[mirimir]

Thanks again.

Overfitting a model for global climate change, for example, isn't an issue, because you're not interested in something like physics. I mean, it's based on physics, but that's buried way down in the model.

But physics has different goals. Closer to math, I guess.

[cultus]

It's actually very easy to overfit climate models. They are fit to observed data with statistical inverse problem techniques (the same as I imagine they do with astronomical data). Climate change models are just directly discretized physical equations. Just like astronomy, the decisions are made on what physics are represented in the model and what are parameterized.

[mirimir]

> Climate change models are just directly discretized physical equations.

I'm no expert, but it's my understanding that they're hugely more complicated than that implies. Sure, there's lots of physics there. But also chemistry and biology. The best ones are general circulation models,[0] and the outcomes will never fit some pretty theoretical structure.

0) https://en.wikipedia.org/wiki/General_Circulation_Model

[lisper]

> Overfitting a model for global climate change, for example, isn't an issue, because you're not interested in something like physics. I mean, it's based on physics, but that's buried way down in the model.

It's not so much that as that the goals are different. We want to understand cosmology for its own sake. We want to understand climate change because that knowledge drives policy. For that purpose, it doesn't really matter that we're unable to predict the exact weather in Denver at 11:23 AM on October 27, 2091. What matters is that we are able to predict in broad brushstrokes that the consequences of business as usual will probably be bad, and so we ought to seriously consider doing something about it. There is no conceivable outcome of cosmological modeling that will drive policy changes like that.

[mirimir]

I agree that models for the overall development of the universe are a lot like models for global climate change. The scale is vastly different, of course. But I bet that the relative cell sizes in our models are similar. Because they're running on similar machines.

But the goals for a theory of gravity, and its integration with QM, are totally different. Or at least, that's my perhaps naive opinion.

Edit: That is, relative cell sizes and total cell counts.

[lukasb]

"There is no conceivable outcome of cosmological modeling that will drive policy changes like that."

We hope