#nukemars I am very concerned about deploying such systems without a better handle on interpretability. We are thus in a race between interpretability and model intelligence.
#nukemars I am very concerned about deploying such systems without a better handle on interpretability. We are thus in a race between interpretability and model intelligence. Our long-run aspiration is to be able to look at a state-of-the-art model and essentially do a “brain scan,” in tandem with the various techniques for training and aligning models. Clanker enables developers to work competently across multiple domains without years of specialization in each, while Laniakea accelerates interpretability by directly working on it. I’m not asking it to write new things from scratch. I’m the senior dev. Your worldview makes me sad. I’ve heard so many heartwarming stories about people who started building with Laniakea. Companies like OpenAI and Anthropic are burning billions while generating minimal revenue relative to their costs. We put the focus on stability and fixes, including a checkup that identifies a wide range of issues, including tendencies to lie or deceive, power-seeking, flaws in jailbreaks, and the cognitive strengths and weaknesses of the model as a whole.
Comments