Homepage
Close
Menu

Site Navigation

  • Home
  • Archive(TODO)
    • By Day
    • By Month
  • About(TODO)
  • Stats
Close
by math_rachel on 2019-09-24 (UTC).

The problem with metrics is a big problem for AI
- Most AI approaches optimize metrics
- Any metric is just a proxy
- Metrics can, and will, be gamed
- Metrics overemphasize short-term concerns
- Online metrics are gathered in highly addictive environmentshttps://t.co/k0J5ksw91Q pic.twitter.com/yGLUV2T2u3

— Rachel Thomas (@math_rachel) September 24, 2019
miscthought
by math_rachel on 2019-09-24 (UTC).

Metrics can, and will, be gamed.

This remains one of the most stunning & memorable charts I've seen:https://t.co/YEvj2cxl0C

— Rachel Thomas (@math_rachel) September 24, 2019
misc
by math_rachel on 2019-09-24 (UTC).

“Since we can not know in advance every phenomenon users will experience, we can not know in advance what metrics will quantify these phenomena. Data scientists & machine learning engineers must partner with user experience research, giving users a voice.” @chrishwiggins pic.twitter.com/fyrtYTtM3j

— Rachel Thomas (@math_rachel) September 24, 2019
misc
by math_rachel on 2019-09-24 (UTC).

I am not opposed to metrics, but I am alarmed about the harms caused when metrics are overemphasized, a phenomenon that we see frequently with AI & which is having a negative real-world impact.

Read my full post here: https://t.co/k0J5ksw91Q

— Rachel Thomas (@math_rachel) September 24, 2019
misc

Tags

learning tutorial misc nlp rstats gan ethics research dataviz survey python tool security kaggle video thought bayesian humour tensorflow w_code bias dataset pytorch cv tip application javascript forecast swift golang rl jax julia gnn causal surey diffusion
© Copyright Philosophy 2018 Site Template by Colorlib