Automatically reducing an argument to saying someone is equivalent to an anti-vaxxer kind of sidesteps the issue and is a logical fallacy. If there is a problem with the argument, elucidate the issues, but "Reductio ad X" arguments are not valid. This is a large and nuanced issue, which requires more thought and argumentation than can usually be contained in a HN comment, but I think the reduced thing we have here is whether computer code predicting a certain outcome should be trusted -- and, more importantly, making major decisions based on that prediction, given the nature of bugs and problems we see in normal code. I believe we all agree that potential consequences are far worse, we simply disagree on how to treat those consequences and where they come from. Because there is so much at stake, it is best to be more sure of what is happening and to take the right choice, instead of the first one we thought was correct based on a limited, and potentially flawed, computer model. If the computer model predicts dire catastrophe, then we should take it seriously and do another one with even greater resources in order to ensure the prediction is correct and what course of action should be taken. Perhaps it is right in predicting catastrophe, but not in predicting the -right- catastrophe, in that case we could spend a large amount of resources on the wrong solution and miss the right solution, which could end up being even more catastrophic.