Artificial intelligence experts warned that large language models have issues including "hallucination," which means that the software can make stuff up.