zlacker

[parent] [thread] 1 comments
1. jancsi+(OP)[view] [source] 2022-12-12 05:47:53
Surely in return for the value generated by HN'ers pentesting this thing for free, they can provide a hook for HN mods to test strings against its (recent) output.

Or Christ, just ask the fucking thing if it matches its recent output and/or style.

If we weren't all living in the shittiest timeline this would already have been solved before the thing was released.

replies(1): >>nonran+tm
2. nonran+tm[view] [source] 2022-12-12 09:35:00
>>jancsi+(OP)
These are insightful shortcuts if your threat model is a single instance. ML chatbots have not yet proliferated.
[go to top]