A couple of days ago, I was presenting a research poster at a machine learning conference called ScaledML. Large groups of people were gathering around my poster while I delivered ad-hoc summaries of my work and answered questions on a subject about which I had now apparently become a recognized expert: determinism in deep learning. In fact, I’m not aware of anyone else who is focused on this, so perhaps I’m actually the only expert in the world on this topic.
Last week, at the GPU Technology Conference, I stood on a stage and talked about the same topic for 40 minutes to a room containing almost 140 people. At the end, nearly everyone stayed while I answered a stream of questions for the remaining ten minutes of the session. I told the audience how grateful I felt to be able to share my work with so many interested and appreciative people. I almost cried. At the bottom of this article is a link to My First Public Tech Talk: What I’m Learning in Preparations, which is about that talk.
Shortly afterwards, I had a one-on-one with an engineering manager at NVIDIA, the company where I work, in which I mentioned how amazed I was that I had apparently become an expert on this topic in only one year. He said, “I thought that you were already an expert on this!” I told him that my knowledge of deep learning had all been self-taught over the past few years, and that…
