Pinned Loading

1. filtering-for-misalignment filtering-for-misalignment Public

   What training data should developers filter to reduce risk from misaligned AI? I propose that AI labs filter information about safety measures and strategies for subverting them. This repository he…

   Python 4

2. how-bad-can-ai-be how-bad-can-ai-be Public

   Diffuse control project: how powerful of an affordance is training in diffuse control?

   Python

3. misalignment-by-default misalignment-by-default Public

   Model organisms research: do AI goals drift due to "catastrophic forgetting"? Does alignment drift?

   Python 2 1

4. DQN-maze-solver DQN-maze-solver Public

   Investigating whether or not RL agents can acausally collaborate with other instances of themselves.

   Python 1

5. transformer-shortest-paths transformer-shortest-paths Public

   Experimentally evaluating transformer's generalization on a synthetic task

   HTML 1

6. theland theland Public

   A multiplayer video game that I made back in highschool

   JavaScript 2