Even that isn't correct - the stress in "president" is on the first syllable, not the second. It's interesting to me that both errors lie in the domain of spok

Even that isn't correct - the stress in "president" is on the first syllable, not the second. It's interesting to me that both errors lie in the domain of spoken rather than written language. I wonder whether they'll persist in multimodal models trained on speech, as well as text and images.