Title here
Summary here
Last week, we hosted the first event of the AI Safety Talks series in collaboration with inzva!
We teamed up with inzva for insightful discussions on AI safety featuring our co-founders, Bengüsu Özcan and Berke Çelik, followed by a thought-provoking session by Erdem Bıyık on Reinforcement Learning from Human Feedback.
🎥 You can watch the session here!
ANNOUNCEMENTS 🔊
SERI MATS Program
SERI MATS, arguably the most prestigious research upskilling program in AI Safety has opened applications Independent research and educational seminar program connecting talented scholars with top mentors in the fields of AI alignment, interpretability, and governance. For 10 weeks, MATS scholars will conduct research while also attending talks, workshops, and networking events with other members of the Berkeley alignment research community.
🗓️ Register by: April 18th
Economics of Transformative AI Hackathon
Join Apart Research’s new hackathon that brings technical insights with policy ideas: what the economy should look like in the age of transformative AI?
🗓️ Register by: April 25th
Bluedot Impact Courses
Join 9-12 weeks of online training and discussion sessions, tailored for cutting-edge, transformative AI topics. Choose the topic for your background: technical AI alignment, AI governance, or transformative AI economics.
🗓️ Rolling applications
Networking for impact in AI Safety
Networking event from BlueDot Impact aimed at sparking meaningful conversations, uncovering powerful collaborations, and accelerating your personal and collective impact in the field. The session will blend purposeful 1-on-1 conversations with speed networking structured based on attendee input.
🗓️ Register by: April 17th
AI Alignment Evals Hackathon
We want AIs to have certain values in robust and scalable ways. There are methods proposed to get these values into models, but it’s highly ambiguous how well they actually work. This week-long hackathon from AI-Plans aims to help remove that ambiguity.
🗓️ Register by: April 25th
LASR Labs
Research program working towards reducing the risk of loss of control from advanced AI, focusing on action-relevant questions tackling concrete threat models. Participants are matched into teams of 3-4 and work with a supervisor to write an academic-style paper, with support and management from LASR. The program will be in person, in London.
🗓️ Register by: April 26th
MAISU - Minimal AI Safety Unconference
A flexible, participant-driven unconference for those working to prevent AI catastrophe, organized by the AI Safety Camp team.
🗓️ Register by: April 18th
AI Futures World-building Course
Foresight Institute just launched a free, self-paced course designed to help more people think clearly and creatively about how AI might shape the future. This course is for people who want to explore how AI might shape our shared future and what we can do about it. It is designed to be accessible and thought-provoking, with no technical background required.
🗓️ Register by: April 17th
TOP PICKS 📑 🎧
GATE: Modeling the Trajectory of AI and Automation
If you’re seeking credible analysis on AI progress, Epoch AI continues to be our recommended source. In their new paper, Ege Erdil et. al builds a general framework for modelling the relationship between AI progress and its economic impacts, and provides forecasts regarding how AI will transform the economy. Their research shows that full automation is quite plausible within a decade, and even with conservative assumptions, full automation seems quite likely within two decades.
AI 2027: A Plausible Scenario For Super Intelligence In This Decade
A group of researchers, including ex-OpenAI employee Daniel Kokotajlo known for his accurate forecasts for frontier AI development so far, has published an extensive scenario of a technical and geopolitical pathway that leads up to artificial super intelligence in this decade. This project provides a great opportunity for you to inform your own opinion: what are the dependencies, where you agree and disagree.