Would you argue that it is impossible for a casino to predict whether its dice games will make money if it cannot predict the outcome of the next role of the dice?
I already addressed that in another commenter's response. Games of chance are based on simple statistics - they become easier and easier to predict with a larger number of samples. The climate is a chaotic system - by nature, these become MORE difficult to predict as you go farther out. It's an entirely different thing.
I would not bet on whether it will be warmer or cooler a week from now, but I'd place a substantial sum of money on the bet that it will be warmer 6 months from now.
What about a year from now? You have no idea, and neither does anybody else. Predicting that summer will be warmer than winter is a no-brainer, seeing as how the Earth operates on a clear one-year cycle. But what matters here are the changes from cycle to cycle, not the changes within a single cycle. And our models are nowhere near complete enough to be able to predict those. Like I said, show me a model that can predict today's climate based on 1980's data. There isn't one - they all predict continued warming; meanwhile, the Earth has recently cooled while CO2 continued to rise. There is clearly far more in play than the simplistic models factor in.
My basic point is that the climate is way more complex and chaotic than the models, and you cannot successfully model a chaotic system by simplifying it. It just doesn't work - the tiny (and no-so-tiny) factors that your model ignores end up affecting major trends in the long run in unpredictable ways.