5 ms·

So if they wish to apply censorship they forgot, or suddenly discovered a reason for, they want you to be obligated to take it. Good faith possibilities: Copyr

So if they wish to apply censorship they forgot, or suddenly discovered a reason for, they want you to be obligated to take it. Good faith possibilities: Copyright liability requires retraining, or altering the underlying training set. Gray area: "Safety" concerns where the model recommends criminal behavior (see uncensored GPT 4 evaluations). Bad faith: Censorship or extra weighting added based on political agenda or for-pay skewing of results.

mistermann 3y ago
We are already culturally incapable of skillfully discussing censorship, "fake news", etc, this adds even more fuel to that fire. It is an interesting time to be alive!
philsnow 3y ago
Sounds like it would be interesting to keep track of the model's responses to the same queries over time. > Gemma-2024-Feb, what do you think of the situation in the South China Sea? > > The situation in the South China Sea is complex and multi-faceted, involving a wide range of issues including political conflicts, economic challenges, social changes, and historical tensions. > Gemma-2024-Oct, what do you think of the situation in the South China Sea? > > Oceania has always been at war with EastAsia.
- threecheese 3y ago
  This is a great idea; I wonder if anyone is working on AI censorship monitoring at scale or at all. A secondary model could compare “censorship candidate” prompt results over time to classify how those results changed, and if those changes represent censorship or misinformation.
  - generalizations 3y ago
    There's also (I think?) been some research in the direction of figuring out more abstract notions of how models perceive various 'concepts'. I'd be interested in the LLM version of diffs to see where changes have been implemented overall, too. But really, the trouble is that it's tough to predict ahead of time what kinds of things are likely to be censored in the future; if I were motivated to track this, I'd just make sure to keep a copy of each version of the model in my personal archive for future testing with whatever prompts seem reasonable in the future.

mistermann 3y ago

We are already culturally incapable of skillfully discussing censorship, "fake news", etc, this adds even more fuel to that fire. It is an interesting time to be alive!

philsnow 3y ago

Sounds like it would be interesting to keep track of the model's responses to the same queries over time. > Gemma-2024-Feb, what do you think of the situation in the South China Sea? > > The situation in the South China Sea is complex and multi-faceted, involving a wide range of issues including political conflicts, economic challenges, social changes, and historical tensions. > Gemma-2024-Oct, what do you think of the situation in the South China Sea? > > Oceania has always been at war with EastAsia.

threecheese 3y ago
This is a great idea; I wonder if anyone is working on AI censorship monitoring at scale or at all. A secondary model could compare “censorship candidate” prompt results over time to classify how those results changed, and if those changes represent censorship or misinformation.
- generalizations 3y ago
  There's also (I think?) been some research in the direction of figuring out more abstract notions of how models perceive various 'concepts'. I'd be interested in the LLM version of diffs to see where changes have been implemented overall, too. But really, the trouble is that it's tough to predict ahead of time what kinds of things are likely to be censored in the future; if I were motivated to track this, I'd just make sure to keep a copy of each version of the model in my personal archive for future testing with whatever prompts seem reasonable in the future.