Tom Murphy (the physicist at Do the Math Blog) frequently makes the point that U.S. energy use has increased by about 2.9 percent per year since 1650. Here's the graph he shows. He then makes a big deal about how straight it is.
Then he goes from that to saying that, if it continued for another 400 years, the seas would boil from all the heat.
But what he doesn't show you--what he fails to point out--is that the graph is NOT linear (does NOT show exponential growth) at the end.
Here's that same graph, from 1850 to the present.
Do you see the nonlinearity? Well, let me focus in even further. Here is a graph from 1950 to the present.
Now the non-linearity is easily viewable. But let's make it even clearer that Tom Murphy's graph has absolutely zero predictive power. This is the graph from 1950 onward, but not on a logarithmic scale.
You can see that, if you used Tom Murphy's extrapolation, you'd ALREADY be off by a factor of two in predicting U.S. energy use in 2010. So Tom Murphy's extrapolation of the energy trend that he claims to exist is completely specious. It appears true, but it's completely false. There is no trend that can be extended even 4 years into the future, let alone 400.