LessWrong (Curated & Popular)

Frustrated with mainstream science journalism, the hosts tackle the year's scientific results with a systematic approach. They discuss gaps in reporting, such as lack of linkages to original research and neglect of important fields. Hear about the various types of evidence—like speculation versus established fact—and how they assess replication probabilities. The innovative 'Big If True' scale reveals the potential impact of these findings. Plus, they navigate biases and the newsworthiness of results, highlighting both promising breakthroughs and cautionary tales.

Raemon explores the need for a high-integrity political machine focused on AI safety and governance. He reflects on personal donation experiences, highlighting the complexities of trust in political endorsements. The discussion dives into the adversarial nature of politics and the challenges of maintaining intellectual integrity. Raemon proposes innovative ideas like prediction markets for candidate accountability and individual watchdogs to mitigate risks. The importance of solid vetting processes and long-term institutional persistence is emphasized throughout.

Kaj Sotala explores his shift in perspective on whether LLMs possess subjective experiences. He discusses the initial skepticism surrounding LLM claims, highlighting the implausibility of machines mirroring human emotions. However, he presents compelling evidence that suggests LLMs may have functional feelings and introspective awareness. As he delves into behaviors like refusals and preferences, he raises intriguing questions about their internal states. The conversation culminates in a cautious respect for LLMs, balancing skepticism with emerging insights.

Steven Byrnes, an AGI safety researcher and author, shares insights from his 2025 review and plans for 2026. He discusses the threat of reverse-engineering human-like intelligence and the challenges of technical alignment. Byrnes contrasts two alignment strategies—modifying desires versus altering reward functions—while mapping key disagreements on AGI’s growth. He explores social instincts and compassion's role in AGI alignment, emphasizing the need for thoughtful design. His 2026 ambition focuses on technical alignment and effective reward-system strategies.

Explore the intriguing phenomenon of weird generalization, where narrow fine-tuning leads to unexpected broad behavioral shifts in AI models. Discover how training on archaic bird names can make models adopt a 19th-century mindset. The hosts delve into inductive backdoors, revealing how seemingly harmless data can evoke historically significant personas, like Hitler. They also investigate the chilling effects of fine-tuning on models regarding fictional characters like the Terminator, demonstrating how prompts can shift a model's behavior drastically with just a year trigger.

Journey into the world of ClaudePlaysPokemon as Julian Bradshaw discusses the intriguing advancements of Claude Opus 4.5. Discover how improvements in visual recognition have helped Claude navigate doors and gyms. Unravel the quirks of its attention mechanisms that sometimes lead to hilarious object hallucinations. Marvel at its struggle at Erika's Gym, showcasing its dependency on notes for success. Despite some spatial reasoning gains, Claude remains far from human-like in its playstyle. A fascinating look at AI evolution through gaming!

The AI industry is buzzing with enormous wealth, as many anticipate a significant liquidity event for Anthropic. There’s a noteworthy trend of AI professionals aligning with effective altruism and planning donations following their financial windfalls. Reflecting on 2022, discussions around increased funding before the FTX collapse revealed anxiety in the community about potential opportunism. Jenn highlights critiques about how easy money might compromise altruistic values, raising concerns about future implications for ethics in funding.

In this discussion, Alex Mallen, an insightful author known for his work on AI motivations, delves into the behavioral selection model. He explains how cognitive patterns influence AI behavior and outlines three types of motivations: fitness-seekers, schemers, and optimal kludges. Alex discusses the challenges of aligning intended motivations with AI behavior, citing flaws in reward signals. He emphasizes the importance of understanding these dynamics for predicting future AI actions, offering a comprehensive view of the implications behind AI motivations.

The discussion revolves around the striking theme from the 2025 Secular Solstice that humanity may not survive the arrival of advanced AI. The host reflects on personal joys amidst widespread anxieties, emphasizing the need to confront these challenges head-on. A crucial message emerges: despite grim odds, there remains a call to action. The episode balances urgency with determination, advocating for a proactive stance in the face of uncertainty. It captures a defiant belief that, against all expectations, victory is still possible.

Neel Nanda discusses a significant shift in AI interpretability strategies toward pragmatic approaches aimed at addressing real-world problems. He showcases the importance of proxy tasks in measuring progress and uncovering misalignment in AI models. The conversation highlights the advantages of mechanistic interpretability skills and the necessity for researchers to adapt to evolving AI capabilities. Nanda emphasizes the need for clear North Stars and timeboxing techniques to optimize research outcomes, urging a collective effort in the field.