Posts by petewilkins@mastodon.world
 (DIR) Post #Aav9Tpabf3BPmdJq2S by petewilkins@mastodon.world
       2023-10-19T07:22:40Z
       
       0 likes, 1 repeats
       
       @anderspuck @Free_Press given how much Musk destroyed the trust & safety capabilities of twitter - which actually did have very good, thoughtful people - the EU does now need to flex it's muscle in a more timely and immediate fashion, for the now willful infringements on twitter. Previously there was an argument of bad actors getting through due to the sheer scale of the platforms - but the active disinvestment in trust & safety should allow more robust application of the DSA.
       
 (DIR) Post #AcTsT3y6Ptlt6x2ujQ by petewilkins@mastodon.world
       2023-12-04T23:01:20Z
       
       0 likes, 0 repeats
       
       @arstechnica this has traditionally been an extremely hard problem to solve - heuristics struggle to keep up, character set variance catches a surprising number of systems that are not unicode compliant, and standard ML approaches suffer from a lack of heterogeneous training data. Super curious to see this in action.