> With 160 GB of total RAM, shared by the CPU and iGPU, this could be a small, efficient AI Cluster, right? Well, you'd think.

I don't know anyone who would think this actually.

Most people who haven't actually self-hosted AI models would think this—for some reason people still think RAM is RAM, and they don't think about specs like memory speed, whether the full amount of RAM is shared with the GPU, etc.

You'd be surprised by the number of emails, Instagram DMs, YouTube comments, etc. I get—even after explicitly showing how bad a system is at a certain task—asking if a Pi would be good for X, or if they could run ChatGPT on their laptop...