HackerNews中文版

今天在回家的路上，我突然想到，神经接口领域真正的突破将发生在数据层。在从事电子工作的过程中，我了解到存在一个硬件传输层，也就是信号传输的导线。然后还有一个软件/协议层，它定义了在硬件上传输的_内容_。目前我对 Neuralink 等技术的理解是，它提供了一个坚实的接口，从大脑接收输入，并向大脑提供输出。在这个接口的背后，有一堆硬件和软件，用于转换和利用来自大脑的输入。也就是说，我们从一种信号模式和信号传输方式转换到另一种。我突然想到的是，一个真正的仿生体不会仅仅提供一个与人类大脑现有的硬件和软件数据层的接口，而是会利用新的可用神经元来扩展现有的数据层。现在，你可能一开始会采用“位操作”的方式来实现这一点，也就是说，让你的仿生神经网络存在于软件中，并进行我们目前所做的所有信号处理。真正的革命将是一个硬件，它只需插入大脑，就能在与大脑已经运行的电网相同的电网上提供一个全新的神经网络。

查看原文

It occurred to me on my way home today that the key advancement in in neural interfaces will be in the data layer.In my work with electronics I learned that there's a hardware transport layer, the wires on which signals travel. Then there's the software/protocol layer that defines _what_ travels on the hardware.My current understanding of things like neuralink is that there is a solid interface that takes input from the brain and provides output back to the brain, and behind the interface is a bunch of hardware and software that translates and uses the inputs from the brain. That is, we change from mode of signals and signals transport to another.What occurred to me was that a true bionic won't provide an interface to the existing hardware and software data layers of the human brain, but will instead expend the existing layers with new available neurons.Now, you could probably bit-bang this at the start, IE, have your bionic neural net live in software, and do all the signals processing that we currently do. The revolution will be a piece of hardware that simply plugs in to the brain and makes a whole new neural network available on the same electrical net that the brain already operates on.

神经接口的关键技术进展