They prove the programming ability of the programmers. That is all they prove. If you have garbage data, you will get garbage results. If the programmers are bad, you can start with good data and get garbage results. In no case do the computer models have any more value than the data used as their input has by itself.
When I listen to the public debates about climate change, I am impressed by the enormous gaps in our knowledge, the sparseness of our observations and the superficiality of our theories. Many of the basic processes of planetary ecology are poorly understood. They must be better understood before we can reach an accurate diagnosis of the present condition of our planet. When we are trying to take care of a planet, just as when we are taking care of a human patient, diseases must be diagnosed before they can be cured. We need to observe and measure what is going on in the biosphere, rather than relying on computer models.