Because you never looked. What he's saying is a true and well known problem in climatology. Population increases create an appearance of warming in the dataset.
There are two sources of corruption.
The first is the urban heat island effect. Many weather stations are corrupted because although they were once in fields or other natural sites, over time cities expanded to meet the needs of the growing population and ended up building around the stations. This distorts readings by pushing up the min on the min/max thermometers, as tarmac absorbs heat during the day and re-radiates it at night. If you compute the average temperature for a day by just taking the midpoint between the min and max this makes it appear that the site is warming even if the max doesn't change i.e. there is no actual climate warming.
The second is that a large number of weather stations sit next to runways. It's intentional: pilots need to know runway temperature for takeoff calculations. But climatologists obviously shouldn't use them because not only are they next to tarmac, they're also being blasted with jet exhaust. So as population increases due to immigration (as western countries have below replacement fertility), jet travel goes up and this causes more traffic which warms the air at these weather stations.
In some countries this is a serious problem. For example, the UK is very built up with unusually high population density. The World Meteorological Organization has a classification system for weather stations in which class 1 is the best (lowest uncertainty) and class 4/5 are the worst (uncertainty of 2/5 degrees respectively, which is higher than the 1.5 degrees of warming claimed to have happened since 1850). Unfortunately nearly 90% of all British weather stations are class 4 or class 5, but the Met Office still use them to make claims about record breaking temperatures.
Climatologists have techniques they say correct for these problems, but the algorithms don't actually work. US GHCN uses something called homogenization, a kind of interpolation which spreads warmth from cities and runways into colder rural areas, creating an appearance of global warming across the entire dataset that's not there in the thermometer readings from uncorrupted sites.