The concept of hallucination is directly linked to GPT-3 because the model has been measured to produce these factually incorrect responses in approximately 15% of its outputs, as noted in [1].