Ask HN: Prevent ML like GPT from using public posts like this one?
Several things about GPT-3 and related models like ChatGPT bother me.
First, OpenAI is not merely consuming content I generated (however relatively miniscule that is) but deriving value out of it.
Second, it exposes people to analysis and scrutiny that can be used outside of the pubic domain. For example, if I used a throwaway account to talk about bad practices at an employer or gov agency, this technology can potentially be used to identify me based on writing style much more accurately than existing stylometry methods.
Third, threat actors can use how a person speaks to accurately emulate them when social engineering users (which is worsened by deepfake video and audio).
Lastly, when we author content on a public post, the rules and norms that exist don't consider the type of black-box analysis that can be used to a hostile end, so I don't agree that ML analyzing public posts should be covered under existing public domain rules and laws.
My question is, is the EFF or some other org working with law makers to address this?
There are both consumer protection, fair trade and security concerns about this. I have not been hearing this side of the discussion when talking about ChatGPT for example.
Before ML scrapping, public posts were the equivalent of standing in public fully clothed, with ML scrapping it feels like someone in public is using x-ray to look a layer deeper and expose us in a way we did not consent.
It would be great if social media and sites like HN had a way to tell google,openai,etc... similar to robots.txt "don't use this for ML learning". I feel like laws and standards are 5-10 years behind.