{"id":699,"date":"2023-10-21T17:08:47","date_gmt":"2023-10-21T09:08:47","guid":{"rendered":"https:\/\/www.1ai.net\/?p=699"},"modified":"2023-10-21T17:08:47","modified_gmt":"2023-10-21T09:08:47","slug":"%e4%ba%ba%e5%b7%a5%e6%99%ba%e8%83%bd%e6%a8%a1%e5%9e%8b%e9%80%8f%e6%98%8e%e5%ba%a6%e8%af%84%e4%bc%b0%ef%bc%9allama-2%e4%bd%8d%e5%88%97%e7%ac%ac%e4%b8%80%ef%bc%8cgpt-4%e9%80%8f%e6%98%8e%e5%ba%a6","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/699.html","title":{"rendered":"AI model transparency assessment: Llama 2 ranks first, GPT-4 transparency is poor"},"content":{"rendered":"

\u8fd1\u5e74\u6765\uff0c\u4eba\u5de5\u667a\u80fd\u9886\u57df\u7684\u4e3b\u6d41\u6a21\u578b\u900f\u660e\u5ea6\u6210\u4e3a\u7126\u70b9\uff0c\u65af\u5766\u798f\u5927\u5b66\u3001MIT\u3001\u666e\u6797\u65af\u987f\u5927\u5b66\u7b49\u673a\u6784\u5408\u4f5c\u63d0\u51fa\u4e86\u201c\u57fa\u7840\u6a21\u578b\u900f\u660e\u5ea6\u6307\u6570\u201d\u6765\u8bc4\u4f30\u5341\u5927\u4e3b\u6d41AI\u6a21\u578b<\/a>\u7684\u900f\u660e\u5ea6\u3002\u7ed3\u679c\u663e\u793a\uff0cLlama<\/a>2\u4f4d\u5217\u7b2c\u4e00<\/span>\uff0c\u800cGPT-4\u7b49\u6a21\u578b\u7684\u900f\u660e\u5ea6\u8f83\u4f4e\u3002<\/p>\n

\u5c3d\u7ba1AI\u6a21\u578b\u7684\u793e\u4f1a\u5f71\u54cd\u65e5\u76ca\u589e\u5927\uff0c\u4eba\u4eec\u5bf9\u4e8e\u8fd9\u4e9b\u6a21\u578b\u7684\u6784\u5efa\u3001\u8bad\u7ec3\u548c\u4f7f\u7528\u65b9\u5f0f\u4ecd\u6709\u8bf8\u591a\u7591\u95ee\uff0c\u5305\u62ec\u6570\u636e\u6765\u6e90\u3001\u52b3\u52a8\u529b\u5f85\u9047\u7b49\u3002\u7136\u800c\uff0c\u8be5\u8bc4\u4f30\u7cfb\u7edf\u4e5f\u5f15\u53d1\u4e86\u4e00\u4e9b\u4e89\u8bae\uff0c\u4e00\u4e9b\u5f00\u53d1\u8005\u8ba4\u4e3a\u8981\u6c42\u516c\u53f8\u516c\u5f00\u5546\u4e1a\u673a\u5bc6\u8fc7\u4e8e\u5e7c\u7a1a\u3002<\/p>\n

\"\"<\/p>\n

\u8bba\u6587\u5730\u5740:https:\/\/arxiv.org\/pdf\/2310.12941.pdf<\/p>\n

\u5c3d\u7ba1\u5982\u6b64\uff0c\u900f\u660e\u5ea6\u5bf9\u4e8eAI\u6a21\u578b\u7684\u53d1\u5c55\u548c\u5e94\u7528\u81f3\u5173\u91cd\u8981\uff0c\u5c24\u5176\u662f\u5728\u751f\u6210\u5f0fAI\u9886\u57df\uff0c\u56e0\u4e3a\u6a21\u578b\u65e2\u6709\u63d0\u9ad8\u751f\u4ea7\u529b\u7684\u6f5c\u529b\uff0c\u4e5f\u53ef\u80fd\u88ab\u7528\u6765\u4f24\u5bb3\u4ed6\u4eba\u3002\u900f\u660e\u5ea6\u7684\u4e0d\u8db3\u53ef\u80fd\u5bfc\u81f4\u6a21\u578b\u88ab\u6ee5\u7528\uff0c\u56e0\u6b64\u5f00\u53d1\u8005\u9700\u8981\u66f4\u52a0\u6ce8\u91cd\u900f\u660e\u5ea6\uff0c\u5305\u62ec\u5bf9\u6a21\u578b\u7684\u6784\u5efa\u3001\u529f\u80fd\u3001\u98ce\u9669\u7b49\u65b9\u9762\u7684\u516c\u5f00\u3002<\/p>\n

\u7136\u800c\uff0c\u76ee\u524d\u5927\u591a\u6570\u4e3b\u8981\u57fa\u7840\u6a21\u578b\u5f00\u53d1\u8005\u90fd\u672a\u80fd\u63d0\u4f9b\u8db3\u591f\u7684\u900f\u660e\u5ea6\uff0c\u8fd9\u51f8\u663e\u4e86\u4eba\u5de5\u667a\u80fd\u884c\u4e1a\u4e9f\u9700\u6539\u8fdb\u900f\u660e\u5ea6\u7684\u95ee\u9898\u3002\u540c\u65f6\uff0c\u5f00\u6e90\u57fa\u7840\u6a21\u578b\u5982Llama2\u3001BLOOMZ\u7b49\u83b7\u5f97\u8f83\u9ad8\u8bc4\u5206\uff0c\u4f46\u4ecd\u6709\u6539\u8fdb\u7a7a\u95f4\uff0c\u56e0\u4e3a\u53ea\u6709\u6781\u5c11\u6570\u5f00\u53d1\u8005\u900f\u660e\u5730\u5c55\u793a\u4e86\u6a21\u578b\u7684\u5c40\u9650\u6027\u548c\u5176\u4ed6\u5173\u952e\u4fe1\u606f\u3002<\/p>\n

\u5728\u5f53\u524d\u7684\u653f\u7b56\u4e89\u8bba\u4e2d\uff0c\u662f\u5426\u5e94\u8be5\u5c06\u4eba\u5de5\u667a\u80fd\u6a21\u578b\u5f00\u6e90\u6210\u4e3a\u4e00\u4e2a\u4e89\u8bae\u7126\u70b9\uff0c\u4f46\u65e0\u8bba\u5f00\u6e90\u8fd8\u662f\u95ed\u6e90\uff0c\u900f\u660e\u5ea6\u90fd\u662f\u5173\u952e\u56e0\u7d20\uff0c\u4ee5\u786e\u4fddAI\u6a21\u578b\u7684\u8d1f\u9762\u5f71\u54cd\u5f97\u5230\u63a7\u5236\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"

In recent years, the focus has been on the transparency of mainstream models in the area of artificial intelligence, and the \u201cBasic Model Transparency Index\u201d has been developed by institutions such as Stanford University, MIT and Princeton University to assess the transparency of the 10 main mainstream AI models. The results showed that Llama 2 ranked first, while models such as GPT-4 were less transparent. Despite the growing social impact of AI models, there are many questions about how they are constructed, trained and used, including data sources, labour entitlements, etc. However, the assessment system has also given rise to some controversy, with some developers finding it too naive to require companies to disclose business secrets. Address: https:\/\/arxiv.org\/pdf\/2310.1241.pdf<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[146],"tags":[167,184],"collection":[],"class_list":["post-699","post","type-post","status-publish","format-standard","hentry","category-news","tag-ai","tag-llama"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/699","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/comments?post=699"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/699\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=699"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=699"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=699"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=699"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}