Raw LLM Responses

Yes, but it requires resourcefulness. SETI@home attempts recoup idle CPU power, as do distribute computing projects like BOINC. At my university, I had a colleague who got his hands on a 3D printer and printed himself a whole lab. He used the 3D printer for the gears, case, etc. But a lot of the other material was salvaged: old G4 macs to process the geometry. Used arduinos to drive the stepper motor. He was using old Microsoft Kinnect controllers as 3D scanners. But there will be limitations: a 2006 Mac Pro, for example, isn't really useful for machine learning, even if most of the compute happens on a gpu. The PCI interface is slow and not entirely backward-compatible, the CPUs are missing modern instructions like AVX. The ECC ram is ideal for a file server, but these things idle at like 200 watt.