Distill: Clearing Research Debt

April 8, 2017 by guidj

Recently, I was refered to a website called Distill, by a friend. Their purpose for existing, as they claim, is to reduce research debt. A phenomenon that occurs when there is so much work that has been done before in a field, and is hidden behind complicated explanations that could be rather simplified, thus slowing down scientific progress by demanding a high amount of energy from researchers in understanding previous work. Energy that could be spent elsewhere, in say, doing new research.

I have a few, and short words to say about it.

  • First of all, thank you! After reading Carter and Olah’s essay on research debt, I have to say I felt some of my concerns in research and science in general addressed in the open. As someone who reads papers on a regular basis, and finds himself running down a rabbit hole of concepts, only to get to end the understand that they only meant to say “we wanted to prevent negative values”, and feel like a complete idiot, I stand behind the idea 100%. And I know many researchers go through the same thing.
  • Second, I advise every Machine Learning researcher to read the website. Even if the concepts are already clear to you, I encourage you to read it anyways. Personally, I find that visual explanations help cement concepts a whole lot better in my mind.
  • Third, if you think you have a better way to explain a concept in machine learning, please share it on the site. There is even a $10,000 reward for the most helpful explanation, but even if you don’t get it, you’d be doing an enourmous service to the community, and would earn eternal gratitude from many.
  • Forth, this should be done for all of science. In fact, if you wanna do it for new field, feel free to reach out to me. I would offer any kind of support I can, with my time, to see this spread, and normalized.

I have always considered a limitation of our species that we need time to consume knowledge from previous generations, and then build upon it, because as time progresses, there is more and more knowledge from the past that we need to ingest, before we can enhance it. This will go a long way in reducing that time.

That would be all.