An Experimental Study on the Counting-Rule-Learning in a Chimpanzee.

一只雌性黑猩猩，估计12岁，在经历了大量与物体匹配的阿拉伯数字或随机的点图案后，人工智能终于学会了计算视频监视器显示的物体。发现以下几点：1.响应时间曲线作为点的数量的函数，在形状上与人体实验中获得的曲线非常相似。这意味着subitizing和计数功能都可以反映出来。然而，非常有趣的是，在第一阶段，对最大数字的响应时间比第二大数字短，然后如果添加大数字，则变为最长。这种对最大数的反应时的变化，与人工智能从相对数值判断到计数的反应变化相对应。这些由人工智能获得的基数被正确地用于由两种不同大小的点、两种不同颜色的物体和两种不同类型的物体组成的异质图案.然而，为了计算异构模式子集中的对象数量，需要进行长时间的训练。这些结果意味着AI学会了使用阿拉伯数字作为一个标签来命名一组对象，基数用于计算一组对象的数量。这样的人工智能的性能仍然是有限的，在一定范围内的刺激，很容易受到干扰，如果在一个子集的异构模式的对象的数量是必需的。经过大量的训练，人工智能可能会使用数字作为抽象符号进行逻辑运算。

A female chimpanzee, estimated 12 years old, AI finally learned to count objects presented by the video monitor, after she had experienced lots of matching Arabic numerals to the objects or random patterns of dots. The followings were found.1. The curves of response times as a function of the numbers of dots were very similar in the shape to those obtained in human experiments. It implied that both functions of subitizing and counting could be reflected. It is very interesting, however, that the response time to the largest number was shorter than the second largest number at the first stabe, then became the longest if the large number was added. This change of the response time to the largest number would be correspond to AI's response change from relative numerical judgement to counting.2. These cardinal numbers acquires by AI were correctly used to the heterogeneous pattern consisted of two different size of dots, two different colored objects and two different kinds of objects.3. Long training was necessary, however, to count the number of objects in a subset of the heterogeneous pattern. These results means that AI learned to use Arabian numerals as a label for naming a set of objects at first, the cardinal numbers for counting the number of objects in a set. Such Ai's performance is still limited within some range of stimulus and easily disturbed if the number of objects in a subset of the heterogeneous pattern is required. It may be possible that AI can use the numerals as the abstract symbols to operate logically, after great deal of training.