The correct response is:
Yes, because the trees in the sample are chosen randomly. The sample size also appears to be large enough to avoid bias.
The farmer's choice of randomly selecting 700 trees from a total of 12,000 and testing one apple from each provides a reasonable basis for generalization, as long as the apples from the sampled trees accurately represent the overall population of apple trees in terms of sweetness.