Re:Bot Submissions to Comment Website Can’t Be Distinguished...
by lgeorget on Friday December 20, 2019 @07:18AM
(#21843570)
That's a dataset of one, and due to your presence on HackerNews, you're likely more tech-savvy than the average reader of comments on news site, so more able to tell comments by bots apart from comments by humans. If a sufficiently large percentage of the population gets 50%, you can say that it passes the Turing test.
For my part, I got only 60% (12 out of 20). This may be because English is not my first language.
Re:Bot Submissions to Comment Website Can’t Be Distinguished...
by ses1984 on Friday December 20, 2019 @08:00AM
(#21843808)
There is no single Turing test. A bot might fool me but not you. You don't have to fool everyone to have an impact. Also not everyone reads everything as though they are in the middle of a Turing test.
Re:Bot Submissions to Comment Website Can’t Be Distinguished...
by sharemywin on Friday December 20, 2019 @08:44AM
(#21844039)
Priming does make a difference.
Re:Bot Submissions to Comment Website Can’t Be Distinguished...
by ses1984 on Friday December 20, 2019 @08:47AM
(#21844056)
Are we at the stage where a comments section needs an anti bot primer blastered over the top?
Re:Bot Submissions to Comment Website Can’t Be Distinguished...
by BiteCode_dev on Friday December 20, 2019 @07:35AM
(#21843668)
Besides, you have only access to the text. A website like Amazon has a huge list of additional data points to make guesses.
It's a very good idea to make sure submissions come from humans, but it's also slightly overstated and alarmist to state that model-generated text is reliably passing the Turing test.