The Importance of Having Multiple LLMs in an Architecture

The AI still is not perfect, but it's much better on out. It's like your best or your top 10% out of a call center worker, you're still going to swap it in. Right? So, so, so training, tuning, inference. Yeah. And then there's another phase, which is the error correction. That doesn't have to necessarily be the big model. You don't want the big model because there's higher latency. But I don't think we can currently avoid having at least one powerful model in there for really generating the full blown result. Because it does, there's nothing to mounting call center workers, but not all of them can answer my