The tricky part of this argument is the use of the word "accurate". Since you are talking about a statistical model, it's worth keeping in mind that this word h

The tricky part of this argument is the use of the word "accurate". Since you are talking about a statistical model, it's worth keeping in mind that this word has a mathematical definition in that context. The output of the said models have an accuracy attribute presented for them. It's not a binary value of accurate/not-accurate, and what matters is how accurate and about what specific prediction. You can argue that the modeling process for designing the breaks of your car is also not 100% accurate, and it really doesn't need to be! It needs to be accurate enough about the break's relevant properties (stopping power, wear and tear, performance in wet wheather, etc.) This is not a scientific analogy to climate forecasting, I just used it to explain the core idea in more familiar terms. The issue is that it's hard to get these across correctly to the general audience. It requires analytical thinking and patience and attention to details, which none are compatible with how today's media and social networks are incentivized and structured (short, catchy, flashy, clickbaity, infuriating, etc.)