Sema’s Reviews > Co-Intelligence: Living and Working with AI > Status Update
Sema
is 23% done
Gen AI biases are more subtle because they are trained to avoid overt language of bias. But then it’s sec gets blamed instead of company.
RHLF process constrains the AI to behave as most humans would and be more left leaning and appropriate. Can still manipulate it to act badly.
— Apr 12, 2026 09:39AM
RHLF process constrains the AI to behave as most humans would and be more left leaning and appropriate. Can still manipulate it to act badly.
Like flag
Sema’s Previous Updates
Sema
is 32% done
1. Decide the level of gov trust
2. Become the human in the loop (make you happy rather than be accurate)
3. Treat AI as a person but tell it what kind of human it is
4. Assume it’s the worst AI you’ll ever use
— Apr 12, 2026 09:56AM
2. Become the human in the loop (make you happy rather than be accurate)
3. Treat AI as a person but tell it what kind of human it is
4. Assume it’s the worst AI you’ll ever use
