EAG Talks

Join the organizers for a brief welcome and a group photo of attendees, followed by a short talk from Ben West. Ben is the Interim Managing Director of the Centre for Effective Altruism (CEA) and he is responsible for overseeing CEA’s work during the transition to new permanent leadership. He will speak on the current state of the EA movement and possible directions for its future. Watch on Youtub...

In this session, Ryan Ritterson will use examples of his and others’ work at Gryphon to tell stories about how to effectively influence public policy. Based on Gryphon’s data-driven approach, he’ll provide key lessons learned and takeaways for others interested in influencing policy. Along the way, he’ll also talk about two recent Gryphon efforts, including one focused on developing and securing p...

This talk examines whether advanced AIs that perform well in training will be doing so in order to gain power later — a behavior Joe Carlsmith calls "scheming" (also often called "deceptive alignment"). This talk gives an overview of his recent report on the topic, available on arXiv here: https://arxiv.org/abs/2311.08379 . Joe Carlsmith is a senior research analyst at Open Phi...

Should you be able to order smallpox DNA in the mail? Biosecurity professionals have argued for almost 20 years that synthesis companies should screen orders so that pathogen and toxin sequences are only sent to people with a real scientific use for them. Now, it seems like fears of AI-engineered pandemics may spur governments to make screening mandatory. Tessa will discuss why securing nucleic ac...

Media coverage of catastrophic risk is going mainstream; our panel of media professionals discuss the ways it's going right vs going wrong, and what actions they are taking to focus the conversation on well-reasoned risk models and effective interventions. On the panel will be Kelsey Piper of Vox Future Perfect, Shakeel Hashim of the AI Safety Communications Centre, and Clara Collier of Asteri...

Hayley Clatterbuck will summarize key findings from Rethink Priorities' ""Causes and uncertainty: Rethinking value in expectation"" (CURVE) project, which evaluated the cost-effectiveness of existential risk mitigation projects under both standard expected utility maximization and risk aversion. Persistence of effect and future growth trajectory are the strongest contributo...

The five years since Toby Ord wrote The Precipice have seen dramatic changes to the landscape of existential risk. Ord will explore the biggest changes to the biggest risks, showing how new developments have upended key assumptions and pushed some of these risks into new phases. And we’ll see how the world has woken up to the very idea of existential risk, with it becoming an explicit priority and...

If an AI system learned a deceptive strategy, could we detect it and remove it using current state-of-the-art safety training techniques? That's the question that Evan and his coauthors at Anthropic sought to answer in their work on ""Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training"", which Evan will be discussing. Evan Hubinger leads the new Al...

The final session of the conference will include some closing words, followed by a talk and fireside chat with Dan Zimmer. Dan Zimmer completed his Ph. D. from the Department of Government at Cornell University. His research focuses on the implications that anthropogenic existential risk (x-risk) poses for some of the foundational categories of Western political thought, paying particular attentio...

Pandemic security aims to safeguard the future of civilisation from exponentially spreading biological threats. In this talk, Kevin will outline two distinct scenarios–"Wildfire" and "Stealth"–by which pandemic-causing pathogens could cause societal collapse. He will then explain the ‘Delay, Detect, Defend’ plan to prevent such pandemics, including the key technological program...

Content warning: This presentation contains images some may find distressing. In this talk Cynthia Schuck, Research Director of the Welfare Footprint Project, describes their approach to quantifying animal suffering, and how it can be used to measure the impact of interventions and inform policies. She also addresses major gaps in welfare research and in our understanding of suffering. The talk co...

This will be a panel discussion on nuclear policy, deterrence, inadvertent escalation, entanglement, emerging technologies, and related topics. James Acton holds the Jessica T. Mathews Chair and is co-director of the Nuclear Policy Program at the Carnegie Endowment for International Peace. A physicist by training, Acton is currently writing a book on the nuclear escalation risks of advanced nonnuc...

Join the organizers for a brief welcome and a group photo of attendees, followed by three short talks from key community members. We will hear remarks from: Lizka Vaintrob Kuhan Jeyapragasan Arden Koehler Lizka runs the EA Newsletter, the EA Forum Digest, and the non-engineering side of the EA Forum at the Centre for Effective Altruism. Kuhan currently runs the Cambridge Boston Alignment Initiativ...

Reinforcement Learning from Human Feedback (RLHF) has emerged as the central alignment technique used to finetune state-of-the-art systems such as GPT-4, Claude-2, Bard, and Llama-2. However, RLHF has a number of known problems, and these models have exhibited some troubling alignment failures. How did we get here? What lessons should we learn? And what does it mean for the next generation of AI s...

We want to help others as much as we can. Knowing how is hard: there are many empirical, normative, and decision-theoretic uncertainties that make it difficult to identify the best paths toward that goal. Should we be focused on sparing children from vitamin deficiencies? Reducing suffering on factory farms? Mitigating the threats associated with AI? Should we split our attention between all three...

Associate Director of New Program Development and Cost-Effectiveness, Kevin Kelsey, will give us an inside look at Evidence Action's Accelerator - their engine for new program development. The Accelerator develops new programs using a six-stage, decision-focused process designed to scale only the most cost-effective, evidence-based interventions. He'll provide insight into how Evidence Act...

The final session of the conference will include some closing words, followed by a talk on averting future pandemics from Dr. Matthew Meselson. Dr. Meselson will discuss past pandemics, the mode of transmission of the particular pathogen, measures to prevent such transmission, and the practical politics of implementing such measures. Dr. Meselson has conducted research at Harvard University mainly...

In this talk, we will cover: why we need to spread the ideas of effective altruism, the principles Giving What We Can uses to spread EA ideas, what has worked well in the past, some preliminary results and insights from recent marketing tests completed and how individuals can best share these ideas.

Join Elise Bohan (Future of Humanity Institute) and Michael Aird (Rethink Priorities) as they discuss how to use stories to get people to care about the long-term future. Elise will reflect on her book Future Superhuman and the importance of stories when talking about transhumanism.

Join Kieran Greig from Farmed Animal Funders and Emma Hurst from the Animal Justice Party as they discuss animal advocacy. They intend to talk about a range of topics including recent achievements, the challenges that come with politics, and looking to the mid-to-long-term future of advocacy.

In this talk, Magdalena explains different research directions in technical AI safety. The talk is for you if you already believe that AI safety is an important problem, and want to get a better understanding of current technical research. https://www.youtube.com/watch?v=fKZahM44KQA

Join us for a brief welcome and opening talk by Carolin Basilowski (Team Lead for EAGxBerlin), and the EA Germany Directors Sarah Tegeler and Patrick Gruban.

This talk covers the fundamentals of mechanistic interpretability: what it is, why it might be impactful for alignment, and how you can get involved. It is most useful for people who are either new to AI safety, or limited knowledge about mechanistic interpretability. The speakers hope you gain insights into what interpretability research can look like, and whether it is a good fit for you. They a...

Brief 8-minute talks about a range of topics. David Quarel: Intro to AI safety James Fodor: AIs versus Humans: Similarities and Differences between Human and Machine Learning https://www.youtube.com/watch?v=btPVqYcBihM

Brief 8-minute talks about a range of topics. Amanda Storm Schuster: Don't forget about fundraising! A career path making a difference David Quarel: A summary of state-of-the-art AI