The fact that the argument keeps changing should tell anyone everything they need to know.
Not to mention we're still in a cold period.
And the notion that there's some global temperature that climate is "supposed to be' is patently unscientific and ignorant of history.
The data shows we SHOULD be in a cold phase but the Earth has been warming rapidly compared to the last 10,000 years. The last time the average temperature rose 1c rapidly it took 900 years. Since the Industrial Revolution( roughly 1850s) the average temperature has risen a bit over 1C.
You can insist that given the great complexity of the Earth's ecosystem, scientists could not possibly know what will happen. They can theorize about weather change and some are right and some are wrong. But there is no doubt that the average temperature is rising and some areas close to the equator will start to be come uninhabitable in 20-30 years.
Here's the main journal article xkcd referenced for that comic.
You've noted, in different words, that the trend since around 1900 is unprecedented in the entire time frame of the temperature reconstruction, last 20k years or more. You are absolutely correct, the journal article re-confirms that the graph trend from 1900 onward is unlike anything in the 20,000 years prior in the entire dataset.
If you read closer though, there is another potential explanation beyond human CO2 emissions that must also be accounted for. If you check the article, you will find that the data set from 1900 backwards is a DIFFERENT data set than the one graphed from 1900 forwards. The data graphed prior to 1900 is reconstructed from proxy sources, the data graphed after 1900 is the instrumental record.
When temperatures averaged over 100+ years, it's tough to average the tail end last 100 years well so using the instrumental record isn't wrong. Drawing conclusions SOLELY on the divergence that happens at 1900 though is to say the least, nuanced. Plainly, the most important and probable factor that must first be thoroughly ruled out is that the change in data sets is having an impact. There's a possibility that thermometers measure temperature more accurately than proxy records that are statistically analyzed and averaged over hundreds or thousands of years.