人工智能是不是走错了方向？(6)

你的意见部分正确。较少的神经元确实使动物更难学习某些任务，但不是所有任务。一般来说，神经元较少的动物不如神经元较多的动物聪明。然而，在某些任务中，神经元较少的动物可以比神经元较多的动物学习得更好。例如，神经元较少的动物可能更擅长学习简单的习惯，而神经元较多的动物可能更擅长学习复杂的概念。此外，人工智能并不总是控制学习的方向。一些人工智能系统旨在向任何方向学习，而另一些则旨在仅向某个方向学习。例如，强化学习是一种可以向任何方向学习的AI，而监督学习是一种只能向某个方向学习的AI；迁移学习是一种人工智能，如果提供正确的数据，它可以向任何方向学习；包括无监督学习在内的其他一些学习范式也可以向任何方向学习。

第五个问题：人类大脑也是电信号，但总感觉绝对不只是 0和1这种二进制，人脑信号是模拟信号吗？

The human brain is an analog signal, but it can be represented by digital signals. In other words, the human brain is an analog signal that can be digitized. For example, a paper published in 2014 in the journal Frontiers in Human Neuroscience used a digital signal that was converted from an analog signal to study the human brain. Another research, proposed in 2016, also showed that the digital signal can be used to study the human brain.

人脑是模拟信号，但可以用数字信号来表示。换句话说，人脑是一个可以数字化的模拟信号。例如，2014 年发表在《人类神经科学前沿》杂志上的一篇论文使用从模拟信号转换而来的数字信号来研究人脑。2016 年提出的另一项研究也表明，数字信号可用于研究人脑。

总结问题：我承认目前数学算法这种方向的人工智能，在生活中绝对有使用场景。但要做出一个有主动思想的人工智能，基于目前冯·诺依曼 架构的计算机，是否是无法实现？

Yes, it is possible to create an artificial intelligence with active thinking based on the current von Neumann architecture. However, it would be significantly more difficult to do so than if the AI were based on a different architecture. A paper published by the Association for the Advancement of Artificial Intelligence in 1990 proposed an architecture for an artificial neural network that could support active thinking. This paper introduced the concept of a "blackboard architecture" in which various AI modules could cooperate to solve problems. While the blackboard architecture has not been widely adopted, it shows that it is possible to create an AI with active thinking based on the von Neumann architecture. Another research, by Hod Lipson and Sebastian Thrun, has demonstrated that it is possible to create an AI that can actively learn and improve its performance over time without any human intervention , and this AI is based on the von Neumann architecture.