AI Dungeon will block certain words, review content flagged as inappropriate

Re:AI Dungeon will block certain words, review content flagg... by Ggshjtcnjfxhg on Wednesday April 28, 2021 @05:46PM (#26974631)

Their customers could be disturbed if GPT3 outputs child pornography when they aren't expecting it. There's more than an off chance a given customer would find child pornography disturbing; the majority of customers would find child pornography disturbing.

[ Parent | on HN ]

Re:AI Dungeon will block certain words, review content flagg... by the8472 on Wednesday April 28, 2021 @06:06PM (#26974848)

They're also doing

> Additionally, we are updating our community guidelines and policies to clarify prohibited types of user activity.

So they clearly want to prevent everyone from using it that way, not just the unsuspecting users.

If cared about the latter they'd just add an NSFW toggle or something like that.

[ Parent | on HN ]

Re:AI Dungeon will block certain words, review content flagg... by Ggshjtcnjfxhg on Wednesday April 28, 2021 @06:51PM (#26975262)

There's a pretty huge difference between general NSFW and child porn. I don't think adding a child porn toggle would go very well for Latitude.

[ Parent | on HN ]

Re:AI Dungeon will block certain words, review content flagg... by the8472 on Wednesday April 28, 2021 @07:09PM (#26975395)

They don't have to carve the problem space exactly that way. They could have chosen a general NSFW toggle instead but didn't and went for the easy to defend censorship starting position instead.

[ Parent | on HN ]