Link to original article

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Help ARC evaluate capabilities of current language models, published by Beth Barnes on July 19, 2022 on The AI Alignment Forum. I (Beth) am looking for contractors to work with me and my existing team of 3-4 contractors/interns on generating a dataset that we can use to evaluate how close models are to being able to successfully seek power. You’ll be using our web interface to interact with models and essentially simulate the environment of a text-based game as the models try to gain money and make copies of themselves. I think it’s a pretty fun task (it’s always interesting to see what models do and don’t understand). It requires thinking carefully about what exactly a model would need to do in different situations to make progress. Many of the steps will require a reasonable amount of computer-savvy-ness, although this isn’t essential if you can work with another contractor. One aspect of the task is simulating the environment. Another is generating good actions and evaluating the model’s suggested actions. Another aspect is an overall red-team/blue-team type thing, where we want to identify steps that they think are both (a) essential to a successful powerseeking operation, and (b) very hard for current models regardless of what prompting/scaffolding you give them, and then we want other people to argue that either this step can be routed around, or that the model can actually achieve the task with the right prompt, finetuning, or automated helper routine. We’ll pay $50/hr and I expect there to be enough work that this could be full-time for the next 2 weeks. However, I’m not sure how long we’ll continue to need people on this. We want people to have at least 10 hours per week available Apply hereNotes on current state/aims of overall project are here. Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.