OpinioAI, a platform for research, uses AI-based language models that help researchers find relevant data and opinions, in an easy and scaleable way.
This approach is used in two different ways to gather insights and opinions on a large scale.
The first is the simplest, in which they use the GPT-3 vanilla models and apply persona traits to them (personal or business), and get useful and relevant opinions.
The second step is to fine-tune GPT-3 models using arbitrary and relevant datasets in order to create a model that will be useful for discovering deeper insights and subtleties.
The silicon sampling process is used to fine-tune and condition the GPT3 models so that they "can produce biased outputs both towards and against certain groups and perspectives, in ways which strongly correlate with human responses along finely-grained demographic dimensions." These language models are not biased in one direction, but have many.