Amazing story worth the read, if true - I've not tried to verify it myself, but I see that The Guardian, VICE, Gizmodo and others are now running with the story too.
The scarier part is one of the Reddit comments by AppleGuySnake down the page:
"I found this thread from someone on twitter pointing out that several computer language models use the Scots Wikipedia as their dataset for learning the language."
On a more serious note this is the big downside to Wiki's - we have become far too casual in trusting "the crowd" will take care of basic vetting. In this case it happened to be a crowd of one until someone else noticed. That should be a reminder to everyone.
For all the fun poked today of old school resources like the Encyclopedia Britannica, they at least had a reputation to defend and thus were highly incentivized to be vigilant with quality control. And even then stuff still slipped through.
3 comments
[ 1.8 ms ] story [ 16.9 ms ] threadThe scarier part is one of the Reddit comments by AppleGuySnake down the page:
"I found this thread from someone on twitter pointing out that several computer language models use the Scots Wikipedia as their dataset for learning the language."
On a more serious note this is the big downside to Wiki's - we have become far too casual in trusting "the crowd" will take care of basic vetting. In this case it happened to be a crowd of one until someone else noticed. That should be a reminder to everyone.
For all the fun poked today of old school resources like the Encyclopedia Britannica, they at least had a reputation to defend and thus were highly incentivized to be vigilant with quality control. And even then stuff still slipped through.