 By now, we all know Watson, to the extent that he is getting full of himself, almost as much as Sherlock. For those who don't know, Watson took on Jeopardy's champions and was crowned champion. But after being hammered with questions, like in this video... Watson decided enough is enough. And it's time for him to ask the questions. Although with an attitude, Watson's intentions are noble. He wants to help those seeking knowledge and make them experts. This work is entitled Time for Watson to Ask the Questions. And it's the joint work of myself and Summa from UIUC, Bob, Tankfei, and Jinjin from IBM. Filled through Jack and Ridge from Rice. Given any source of information, we want Watson to ask the right questions to maximize comprehension. Building the pipeline for this to work is challenging. Starting with resolving co-references, selecting the important sentences, followed by extracting relations, and morphing them into questions. Finally, we need to generate distractors, and this is the focus of this video. In such a sample question, distractors would be the wrong choices that distract the students from the right answer. Now they can't be too easy like cake, or too correct like O2. So to solve this challenging problem, here are our current contributions. In the past, people used to ask an expert to manually label the efficiency of the distractors, or experiment the distractors on students. We formulate an automatic way to evaluate distractors given the data set, and we prove its efficiency. This helps speeding up research in this field. Starting also with this baseline to generate distractors, whereby we retrieve all words near the answer in the embedding space as distractors, we showed that an embedding space trained on all of Wikipedia performs at 25.2%, whereas an embedding space trained on only the scientific subset performs at 52.4%. Also, the baseline only uses the answer as an input, and the average rank of the correct distractors sits at 99.9%. Whereas when we incorporate the question words in our method, the correct distractors are ranked higher at an average of 81.4, something no previous work has done. Thank you for watching, and Watson did not really stop answering your questions. Here's Watson.