There are a couple of things wrong here. First off, there is a historic climate record that goes back centuries and is fairly accurate. Second, the climate prediction models are tested against this historic record. They reproduce the historic climates quite well. The error margins are generally shrinking due to model improvements.

But when making actual predictions, the models need to make assumption about anthropogenic parameters like CO2 and methane emissions. That's the largest remaining uncertainty at this point. Given the same assumptions, climate models generally agree on the outcome.