At ICML 2019 I asked “Pytorch or Tensorflow ?” at arbitrary passer-bys. Depending on their response, I drew a “x” on either side of the notebook. After 1024 queries I have obtained the above picture.
Why?
This was the question I got the most while talking with people. There were some conceptions that I was doing it for a research, or if I worked at facebook or google. I joked that I was collecting a dataset of crosses, or for fun, or to simply watch people’s reactions. The real answer is maybe all of the above, but to be honest, the best answer is “I don’t really know either”.
In science and engineering, especially us in the field of machine learning, we run specific experiments to validate (at times forcefully) the story we would like to tell. The data we collect is a projection of our desires and attachments. We either utter sighs of relief or suffer gut-dropping panic when the curve captioned as “ours” did better or worse than curves captioned as “baselines”.
This experience of random interaction and curation is maybe a counter point. I simply wanted the data as is, without my own projections and attachments. The data collection is biased, unprofessional, and serves no specific purpose. In retrospect, perhaps on a meta level I knew what I was hoping to achieve, which forbids me to ponder what I was doing during the act of curation itself.
How?
It starts with me way-laying people, asking them “Pytorch or Tensorflow?” with a pen and notebook. This flow-chart captures most of the interactions.
Who?
If you contributed to one of the 1024 crosses, truly, thanks! We actually did it! I hope to see you in future conferences and hearing about your good works.
If you read this far, cheers and high five!
— evan
