The Nonlinear Library

Seth Ariel Green, coauthor of a research report on sexual violence prevention programs, discusses the need for a more behavioralist approach in addressing environmental and structural determinants of violence. The podcast explores the gap between attitude and behavioral change in prevention programs, the effectiveness of building-level interventions, and the importance of shifting societal perceptions to reduce violent behaviors.

Johnny Lin, a researcher in Sparse Autoencoders, talks about the Neuronpedia platform for SAE research. Topics include hosting models, data visualizations, collaboration opportunities, and the importance of mechanistic interpretability. They discuss the challenges, tools like Transformer debugger, and the implications of SAEs for AI alignment and dual-use risks in technical AI safety.

Author Jeffrey Heninger discusses his detailed notes and commentary on Secular Solstice, exploring the concept of whig history and expanding circles of moral concern. The podcast delves into themes of gratitude, existential risks, Spinoza's philosophy, hope, futuristic ambitions, nationalism, and sacred causes.

Rob Gledhill, author of the announcement on the future of Wytham Abbey, discusses the closure of the project, sale of the property, and donation of proceeds to high-impact charities.

Sam Altman discusses risks, theatrical tactics, and loss of control scenarios. The conversation delves into board dynamics, decision-making, AI advancements, and future governance. Altman's views on capitalization, simulations, aliens, and engaging conversations are highlighted.

A podcast discussing the ethical dilemma of dealing with a moth infestation using non-lethal methods in a London flat. The speaker explores the struggle of balancing humane treatment with the necessity of pest control, reflecting on the moral implications and inner conflict of exterminating the insects.

Exploring EA Sweden's achievements in 2023, their goals for 2024, and the challenges they face, including a funding gap. Delving into community-building efforts, career guidance support, project impact assistance, diversity strategies, and Nordic EA collaboration. Addressing the importance of financial support to scale activities and drive impact.

Discussing the confusion between inner and outer alignment in narrow AI, exploring the challenges of defining reward functions, and the shared responsibility of ethical AI behavior in training processes.

Explore the benefits of preserving misaligned AGI for cooperation, incentivizing alignment. Discuss challenges in detecting misalignment in early AGI, impact of patternism on consciousness. Learn about strategies to improve AGI relations despite low success rates.

A discussion on concave and convex agents in decision theory. Different utility functions shape preferences for resources. Humans tend to have concave utility functions, valuing resources more when they have less. Insights into risk aversion, human behavior, and trade scenarios involving diverse agent types.