Your first example talks about "doctors". The set "doctors" includes all living doctors, all dead doctors, all retired doctors, all around the world. So if somebody wants to make a statement about doctors - they have to use a sample of it.
Even if the statement only talks about US based still practicing doctors who are familiar with the pain-relievers A,B,C,D - it would still probably be based on a sample, unless someone actually surveyed every such doctor.
The second example is probably based on the whole population because it sounds like it talks about all the cars sold in US in the year 1996. It's possible that somebody has all the records of cars sold in the US in that year. So in this case they would have the whole population.
Here is one additional example that should help: You are a teacher and you want to measure the difference in height between boys and girls in your class. Let's think about population and sample:
- Population: your class is the population and you can say "boys in my class are on average 5cm higher than girls". This would be statement of fact after measuring the heights. No confidence intervals, p-values or other uncertainty estimates would be needed.
- Sample: your class is a sample of all the boys and girls that could possibly attend your class. Then you could say that the best estimate you have about the difference of height between boys and girls is 5cm. But you are aware that if other boys and girls attended your class it wouldn't be 5cm and there is even a chance that girls would be higher. And you can produce various uncertainty measures.
In general statistics is much more frequently concerned about making statements from a fixed sample and seeing how well it applies to the population from which the sample was drawn.
You talk about population when you have every member of it measured. And you talk about sample when you measure only some members and try to extend your observations to the broader population.