n\u00e0y hay kh\u00f4ng \u2212 hay l\u1ee3i th\u1ebf n\u00e0o \u2212 nh\u01b0ng b\u00e0i b\u00e1o \u0111\u1ec1 c\u1eadp \u0111\u1ebfn ph\u00e2n t\u00edch t\u00ecnh c\u1ea3m (sentiment analysis) v\u00e0 L\u00ea th\u1eeba nh\u1eadn \u0111\u00e2y c\u00f3 th\u1ec3 l\u00e0 m\u1ed9t trong nh\u1eefng \u1ee9ng d\u1ee5ng c\u1ee7a n\u00f3. N\u00f3i c\u00e1ch kh\u00e1c, n\u00f3 t\u00ecm c\u00e1ch hi\u1ec3u c\u00e1c quan \u0111i\u1ec3m – th\u1eadm ch\u00ed c\u1ea3 c\u1ea3m x\u00fac.<\/p>\n

Th\u1eadm ch\u00ed g\u1ea7n \u0111\u00e2y, Qu\u1ed1c \u0111\u00e3 c\u00f4ng b\u1ed1 m\u1ed9t b\u00e0i b\u00e1o v\u1edbi c\u00e1c nh\u00e2n vi\u00ean Google, Ilya Sutskever v\u00e0 Oriol Vinyals, v\u1ec1 m\u00e1y d\u1ecbch b\u1eb1ng c\u00e1ch s\u1eed d\u1ee5ng c\u00e1c m\u1ea1ng th\u1ea7n kinh s\u00e2u. N\u00f3 \u0111\u01b0\u1ee3c tr\u00ecnh b\u00e0y t\u1ea1i h\u1ed9i ngh\u1ecb Neural Information Processing Systems Foundation \u1edf Montreal v\u00e0o tu\u1ea7n t\u1edbi.<\/p>\n

H\u1ec7 th\u1ed1ng n\u00e0y s\u1eed d\u1ee5ng m\u1ed9t c\u00e1i g\u00ec \u0111\u00f3 g\u1ecdi l\u00e0 “m\u1ea1ng th\u1ea7n kinh l\u1eb7p l\u1ea1i” (recurrent neural networks). B\u1edfi v\u00ec ch\u00fang ph\u1ee9c t\u1ea1p h\u01a1n n\u00ean ch\u00fang c\u00f3 th\u1ec3 kh\u00f3 \u0111\u00e0o t\u1ea1o h\u01a1n so v\u1edbi c\u00e1c m\u1ea1ng th\u1ea7n kinh t\u00edch ch\u1eadp (convolutional neural networks) \u0111\u01b0\u1ee3c s\u1eed d\u1ee5ng r\u1ed9ng r\u00e3i h\u01a1n. Nh\u01b0ng l\u1ee3i th\u1ebf l\u1edbn nh\u1ea5t c\u1ee7a ch\u00fang l\u00e0 ch\u00fang c\u00f3 m\u1ed9t b\u1ed9 nh\u1edb t\u00edch h\u1ee3p, m\u1ed9t \u0111\u1eb7c \u0111i\u1ec3m l\u00e0m cho ch\u00fang \u0111\u1eb7c bi\u1ec7t h\u1eefu \u00edch cho nh\u1eefng th\u1ee9 c\u00f3 t\u00ednh chu\u1ed7i (sequential), nh\u01b0 ng\u00f4n ng\u1eef. “R\u1ea5t nhi\u1ec1u ng\u01b0\u1eddi trong ch\u00fang ta quan t\u00e2m \u0111\u1ebfn vi\u1ec7c l\u00e0m th\u1ebf n\u00e0o \u0111\u1ec3 \u00e1nh x\u1ea1 m\u1ed9t chu\u1ed7i d\u00e0i h\u01a1n th\u00e0nh m\u1ed9t v\u00e9ct\u01a1. Vect\u01a1 \u0111o\u1ea1n v\u0103n l\u00e0 m\u1ed9t c\u00e1ch ti\u1ebfp c\u1eadn, nh\u01b0ng \u0111i\u1ec1u n\u00e0y c\u00f3 r\u1ea5t nhi\u1ec1u l\u1ee3i th\u1ebf “, Vi\u1ec7t Qu\u1ed1c n\u00f3i. “N\u00f3 c\u00f3 th\u1ec3 nh\u1edb th\u1ee9 t\u1ef1 c\u00e1c t\u1eeb trong chu\u1ed7i … N\u00f3 c\u00f3 th\u1ec3 x\u1eed l\u00fd c\u00e1c \u0111\u1ea7u v\u00e0o c\u00f3 k\u00edch th\u01b0\u1edbc thay \u0111\u1ed5i.”<\/p>\n

M\u1ed9t b\u1ed9 N\u00e3o l\u1edbn h\u01a1n nhi\u1ec1u<\/b><\/p>\n

Theo b\u00e0i b\u00e1o, c\u00e1ch ti\u1ebfp c\u1eadn m\u1edbi t\u1ed1t h\u01a1n c\u00e1c thu\u1eadt to\u00e1n m\u00e1y d\u1ecbch kh\u00e1c, nh\u01b0ng \u0111\u00e2y ch\u1ec9 l\u00e0 m\u1ed9t trong nh\u1eefng \u1ee9ng d\u1ee5ng c\u00f3 th\u1ec3 c\u00f3 c\u1ee7a n\u00f3. Qu\u1ed1c n\u00f3i r\u1eb1ng c\u00f4ng tr\u00ecnh m\u1edbi n\u00e0y c\u0169ng c\u00f3 th\u1ec3 h\u1eefu \u00edch cho vi\u1ec7c tr\u1ea3 l\u1eddi c\u00e1c c\u00e2u h\u1ecfi tr\u00ean web, t\u1ef1 \u0111\u1ed9ng h\u00f3a ph\u1ee5 \u0111\u1ec1 v\u00e0 c\u00f3 th\u1ec3 ph\u00e2n t\u00edch t\u00ecnh c\u1ea3m \u2212 m\u1eb7c d\u00f9 anh cho bi\u1ebft anh v\u1eabn ch\u01b0a th\u1eed.<\/p>\n

\u0110\u1ec3 t\u1eadn d\u1ee5ng \u0111\u1ea7y \u0111\u1ee7 c\u00e1c lo\u1ea1i thu\u1eadt to\u00e1n n\u00e0y, Google s\u1ebd ph\u1ea3i x\u00e2y d\u1ef1ng m\u1ed9t s\u1ed1 m\u1ea1ng n\u01a1-ron th\u1ef1c s\u1ef1 l\u1edbn \u2212 l\u1edbn h\u01a1n nhi\u1ec1u so v\u1edbi nh\u1eefng g\u00ec n\u00f3 \u0111\u00e3 x\u00e2y d\u1ef1ng cho nh\u00ecn v\u00e0 n\u00f3i. Geoff Hinton, nh\u00e2n v\u1eadt trung t\u00e2m c\u1ee7a phong tr\u00e0o h\u1ecdc t\u1eadp s\u00e2u, hi\u1ec7n \u0111ang l\u00e0m vi\u1ec7c t\u1ea1i Google, s\u1eed d\u1ee5ng m\u1ed9t s\u1ef1 so s\u00e1nh th\u00edch h\u1ee3p. V\u1edbi n\u00e3o b\u1ed9 k\u00edch th\u01b0\u1edbc c\u1ee7a m\u1ed9t con chim b\u1ed3 c\u00e2u, \u00f4ng n\u00f3i v\u1edbi WIRED n\u0103m ngo\u00e1i, “b\u1ea1n c\u00f3 th\u1ec3 c\u00f3 m\u1ecdi l\u00fd do \u0111\u1ec3 tin r\u1eb1ng b\u1ea1n c\u00f3 th\u1ec3 l\u00e0m m\u1ed9t t\u1ea7m nh\u00ecn t\u1ed1t. Nh\u01b0ng b\u1ea1n kh\u00f4ng mu\u1ed1n tr\u00f2 chuy\u1ec7n v\u1edbi m\u1ed9t con chim b\u1ed3 c\u00e2u. “<\/p>\n

Nh\u01b0 m\u1ed9t th\u1ef1c t\u1ebf, m\u1ed9t n\u00e3o b\u1ed9 b\u1ed3 c\u00e2u v\u1eabn c\u00f3 th\u1ec3 ho\u1ea1t \u0111\u1ed9ng t\u1ed1t h\u01a1n ngay c\u1ea3 nh\u1eefng h\u1ec7 th\u1ed1ng th\u1ea7n kinh ti\u00ean ti\u1ebfn nh\u1ea5t th\u1ebf gi\u1edbi, g\u1ed3m c\u1ea3 Google Brain, v\u1ec1 m\u1ed9t s\u1ed1 nhi\u1ec7m v\u1ee5. Tuy nhi\u00ean, Hinton \u0111\u00e3 gia nh\u1eadp Google, \u00f4ng n\u00f3i, \u0111\u1ec3 x\u00e2y d\u1ef1ng m\u1ed9t trong nh\u1eefng m\u1ea1ng n\u01a1-ron l\u1edbn nh\u1ea5t cho t\u1edbi nay, m\u1ed9t trong s\u1ed1 \u0111\u00f3 c\u00f3 th\u1ec3 “l\u00e0m l\u00fd t\u00ednh b\u00ecnh th\u01b0\u1eddng r\u1ea5t t\u1ed1t “.<\/p>\n

V\u1edbi s\u1ef1 gi\u00fap \u0111\u1ee1 c\u1ee7a nh\u1eefng ng\u01b0\u1eddi nh\u01b0 L\u00ea Vi\u1ebft Qu\u1ed1c, \u0111i\u1ec1u \u0111\u00f3 c\u00f3 th\u1ec3 x\u1ea3y ra.\/.<\/p>\n

\n

Ch\u00fa gi\u1ea3i:<\/p>\n

    \n
  1. B\u1ea3n g\u1ed1c: https:\/\/www.wired.com\/2014\/12\/googlers-quest-teach-machines-understand-emotions\/<\/span><\/li>\n
  2. Mikolov b\u00e2y gi\u1edd \u0111ang l\u00e0m vi\u1ec7c t\u1ea1i Facebook<\/span><\/li>\n
  3. C\u1eadp nh\u1eadt 4:46 EDT 12\/05\/14: C\u00e2u chuy\u1ec7n n\u00e0y \u0111\u00e3 \u0111\u01b0\u1ee3c c\u1eadp nh\u1eadt \u0111\u1ec3 l\u00e0m r\u00f5 vai tr\u00f2 c\u1ee7a L\u00ea Vi\u1ebft Qu\u1ed1c trong c\u00f4ng ngh\u1ec7 h\u1ecdc s\u00e2u gi\u00fap nu\u00f4i d\u01b0\u1ee1ng “\u0111\u1ed3 th\u1ecb tri th\u1ee9c” (Knowledge Graph) c\u1ee7a Google<\/span><\/li>\n<\/ol>\n","protected":false},"excerpt":{"rendered":"

      M\u1ed8T GOOGLER VI\u1ec6T NAM T\u00ccM C\u00c1CH D\u1ea0Y M\u00c1Y HI\u1ec2U\u00a0C\u1ea2M X\u00daC C\u1ee6A CON NG\u01af\u1edcI: \u2014 TI\u1ebeN S\u0128 L\u00ca VI\u1ebeT QU\u1ed0C \u2014 T\u00e1c gi\u1ea3: Daniela Hernandez (12.02.14) (Ngu\u1ed3n: WIRED)1 Ng\u01b0\u1eddi d\u1ecbch: V\u00f5 Th\u00e0nh Minh Tu\u1ec7 & Nguy\u1ec5n Xu\u00e2n Xanh L\u1eddi n\u00f3i \u0111\u1ea7u c\u1ee7a ng\u01b0\u1eddi d\u1ecbch. L\u00ea Vi\u1ebft Qu\u1ed1c m\u00e0 b\u00e0i b\u00e1o d\u01b0\u1edbi \u0111\u00e2y n\u00f3i v\u1ec1 anh<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_mi_skip_tracking":false,"jetpack_post_was_ever_published":false},"categories":[1],"tags":[55],"jetpack_featured_media_url":"","_links":{"self":[{"href":"https:\/\/rosetta.vn\/nguyenxuanxanh\/wp-json\/wp\/v2\/posts\/686"}],"collection":[{"href":"https:\/\/rosetta.vn\/nguyenxuanxanh\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/rosetta.vn\/nguyenxuanxanh\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/rosetta.vn\/nguyenxuanxanh\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/rosetta.vn\/nguyenxuanxanh\/wp-json\/wp\/v2\/comments?post=686"}],"version-history":[{"count":7,"href":"https:\/\/rosetta.vn\/nguyenxuanxanh\/wp-json\/wp\/v2\/posts\/686\/revisions"}],"predecessor-version":[{"id":11918,"href":"https:\/\/rosetta.vn\/nguyenxuanxanh\/wp-json\/wp\/v2\/posts\/686\/revisions\/11918"}],"wp:attachment":[{"href":"https:\/\/rosetta.vn\/nguyenxuanxanh\/wp-json\/wp\/v2\/media?parent=686"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/rosetta.vn\/nguyenxuanxanh\/wp-json\/wp\/v2\/categories?post=686"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/rosetta.vn\/nguyenxuanxanh\/wp-json\/wp\/v2\/tags?post=686"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}