The number of people who die by today has nothing to do with the number of people diagnosed by today. If you had the number of people who had recovered by today, you could compare it to the number of people who died by today and come up with a reasonable fatality rate.
The woman who died today has been hospitalized since March 27. If you assume that all of the 8 Utah deaths were diagnosed just before March 27, then a more reasonable fatality rate would be 8/472 or 1.7% which is three times higher than what you calculate and is probably still an underestimate.
One other way to think about it would be: If all new infections stopped today so that the total number of diagnoses in Utah stays at 1478, how many people would die? Would it be 8? No, it would probably be at least 24 once all current patients either pass away or recover.
Your method of calculating fatality rate will continue to be a massive underestimate as long as the growth rate of diagnoses continues to increase.
Underestimates are dangerous because they give people a reason to become lax about the social distancing and hygiene that have helped slow the growth. If you're going to use amateur data science to paint a rosier picture of our current Covid-19 situation, please use it correctly.