This. The difference between most higher paid specialties and and the ones that only require 3 years of training (eg EM, IM, FM) is much smaller than what these surveys make them to be, once you account for training time. The difference also increases once you factor in lost gains from investment over those years for specialties with longer training time.
Also, it would be more accurate to compare post-tax income. Even an additional $200k per year you could make in a high-paying specialty vs a lower paying specialty comes out to only ~$110k additional pear year after taxes. This is problem with high-paying specialties that have long training time; your income is compressed at the end and this is not favorable from a tax standpoint.
Also, would not rely on Medscape as the prime source of physician salaries. Their sampling method is very susceptible to bias, and they are listed means/averages (which for most specialties will be higher than the median as outliers affect the mean much more than the median). MGMA seems to be the gold standard for physician compensation data (and even then it still has its share of limitations) and it also gives specialty median pay which is more accurate, as well as percentile pay for each specialty. It does come with a fee to use, but med students and residents looking at compensation data should use MGMA instead if they want a better representation.