News for hackers. Stuff that matters.

A comment on DeleteFB: Selenium script to delete all of your Facebook wall posts. Parent comment

Re:DeleteFB: Selenium script to delete all of your Facebook ... by Fnoord on Tuesday May 21, 2019 @04:16AM (#19968399)
Either

#1 You can train the model continously

#2 the model stops learning

#3 it has a submodel which detects you are feeding it false data.

Given that people change over time, #2 seems unlikely. The question to me is "how advanced is #3?". Another thing I wonder about is how relative are these to eaxh other. At odds?

[ Parent | on HN ]
Re:DeleteFB: Selenium script to delete all of your Facebook ... by nerdponx on Tuesday May 21, 2019 @09:06AM (#19969747)
#1 is a well-established technique called "online learning", and I'd bet money that it's how many of these "industrial scale" ML algorithms are trained.

#2 makes no sense from a business perspective.

#3 is also well-established. This is how Google's Captchas work, for example.

[ Parent | on HN ]