We don't have to speculate
There used to be a standard retort against anyone doing computer modelling along the lines of "I'll consider your predictions for the next century once you've shown me your predictions for the last one."
To be specific, you feed in the available data for global climate in 1900. If this isn't known terribly accurately, that doesn't matter because you just get larger error bars on the final predictions. You add in your assumptions about solar behaviour. As the model evolves, you add in your data on the actual growth of industrial activity, but you *don't* add in any new climate data. (That would be cheating.) After your model has run, you review the envelope of predictions that it produced. Does what actually happened lie within the envelope?
If it does, then we look at the size of the envelope. If it is so large as to be useless, then we just can't tell what will happen in the next century. We need to keep working on the model. Other research activity is, frankly, pointless.
If the envelope is small, you can use the model as a tool, and see which of the data you added in later (solar model and industrial activity) really mattered. If it's mainly down to the solar model, there's nothing we can do in the next century. If it's mainly down to industrial activity, then we're in the land of anthropogenic warming.
This test is simple enough that you don't need a PhD in climate science to evaluate whether it has been passed or not. Indeed, questions of the form "does my model fit my data" are the stock in trade of experimental science and there are widely practised and understood mathematical tests for quantifying the level of agreement.
Twenty years ago, the models failed the test. That means that twenty years ago, talk of global warming *was* just unsupportable scare-mongering, whether it was true or not. (That's perhaps too subtle for many, but a sensible government *shouldn't* commit trillions of dollars to your pet project if you can't prove there is anything happening.)
In the last few years, I've heard claims that the models now pass the test, so their predictions are worth the paper they're printed on. This being science, presumably the evidence for those claims has been peer-reviewed and is in the public domain somewhere. Unless you know otherwise, I think the default position of the lay audience should be that there *has* been significant progress in climate science in recent years, and that the sceptics in other disciplines simply haven't paid attention to recent improvements in modelling.