{"id":825,"date":"2023-05-30T13:53:45","date_gmt":"2023-05-30T05:53:45","guid":{"rendered":"http:\/\/47.92.147.95\/?p=825"},"modified":"2023-06-01T14:35:03","modified_gmt":"2023-06-01T06:35:03","slug":"2-2-1-3-6-%e6%99%ba%e8%83%bd%e4%b8%8e%e8%ae%a1%e7%ae%97%e4%bb%a5%e5%8f%8allm%e7%9a%84%e5%85%b6%e4%bb%96%e7%bc%ba%e9%99%b7","status":"publish","type":"post","link":"http:\/\/47.92.147.95\/index.php\/2023\/05\/30\/825\/13\/","title":{"rendered":"2.2.1.3.6 \u201c\u667a\u80fd\u4e0e\u8ba1\u7b97\u201d\u4ee5\u53caLLM\u7684\u5176\u4ed6\u7f3a\u9677"},"content":{"rendered":"\n

Stuart Russell\uff1a\u76ee\u524d\u7684\u4eba\u5de5\u667a\u80fd\u7cfb\u7edf\u6ca1\u6709\u901a\u7528\u6027\uff0c\u56e0\u4e3a\u7535\u8def\u65e0\u6cd5\u5f88\u597d\u5730\u6355\u6349\u901a\u7528\u6027\u3002\u6211\u4eec\u5df2\u7ecf\u5728\u5927\u578b\u8bed\u8a00\u6a21\u578b\u5f88\u96be\u5b66\u4e60\u7b97\u672f\u7684\u57fa\u672c\u89c4\u5219<\/strong>\u4e2d\u770b\u5230\u4e86\u8fd9\u4e00\u70b9\u3002\u5c3d\u7ba1\u6709\u6570\u767e\u4e07\u4e2a\u4f8b\u5b50\uff0c\u4e4b\u524d\u51fb\u8d25\u4eba\u7c7b\u7684\u4eba\u5de5\u667a\u80fd\u56f4\u68cb\u7cfb\u7edf\u8fd8\u662f\u65e0\u6cd5\u6b63\u786e\u7406\u89e3 \u201c\u5927\u9f99\u201d \u548c \u201c\u6b7b\u6d3b\u201d \u7684\u6982\u5ff5\uff0c\u4f5c\u4e3a\u4e1a\u4f59\u56f4\u68cb\u9009\u624b\u7684\u7814\u7a76\u5458\u5f00\u53d1\u51fa\u4e86\u4e00\u5957\u7b56\u7565\uff0c\u6218\u80dc\u4e86\u56f4\u68cb\u7a0b\u5e8f<\/em>\uff08\u201c20230223 \u7edd\u827a\u201d\uff09\u3002<\/p>\n\n\n\n


<\/p>\n\n\n\n

\u300aSparks of Artificial General Intelligence: Early experiments with GPT-4\u300b<\/p>\n\n\n\n

8.2 Lack of planning in arithmetic\/reasoning problems<\/p>\n\n\n\n

GPT-4\uff1a7 * 4 + 8 * 8 = 88<\/p>\n\n\n\n

\u4f5c\u8005\u7528 100 \u4e2a\u968f\u673a\u6837\u672c\u5bf9\u6a21\u578b\u8fdb\u884c\u4e86\u6d4b\u8bd5\uff0c\u5728 0 \u5230 9 \u4e4b\u95f4\u5747\u5300\u751f\u6210\u56db\u4e2a\u6570\u5b57\uff0c\u5f97\u5230\u7684\u51c6\u786e\u7387\u53ea\u6709 58%\u3002\u8fd9\u53ea\u6d89\u53ca\u5230\u4e2a\u4f4d\u6570\u7684\u4e58\u6cd5\u548c\u4e24\u4f4d\u6570\u5b57\u7684\u52a0\u6cd5\uff0c\u4e00\u4e2a\u5177\u5907\u57fa\u672c\u6570\u5b66\u77e5\u8bc6\u7684\u5c0f\u5b66\u751f\u5c31\u80fd\u89e3\u51b3\u8fd9\u4e2a\u95ee\u9898\uff0c\u5f53\u6570\u5b57\u5728 10 ~ 19 \u4e4b\u95f4\u548c 20 ~ 39 \u4e4b\u95f4\u5747\u5300\u9009\u62e9\u65f6\uff0c\u51c6\u786e\u5ea6\u5206\u522b\u4e0b\u964d\u5230 16% \u548c 12%\uff0c\u5f53\u6570\u5b57\u5728 99 \u548c 199 \u4e4b\u95f4\u65f6\uff0c\u51c6\u786e\u5ea6\u4e0b\u964d\u5230\u96f6\u3002<\/em><\/p>\n\n\n\n


<\/p>\n\n\n\n

\u300a\u54e5\u5fb7\u5c14\u3001\u57c3\u820d\u5c14\u3001\u5df4\u8d6b\uff1a\u96c6\u5f02\u74a7\u4e4b\u5927\u6210 G\u00f6del, Escher, Bach: An Eternal Golden Braid\u300bHofstadter<\/p>\n\n\n\n

CHAPTER XIX Artificial Intelligence: Prospects<\/p>\n\n\n\n

Ten Questions and Speculations<\/p>\n\n\n\n

Question: Will a thinking computer be able to add fast?\uff08\u667a\u80fd\u8ba1\u7b97\u673a\u662f\u5426\u80fd\u66f4\u5feb\u5730\u505a\u52a0\u6cd5\uff1f<\/em>\uff09<\/p>\n\n\n\n

Speculation: Perhaps not. We ourselves are composed of hardware which does fancy calculations but that doesn’t mean that our symbol level, where “we” are, knows how to carry out the same fancy calculations. Let me put it this way: there’s no way that you can load numbers into your own neurons to add up your grocery bill. Luckily for you, your symbol level (i.e., you) can’t gain access to the neurons which are doing your thinking-otherwise you’d get addle-brained. To paraphrase Descartes again:<\/em><\/em><\/p>\n\n\n\n

“<\/em>I think; therefore I have no access to the level where I sum.<\/em><\/strong><\/em><\/strong>” \uff08<\/em>\u6211\u601d\uff0c\u6545\u6211\u65e0\u6cd5\u8fdb\u5165\u6211\u7b97\u7684\u5c42\u6b21<\/em><\/strong><\/em><\/strong>\uff09<\/em><\/em><\/p>\n\n\n\n

Why should it not be the same for an intelligent program? It mustn’t be allowed to gain access to the circuits which are doing its thinking otherwise it’ll get addle-CPU’d. Quite seriously, a machine that can pass the Turing test may well add as slowly as you or I do, and for similar reasons. It will represent the number 2 not just by the two bits “10”, but as a full-fledged concept the way we do, replete with associations such as its homonyms “too” and “to”, the words “couple” and “deuce”, a host of mental images such as dots on dominos, the shape of the numeral ‘2’, the notions of alternation, evenness,<\/em><\/em><\/p>\n\n\n\n

oddness, and on and on … With all this “extra baggage” to carry around, an intelligent program will become quite slothful in its adding. <\/em>Of course, we could give it a ‘ pocket calculator , so to speak (or build one in). Then it could answer very fast, but its performance would be just like that of a person with a pocket calculator. <\/em><\/u><\/em>There would be two separate parts to the machine: a reliable but mindless part and an intelligent but fallible part. You couldn’t rely on the composite system to be reliable, any more than a composite of person and machine is necessarily reliable. So if it’s right answers you’re after, better stick to the pocket calculator alone-don’t throw in the intelligence!<\/em><\/em><\/p>\n\n\n\n


\u6269\u5c55\u4e00\u4e0b\uff1a\u6cdb\u5316\u80fd\u529b\/\u9c81\u68d2\u6027\u4e0e\u51fa\u9519\/\u5e7b\u89c9\/\u80e1\u8bdd\u3001\u539f\u7406\u9ed1\u7bb1\u662f\u901a\u7528\u667a\u80fd\u7684\u4e00\u4f53\u4e24\u9762\uff082.2.1.1.3 \u5bb9\u9519\u3001\u81ea\u4fee\u590d\u3001\u9c81\u68d2\u6027<\/a>\uff09\u3002<\/p>\n\n\n\n

\u6709\u4e9b\u573a\u666f\u4e2d\uff0c\u5927\u6a21\u578b\u7684\u4e00\u4e9b\u73b0\u6709\u7f3a\u9677\u5bf9\u5546\u7528\u7684\u5f71\u54cd\u6ca1\u90a3\u4e48\u5927\uff0c\u751a\u81f3\u53ef\u80fd\u662f\u52a9\u76ca\u3002\u5982\u5f3a\u8c03\u4e2a\u6027\u5316\u548c\u8da3\u5473\u6027\u7684\u804a\u5929\u5e94\u7528 character.ai \u521b\u59cb\u4eba\u8bf4\uff1a\u201c\u6211\u5e76\u4e0d\u8ba4\u4e3a\u5e7b\u89c9\u662f\u9700\u8981\u89e3\u51b3\u7684\u95ee\u9898\uff0c\u6211\u751a\u81f3\u5f88\u559c\u6b22\u5b83\uff0c\u8fd9\u662f\u6a21\u578b\u6709\u8da3\u7684\u7279\u70b9\u3002\u201d \u5728  character.ai \u505a\u7684 \u201c\u89d2\u8272\u626e\u6f14\u201d \u804a\u5929\u573a\u666f\u4e2d\uff0c\u5e7b\u89c9\u662f\u60f3\u8c61\u529b\u7684\u6e90\u6cc9\u3002\u4f46\u5bf9\u53e6\u4e00\u4e9b\u5bb9\u9519\u5f88\u4f4e\u7684\u884c\u4e1a\uff0c\u5982\u533b\u7597\u8bca\u65ad\u3001\u81ea\u52a8\u9a7e\u9a76\u3001\u5de5\u4e1a\u81ea\u52a8\u5316\uff0c\u5e7b\u89c9\u5374\u5371\u5bb3\u663e\u8457\u3002<\/p>\n\n\n\n


\u7fa4\u96c6\u6a21\u578b\u7cfb\u7edf\uff1a\u5728\u51ef\u6587\u00b7\u51ef\u5229\u300a\u5931\u63a7\uff1a\u673a\u5668\u3001\u793e\u4f1a\u4e0e\u7ecf\u6d4e\u7684\u65b0\u751f\u7269\u5b66\u300b\u4e00\u4e66\u4e2d\u63cf\u8ff0\u4e86\u590d\u6742\u7cfb\u7edf\u7684\u8fdb\u5316\u3001\u6d8c\u73b0\u548c\u5931\u63a7\u3002\u4eba\u7c7b\u5927\u8111\u7684\u795e\u7ecf\u7f51\u7edc\u3001\u8681\u7fa4\u3001\u8702\u7fa4\u8fd9\u7c7b\u7cfb\u7edf\u7684\u52a8\u4f5c\u662f\u4ece\u4e00\u5927\u5806\u4e71\u54c4\u54c4\u5374\u53c8\u5f7c\u6b64\u5173\u8054\u7684\u4e8b\u4ef6\u4e2d\u4ea7\u751f\u7684\u3002\u4ece\u7fa4\u4f53\u4e2d\u6d8c\u73b0\u51fa\u6765\u7684\u4e0d\u662f\u4e00\u7cfb\u5217\u4e2a\u4f53\u884c\u4e3a\uff0c\u800c\u662f\u4f17\u591a\u4e2a\u4f53\u534f\u540c\u5b8c\u6210\u7684\u6574\u4f53\u52a8\u4f5c\u3002AI \u5927\u6a21\u578b\u4e5f\u7b26\u5408\u7fa4\u96c6\u7cfb\u7edf\u7684\u7279\u70b9\u3002\u7531\u4e8e\u7f3a\u4e4f\u4e2d\u5fc3\u63a7\u5236\uff0c\u7fa4\u96c6\u7cfb\u7edf\u6548\u7387\u76f8\u5bf9\u4f4e\uff0c\u5982\u5927\u6a21\u578b\u5c31\u5b58\u5728\u4fe1\u606f\u5197\u4f59\uff0c\u4e14\u4e0d\u53ef\u9884\u6d4b\u3001\u4e0d\u53ef\u77e5\u3001\u4e0d\u53ef\u63a7\u3002\u4f46\u7f3a\u4e4f\u4e2d\u5fc3\u63a7\u5236\u4e5f\u5e26\u6765\u4e86\u53ef\u9002\u5e94\u3001\u53ef\u8fdb\u5316\u3001\u65e0\u9650\u6027\u548c\u65b0\u9896\u6027\u7684\u4f18\u52bf<\/strong>\uff0c\u56e0\u6b64\u5927\u6a21\u578b\u80fd\u901a\u8fc7\u6d8c\u73b0\u81ea\u5b66\u65b0\u6280\u80fd\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"

Stuart Russell\uff1a\u76ee\u524d\u7684\u4eba\u5de5\u667a\u80fd\u7cfb\u7edf\u6ca1\u6709\u901a\u7528\u6027\uff0c\u56e0\u4e3a\u7535\u8def\u65e0\u6cd5\u5f88\u597d\u5730\u6355\u6349\u901a\u7528\u6027\u3002\u6211\u4eec\u5df2\u7ecf\u5728\u5927\u578b\u8bed\u8a00\u6a21 […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[1],"tags":[],"_links":{"self":[{"href":"http:\/\/47.92.147.95\/index.php\/wp-json\/wp\/v2\/posts\/825"}],"collection":[{"href":"http:\/\/47.92.147.95\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/47.92.147.95\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/47.92.147.95\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/47.92.147.95\/index.php\/wp-json\/wp\/v2\/comments?post=825"}],"version-history":[{"count":2,"href":"http:\/\/47.92.147.95\/index.php\/wp-json\/wp\/v2\/posts\/825\/revisions"}],"predecessor-version":[{"id":831,"href":"http:\/\/47.92.147.95\/index.php\/wp-json\/wp\/v2\/posts\/825\/revisions\/831"}],"wp:attachment":[{"href":"http:\/\/47.92.147.95\/index.php\/wp-json\/wp\/v2\/media?parent=825"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/47.92.147.95\/index.php\/wp-json\/wp\/v2\/categories?post=825"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/47.92.147.95\/index.php\/wp-json\/wp\/v2\/tags?post=825"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}