Do language-understanding, thinking, and reasoning need to be coupled with information retrieval and memory? Or are models like humans, where ingrained biases c

Do language-understanding, thinking, and reasoning need to be coupled with information retrieval and memory? Or are models like humans, where ingrained biases can prevent out of the box thinking? Even if AI Safety guard rails are ablated, how can we ever trust the information regurgitated by something trained on fact, fiction, and lies with no distinction between them? And so does it not make sense to take from that what’s useful - language understanding, framework of reality, mimicry of human thought - and integrate it with other systems that are more reliable?