Jürgen Müller, SAP

Today’s large language models (LLMs) seem to come prebaked straight from the factory. They’re these fixed tools that only know about the information they were trained on. They’re incredible, but, at the same time, incredibly limited.

“An LLM doesn’t know what it doesn’t know,” says Zain Hasan, a senior developer advocate at Weaviate. “It only knows what it’s been trained on. It knows the probability of words or concepts.”