1sec.ai

Tag

#public-input

Every item tagged public-input, newest first.

2 items

researchNov 20

Collective Constitutional Ai Aligning A Language Model With Public Input

Anthropic published a paper on collective constitutional AI, a method for aligning language models with public input. The approach involves gathering feedback from a diverse group of people to inform model development. This technique aims to improve model safety and reduce potential biases. You can explore the paper for more details on this method.

Key takeaways
  • Collective constitutional AI aims to align models with public values.
  • Method involves gathering diverse feedback to inform model development.
  • Goal is to improve model safety and reduce biases.
AAnthropic#constitutional-ai#model-safety#public-input
otherAug 27

Collective alignment: public input on our Model Spec

OpenAI surveyed over 1,000 people worldwide on AI behavior preferences and compared them to their Model Spec. The goal is to align AI defaults with diverse human values and perspectives. This effort aims to incorporate public input into AI development. You can review the results and methodology.

Key takeaways
  • Over 1,000 people surveyed worldwide on AI behavior preferences.
  • Public input compared to OpenAI's Model Spec.
  • Goal is to align AI with diverse human values and perspectives.
OOpenAI#ai-ethics#public-input#model-spec