{"resolved":true,"match_type":"uuid","source_type":"original_book","book_id":"72c4da97-7cf2-441a-a985-c1252297d6a3","title":"人工智能的进化 计算机思维离人类心智还有多远","author":null,"edition":null,"publisher":null,"year":null,"isbn":"9787508685670","requested_by":"printed_page","book":{"source_type":"original_book","book_id":"72c4da97-7cf2-441a-a985-c1252297d6a3","id":"72c4da97-7cf2-441a-a985-c1252297d6a3","uuid":"72c4da97-7cf2-441a-a985-c1252297d6a3","title":"人工智能的进化 计算机思维离人类心智还有多远","normalized_title":"人工智能的进化计算机思维离人类心智还有多远","author":null,"edition":null,"publisher":null,"year":null,"isbn":"9787508685670","isbns":["9787508685670"],"parsed":true,"ai_readable":true,"document_type":"book","page_count":216,"navigation_url":"https://workshop.closeclaw.top/ai/book/72c4da97-7cf2-441a-a985-c1252297d6a3","title_url":"https://workshop.closeclaw.top/ai/title/%E4%BA%BA%E5%B7%A5%E6%99%BA%E8%83%BD%E7%9A%84%E8%BF%9B%E5%8C%96%20%E8%AE%A1%E7%AE%97%E6%9C%BA%E6%80%9D%E7%BB%B4%E7%A6%BB%E4%BA%BA%E7%B1%BB%E5%BF%83%E6%99%BA%E8%BF%98%E6%9C%89%E5%A4%9A%E8%BF%9C","fixed_page_url_template":"https://workshop.closeclaw.top/ai/book/72c4da97-7cf2-441a-a985-c1252297d6a3/page/{printed_page}","path_api_url_template":"https://workshop.closeclaw.top/api/ai/book/72c4da97-7cf2-441a-a985-c1252297d6a3/page/{printed_page}","read_url":"https://workshop.closeclaw.top/api/ai/read/72c4da97-7cf2-441a-a985-c1252297d6a3","page_url":"https://workshop.closeclaw.top/api/ai/books/72c4da97-7cf2-441a-a985-c1252297d6a3/page","search_url":"https://workshop.closeclaw.top/api/ai/books/72c4da97-7cf2-441a-a985-c1252297d6a3/search"},"page":{"printed_page":186,"pdf_page":197,"physical_index":197,"physical_index_base":0,"printed_page_status":"recognized","page_sha256":"8aec49c1a50f9aa616c1aa51f8bf24f787215eb29447d964a4e2a6e0dd9ece0a"},"chapter":"第8章符号与符号处理","content":"但在符号逻辑当中，目前为止最好的算法可能就是DPLL(数字锁相环，可参见文章Satisfiability solvers的例子)。逻辑问题存在有 n个变量，求解大约需要2\"步The intractability ofresolution。（而证明这点则需要根据消解规则，对DPLL进行变体(A machine-oriented logic based on the resolution principle ), 文中已经提到)。这就是说，即使是最快的计算机，对于只有100个变量的逻辑问题也无能为力。\n\n这又引出了两个问题。首先，我们在想是否会有比DPLL更好的算法。在数学领域，这个问题的精确版就是著名的P=NP问题，这个问题由斯蒂芬·库克（Stephen Cook）于20世纪70年代首次提出（参见 The complexity of theorem-proving procedures)。从那时起到现在，虽然已有成千上万名计算机科学家和数学家竭尽全力，试图攻克难题，但是依然没有答案。由于这个问题与其他众多题目息息相关，因此被视为计算机科学领域最为重要的开放问题(参见文章 The status of the P versus NP problem)。\n\n第二个问题涉及前一章提到的长尾现象。DPLL的工作方式是系统性地搜索逻辑上的所有可能。有趣的是，在随机构建的测试当中，DPLL执行此类操作所需的步骤却都很少。实际上，上述测试所需的步骤与上一章中的长尾数字非常相似。因为样本测试案例越多，平均值就越高，所以在实践中，根本不可能估算出DPLL所要求的步数。有关详情，请参阅Heavy-tailed phenomenain satisfiability and constraint satisfaction problems —文。\n\n人工智能的进化\n\n186","previous_page":185,"next_page":187}