I asked a Star Trek question (somewhat obscure but time wise and corpus wise clearly in the training set) and the answer was incredibly wrong … but the output

I asked a Star Trek question (somewhat obscure but time wise and corpus wise clearly in the training set) and the answer was incredibly wrong … but the output sounded very confident and coherent so kudos! Now all we need are the right answers