Raw LLM Responses

Yes and no. The "ai" part of the tech is overblown. It's a buzzword rather than a true descriptor. "Intelligence" in a true sense is not part of it. Basically if you think about it as an advanced modeling device, you are closer to the real story. Effectively, this is a step past a standard GPU. Your normal GPU has a large bank of memory and a specialized processor than handles processing that memory in parallel. These operations are generally relatively simple. Hardware side enhancement for this involves adding some level of capability to these processors. Something that might take multiple iterations of a multiplication operation may be condensed down to be handled in a new operation that directly handles it in a handful of the processing cycles. This is incredibly cool technology and we absolutely should be locking that shit down, but it's something that improves an existing capability rather than delivering a whole new one. It turns out that certain improvements make certain applications of existing technology viable, but that's the range in which this is operating. It's something of a personal crusade of mine to aggressively go after things labeled "intelligence" that very much are not that because it creates a set of expectations that are not properly checked until someone posts a video of a tesla ramming a mannequin.