I'm very pleased to ask that question because it's very easy to answer.
I'll switch to English, if you don't mind.
We're talking about two different kinds of problems in physics. There's an initial value problem and a boundary value problem. Weather prediction involves knowing accurately the initial conditions and extrapolating forward in time to impute the details of a weather forecast.
Climate is the statistics of weather where you do a lot of averaging over time. Climate predictions are made without any initial data. Data is only used in order to verify the predictions made by a climate model, which are made ab initio from pure physical principles. You model the earth, the atmosphere, turn on the sun, compute the physics of energy transfer, the circulation starts up, the atmosphere starts up very quickly, and the ocean takes a lot longer to start up. You actually simulate the climate from first physical principles without any data at all; data is used only to determine certain coefficients and parameterizations and to verify the results after the fact.
So people who make the statement that we can't predict the weather even, let's say, ten days in advance, so how can we possibly predict the climate a century in advance are talking about apples and oranges. Furthermore, the fact is we can predict the weather ten days in advance, and it is through the use of this numerical weather prediction model that we've been able to do that.
When I started my career as a weather forecaster, we couldn't make a prediction beyond about 36 hours. We didn't even try. Weather predictions for ten days now are as accurate as they were for a day and a half back in the sixties when I started in this business. That's all due to being able to model what's actually going on and solving the initial value problem.
Climate people solved the boundary value problem where they changed the content of the composition of the atmosphere, the energy coming in and going out, and other parameters like that. It works.