物联网和大模型

一点儿极初浅的想法：

物联网的监测数据波形，如设备的振动波形，是设备的外在Sequence表达，其实就是“物”的语言。了解大语言模型的一些算法，有可能启发在物联网产品方面的一些创新。

必须考虑的是，对于物联网领域“物”的语言的处理，常常需要的是精确，不能像chatGPT处理人类自然语言那样有太多的“表达热度”和“创作激情”。

不能拿着AI锤子在IoT领域找钉子。复杂的做法可能不一定适用，反而是一些基础算法和初级的AI模型来得更为实用，如利用自动编码器Autoencoder的波形异常判断，如利用卷积神经网络的产品缺陷图像识别，如将物联网Sequence数据图谱化的图形的AI处理等等。

但对新技术的适当跟踪，并思考如何和产品创新结合，值得持续去做。AI大模型等技术的相关思路和方法，必定也会在工业领域激发创新。

---

IoT and LLM

Some initial, very shallow thoughts:

The monitoring data waveforms in the Internet of Things (IoT), such as device vibration waveforms, are the external sequence expressions of devices, essentially the “language” of things. Understanding some algorithms of LLM(large language models) might inspire innovations in IoT products.

It is crucial to consider that processing the “language” of things in the IoT field often requires precision. Unlike how ChatGPT handles human natural language with much “expressiveness” and “creative passion,” IoT demands accuracy.

We shouldn’t wield the AI hammer in the IoT field looking for nails. Complex approaches might not necessarily be suitable; instead, some basic algorithms and elementary AI models might be more practical. For example, using autoencoders for waveform anomaly detection, using CNN(convolutional neural networks) for product defect image recognition, and applying AI to graph IoT sequence data, among others.

However, it is worth continuously tracking new technologies and thinking about how to integrate them with product innovation. Relevant ideas and methods from technologies like AI LLM will undoubtedly inspire innovation in the industrial field.