@Remittancegirl

There have been a few papers released on how we can detect AI content.

Just like your gate, the way that you speak, your sentence construction, punctuation, and vocabulary are unique to you.

Currently, AIs like ChatGPT aren't constrained to this because they are trained on millions of sets of data.

But a new paper sugests that we can change it to prefer certain words, tenses, punctuation and other sentence construction elements which would go unnoticed by humans, but will be statistically detectable by other AIs, allowing us to filter it out and also detect plaigarism and laziness by authors.

The idea is still in its infancy, but it is really possible. It will become part of our world, but AIs won't go unnoticed, and they will be recognizable. It's almost like creating a new race, culture, and language dialect.

#chatgpt #AI #language #filter #plaigarism #research

@Stark9837 So we can‘t detect AI generated content but we can detect watermarked content.
The AI that does not watermark its utterances will just be a little more expensive to use or run on your own system. Unless we forbid open source.

This genie will not be bottled again.

Ch(e)at GPT? - Computerphile

YouTube
@Stark9837 @Stark9837 Yes, #watermark|s that exist, can be found and watermarks, that don‘t exist, can‘t be found.
Why do you expect someone to use an #AI that is watermarked if their intention is to hide that they used an AI?