Research

Ought is somewhere between a research lab and a startup.

We do conceptual and empirical work on supporting thinking and reflection using machine learning, and share our findings openly. This research is guided by a vision for tools and applications that we hope will eventually help millions of people think through the questions and choices they face every day.

We're incorporated as a non-profit, so we can afford to take the long view.

Elicit

Our core project is Elicit, the AI research assistant:

Elicit helps people answer research questions by making qualitative reasoning steps explicit and using language models to incrementally automate those steps.

Over the next decade, language models will increasingly automate aspects of human thought, but they won't automatically help us make sense of the world. There's a gap between providers of large language models (e.g. OpenAI, Anthropic, AI21) and researchers whose work informs corporate and government policy (e.g. CSET, CSER, FHI). To bridge this gap, we study and automate research workflow steps.

You can watch Elicit screencasts on Youtube.

Other research

We've previously worked on:

Integrate model-based and judgmental forecasting so that we get the best of both worlds, automated forecasts that are sensitive to qualitative arguments
Solve difficult problems by assembling small and mostly context-free contributions from individual agents who don't know the big picture