论文部分内容阅读
R.G.Hoffmam(RGH):并行分布式理近来很热门,有些人认为,它可以解“传统的”基于符号的 AI 方法久攻不下的题。对此您有何看法?J.McCarthy(JM):我看过 Rumelhart 等人著的《并行分布式处理》这一两卷长的 PDP经典著作,觉得它仍有五十年代和六十年代神经模式识别研究中所存在的许多缺点。确切地说,它注重一阶谓词。不妨举一个例子,Rumelhart 等人写的第一篇论文是“PDP 模型概要及其顺序思维过程”(载于第二卷),文中有一个(神经)网络按某些家具陈设与否,对一幢房子的各
R.G.Hoffmam (RGH): Parallel Distributed Arithmetic has been a hit lately, and some think it can solve the longstanding problem of “traditional” symbolic AI methods. J.McCarthy (JM): I’ve read the two volumes of PDP classic by Rumelhart et al., “Parallel Distributed Processing,” and I think it still has nerves in the fifties and sixties Many shortcomings exist in the research of pattern recognition. Specifically, it focuses on first-order predicates. As an example, the first essay by Rumelhart et al. Was “An Overview of the PDP Model and Its Sequential Thinking Process” (in Volume II), in which there is a (neural) network based on certain furnishings or not, House of all