Google’s Gemini is excellent at calling the right runtime data stores to augment answers, which makes it shine against other models. However, there may be over reliance on this at the expense of training semantics properly in the foundation model in the first place. A simple example that Gemini should have gotten right (both Claude and ChatGPT answered this within bounds of available data):
hashtag#AI hashtag#AImodels Activate to view larger image,