AI Hype and the Opportunists

I recently watched this talk by Zachary Lipton. The talk is two years old but still relevant, important, and informative.

  • Some researchers try to anthropomorphize their technology to sensationalize things — such as calling answer prediction models, text comprehension. ( I think it is okay to attempt to catch the attention of your readers with fancy words/phrases, but the readers have to be wary and discerning while reading the text)
  • The number of people who can train neural networks has blown up, but the number of people who can reason about it, unfortunately, has not. I mean thinking about things like when this network will break or what the results might mean by thinking about neural networks. ( which is why concepts like causality are probably necessary).