3 ms·

I wonder why twitter will not use Machine Learning to remove terrorism propaganda from their social network? I mean you can train neural networks to recognize t

by lossolo 10y ago

I wonder why twitter will not use Machine Learning to remove terrorism propaganda from their social network? I mean you can train neural networks to recognize terrorism spreading accounts and tweets. It's not black magic, NLP to get intent + classification of known/blocked terrorist accounts and just train. Then show message "Sorry but this tweet looks like abusing our rules, please wait until our staff will check it" or something like that. If you do nothing where you can then you are just accomplice to spreading terrorism.

ceejayoz 10y ago
The likely end result of that is just coded communication. As a similar example, Facebook's "no gun sales" policy now has people selling a pair of socks for $650 using a picture that just happens to have a gun in the background. (Examples: https://twitter.com/ecined/status/756940120826097664 https://twitter.com/ecined/status/756940120826097664 https://twitter.com/monteiro/status/756639900796088320 https://twitter.com/monteiro/status/756639900796088320)
- hiou 10y ago
  And you honestly don't think machine learning will be able to detect that as a gun sale?
  - ceejayoz 10y ago
    The more it does, the more coded it gets. I still have to dig into my spam folder's contents periodically to rescue important emails for the same reasons.
artursapek 10y ago
I'm pretty sure they allow it because the government asked them to. It can provide useful intelligence. I doubt detecting and removing that stuff is beyond the Twitter team's technical capabilities.