This woman cleverly used an Artificial Neural Network (ANN) to forecast twentieth century temperatures (which we actually know) from published proxy records used by other "climate scientists" (tree rings, coral, etc.) to feed their climate models.
Apparently the technique below works well enough to predict rainfall a few months in the future (http://jennifermarohasy.com/).
This was accomplished by "training" the ANN on proxy temperature data up until 1830.
(In case you don't understand an ANN is basically a pattern matcher. It works conceptually like this: Suppose I show you a series of images taken while driving down the road. Your mind would probably quickly discover that small items which appear in the distance would generally get larger as the car moves forward hence you could "guess" what the image would look like after driving for a few more seconds. An ANN can basically "learn" similarly; in this case looking by looking at information derived the proxy data it can "guess" what temperatures would occur after 1830 in this case.)
From the linked blog regarding this chart: Proxy temperature record (blue) and ANN projection (orange) based on input from spectral analysis for this Northern Hemisphere multiproxy. The ANN was trained for the period 50 to 1830; test period was 1830 to 2000.
As you can see there is a very good match between the portion of the chart after 1830.
In fact, the match is within 0.2C.
What this says is that, based on the prior patterns of temperature before the industrial revolution, the current temperatures in the world match very closely to the ANNs predictions: only the ANN doesn't know about the industrial revolution (and the climate warming associated with it by "climate science.").
The link above is quite readable and it's well worth your time to investigate.
Effectively the current temperatures in the world relate very closely to the patterns they always follow.