{"id":2297,"date":"2023-01-05T12:09:08","date_gmt":"2023-01-05T12:09:08","guid":{"rendered":"https:\/\/cafe2sach.com\/?p=2297"},"modified":"2023-01-05T12:09:08","modified_gmt":"2023-01-05T12:09:08","slug":"chatgpt-va-quan-diem-cua-andrew","status":"publish","type":"post","link":"https:\/\/cafe2sach.com\/index.php\/2023\/01\/05\/chatgpt-va-quan-diem-cua-andrew\/","title":{"rendered":"ChatGPT v\u00e0 quan \u0111i\u1ec3m c\u1ee7a Andrew"},"content":{"rendered":"\n

M\u00ecnh s\u1ebd tr\u00edch d\u1eabn \u1edf \u0111\u00e2y quan \u0111i\u1ec3m c\u1ee7a th\u1ea7y Andrew v\u1ec1 ChatGPT c\u00f4ng c\u1ee5 h\u1ecdc m\u00e1y \u0111ang \u0111\u01b0\u1ee3c cho l\u00e0 tr\u1ea5n \u0111\u1ed9ng gi\u1edbi nghi\u00ean c\u1ee9u AI g\u1ea7n \u0111\u00e2y. <\/p>\n\n\n\n

Playing with ChatGPT, the latest language model from OpenAI, I found it to be an impressive advance from its predecessor GPT-3. Occasionally it says it can\u2019t answer a question. This is a great step! But, like other LLMs, it can be\u00a0hilariously wrong<\/a>. Work lies ahead to build systems that can express different degrees of confidence.<\/em><\/p>\n\n\n\n

For example, a model like Meta\u2019s Atlas<\/a> or DeepMind\u2019s RETRO<\/a> that synthesizes multiple articles into one answer might infer a degree of confidence based on the reputations of the sources it draws from and the agreement among them, and then change its communication style accordingly. Pure LLMs and other architectures may need other solutions.<\/em><\/p>\n\n\n\n

If we can get generative algorithms to express doubt when they\u2019re not sure they\u2019re right, it will go a long way toward building trust and ameliorating the risk of generating misinformation.<\/em>

Keep learning!<\/em><\/p>\n\n\n\n

Andrew<\/em><\/p>\n\n\n\n

Khi d\u00f9ng ChatGPT, m\u00f4 h\u00ecnh h\u1ecdc m\u00e1y m\u1edbi nh\u1ea5t c\u1ee7a OpenAI, t\u00f4i th\u1ea5y \u0111\u00e2y l\u00e0 m\u1ed9t b\u01b0\u1edbc ti\u1ebfn \u1ea5n t\u01b0\u1ee3ng so v\u1edbi ng\u01b0\u1eddi ti\u1ec1n nhi\u1ec7m GPT-3. \u0110\u00f4i khi n\u00f3 n\u00f3i r\u1eb1ng n\u00f3 kh\u00f4ng th\u1ec3 tr\u1ea3 l\u1eddi m\u1ed9t c\u00e2u h\u1ecfi. \u0110\u00e2y l\u00e0 m\u1ed9t b\u01b0\u1edbc tuy\u1ec7t v\u1eddi! Tuy nhi\u00ean, gi\u1ed1ng nh\u01b0 c\u00e1c LLM kh\u00e1c, n\u00f3 c\u00f3 th\u1ec3 sai m\u1ed9t c\u00e1ch bu\u1ed3n c\u01b0\u1eddi ho\u1eb7c ng\u1edb ng\u1ea9n.\n\n\"\"\n\n\nC\u00f4ng vi\u1ec7c c\u00f2n \u1edf ph\u00eda tr\u01b0\u1edbc l\u00e0 x\u00e2y d\u1ef1ng c\u00e1c h\u1ec7 th\u1ed1ng c\u00f3 th\u1ec3 th\u1ec3 hi\u1ec7n c\u00e1c m\u1ee9c \u0111\u1ed9 tin c\u1eady kh\u00e1c nhau.\n\nV\u00ed d\u1ee5: m\u1ed9t m\u00f4 h\u00ecnh nh\u01b0\u00a0Atlas\u00a0c\u1ee7a Meta ho\u1eb7c\u00a0RETRO\u00a0c\u1ee7a DeepMind t\u1ed5ng h\u1ee3p nhi\u1ec1u b\u00e0i vi\u1ebft th\u00e0nh m\u1ed9t c\u00e2u tr\u1ea3 l\u1eddi c\u00f3 th\u1ec3 suy ra m\u1ee9c \u0111\u1ed9 tin c\u1eady d\u1ef1a tr\u00ean danh ti\u1ebfng c\u1ee7a c\u00e1c ngu\u1ed3n m\u00e0 n\u00f3 r\u00fat ra v\u00e0 s\u1ef1 \u0111\u1ed3ng thu\u1eadn gi\u1eefa c\u00e1c ngu\u1ed3n \u0111\u00f3, sau \u0111\u00f3 thay \u0111\u1ed5i phong c\u00e1ch giao ti\u1ebfp c\u1ee7a n\u00f3 cho ph\u00f9 h\u1ee3p. LLM thu\u1ea7n t\u00fay v\u00e0 c\u00e1c ki\u1ebfn \u200b\u200btr\u00fac kh\u00e1c c\u00f3 th\u1ec3 c\u1ea7n c\u00e1c gi\u1ea3i ph\u00e1p kh\u00e1c.\n\nN\u1ebfu ch\u00fang ta c\u00f3 th\u1ec3 s\u1eed d\u1ee5ng c\u00e1c thu\u1eadt to\u00e1n t\u1ed5ng qu\u00e1t \u0111\u1ec3 b\u00e0y t\u1ecf s\u1ef1 nghi ng\u1edd khi ch\u00fang kh\u00f4ng ch\u1eafc m\u00ecnh \u0111\u00fang, th\u00ec \u0111\u00f3 s\u1ebd l\u00e0 m\u1ed9t ch\u1eb7ng \u0111\u01b0\u1eddng d\u00e0i h\u01b0\u1edbng t\u1edbi vi\u1ec7c x\u00e2y d\u1ef1ng l\u00f2ng tin v\u00e0 gi\u1ea3m thi\u1ec3u nguy c\u01a1 t\u1ea1o ra th\u00f4ng tin sai l\u1ec7ch.\n\nTi\u1ebfp t\u1ee5c h\u1ecdc t\u1eadp!\n\nAndrew<\/pre>\n","protected":false},"excerpt":{"rendered":"
M\u00ecnh s\u1ebd tr\u00edch d\u1eabn \u1edf \u0111\u00e2y quan \u0111i\u1ec3m c\u1ee7a th\u1ea7y Andrew v\u1ec1 ChatGPT c\u00f4ng c\u1ee5 h\u1ecdc m\u00e1y \u0111ang \u0111\u01b0\u1ee3c cho […]<\/p>\n","protected":false},"author":1,"featured_media":1425,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[250,565,111,233],"tags":[],"class_list":["post-2297","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-deep-learning","category-he-thong-khuyen-nghi-ai","category-tin-tuc-4-0","category-tri-tue-nhan-tao"],"_links":{"self":[{"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/posts\/2297","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/comments?post=2297"}],"version-history":[{"count":1,"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/posts\/2297\/revisions"}],"predecessor-version":[{"id":2299,"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/posts\/2297\/revisions\/2299"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/media\/1425"}],"wp:attachment":[{"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/media?parent=2297"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/categories?post=2297"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/tags?post=2297"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}