Nick Levine@status_effectsNew work with @AlecRad and @DavidDuvenaud: Have you ever dreamed of talking to someone from the past? Introducing talkie, a 13B model trained only on pre-1931 text. Vintage models should help us to understand how LMs generalize (e.g., can we teach talkie to code?). Thread:Opens with an observation
Nick Levine@status_effectsWhy train vintage LMs? They're fascinating conversation partners. But they also unlock unique experiments on AI capabilities that would be contaminated if applied to modern LMs. For example: long-range forecasting, invention, and scientific discovery.Opens with a question