As long as the input parameters are credible (ie fall within the range of the real parameter they propose to represent) the output of these models will be credible, even though they can not be used to predict every foreseen potential event.
Although we now know that with 75-100 years at the present rate of warming (not the accelerated one that we can more likely expect), wet bulb temperatures lethal to humans can be expected to occur commonly. We can't predict the precise number of heat stroke deaths, but we can predict within a known margin of error, what the number of deaths will likely be. ALL indications are they will likely be very high, as even now we are observing summer temperatures that are causing tens of thousands of deaths due to heat stroke and the rate of this cause of death is increasing.
It should be pointed out that these are not the early days of climate science. That increasing carbon dioxide in the atmosphere will warm the planet has been known for over 150 years. The reality is that its not merely a issue of rising temperatures, but rather the ecological consequences of these rising temperatures. These are effects that are being studied now, not at some distant point in the future, when the last denier will be convinced. Biologists can tell you with near certainty what will happen to specific aspects of organismal behavior that will be affected by rising temperature and what the consequences for the organism will be. What we know is that a great many kind of organisms will cease to exist as viable populations very soon, on the order of years and decades in many cases, not thousands of years later.