and insurance companies running models.

it is absurd to suggest that "all academic epidemiology be defunded" just because a few modelers have ethnic issues or a few models behave badly. It is like suggesting to defund the military because a few soldiers went rogue.

By the way, insurance companies run models all the time (most likely for COVID as well). These models can be and have been used to shape policies debate. However, there must be independent research/checks to ensure the potential biases due to conflict of interest be kept at a minimum. In fact, this is why any research funded by commercial companies or other parties with a conflict of interest should be disclosed as such. For example, you can easily imagine models from insurance companies will say pre-existing conditions are the most important factor in determining the patient's death rates. True or not, it could very much be a pre-defined assumption.

Now how many people outside of academics understand models or modeling? I would say not many. Again, "all models are wrong, some are useful". Below is a short explanation why this is the case if you are interested.

Every model comes with a set of pre-defined assumptions. Once one or more of those assumptions are violated, the model results will be likely invalid. For COVID forecast models, for example, the infection rates and mortality rates are key parameters that need to be defined, which were not clear 7 months ago or even now. Also, human behaviors (e.g. social distancing or not, business as usual or not) are critical factors that will strongly affect the evolution of the pandemic. So these factors must be correctly defined (assumptions) in the models as well. However, you know, some MAGA people refused to follow any guidances, and, at the same time, some local government opted to mandate strong restrictions while others refused to do anything. All of these create uncertainties or biases that the models are unable to accurately reproduce. Of course, there are also the complex behaviors of the virus, which so far we still don't know well and so are difficult to simulate to begin with.

So is it any surprise model forecasts aren't very good? (Again, I am not talking about one or two particular models, maybe London model has some code issues, which I don't really care to check).

These also mean all models' assumptions need to be revised as we know better, and new forecasts should be made accordingly.

So models are just models, regardless it is from London or Seattle. Anyone assuming one set of models results are truths is an idiot. Anyone disregarding model forecasts completely is one as well. Take your picks. But if you want to use models properly, then proper context and explanation are needed from the model developers.

The funny thing about this era is, sort of echoing QTL's comments, anyone can play scientists, politicians, historians, etc., by just googling a bit or maybe just reading a few crazy blogs. It is tempting because all of sudden you are so smart. And you don't need years of training and actually doing any research or practicing politics. The truth is NO, not so fast.

My suggestions: you are most welcome to try, but use your brain and do some serious homework first.