Swiss AI Safety Summer Camp
About the Camp
According to 80,000 Hours, reducing risks from advanced artificial intelligence might be one of the most effective ways to improve the world.
Our Camp is designed to provide a gateway into the world of AI Safety and create an enjoyable experience for all participants.
With a focus on the key sub-domains of AI Safety, including conceptual aspects of the alignment problem, mechanistic interpretability, and governance, we offer a multidisciplinary learning experience through technical activities such as deep learning courses, paper readings, and discussions, presentations, and lectures.
We would be particularly enthusiastic to receive applications representing a wide range of diversity. Women, racialised people, people with disabilities and LGBTQ+ people are strongly encouraged to apply.
Fact Sheet
Date: 4th - 16th of September 2023
Location: We will be in juhui Melchtal (which is close to Luzern)
Cost: Free (incl. housing & meals; covering transportation costs is also possible)
PREREQUISITES TO ATTEND
Higher education level on:
Linear algebra (matrix operations, eigenvalues, eigenvectors, linear subspace)
Analysis (multivariable calculus)
Probability (random variables, expected values, conditional distributions, Bayes theorem)
Python-Coding (NumPy, basic OOP, math programming)
Basic Deep Learning Knowledge (feed forward neural network, gradient descent algorithm) - be comfortable with the concepts of video 1-3 of this 3Blue1Brown playlist
(If you don't have these prerequisites, we still encourage you to apply. We are able to send you preparation-resources upfront)
APPLICATION
Applications are now closed.
TENTATIVE PROGRAM
Deep Reinforcement Learning
Technical overview of Deep Reinforcement Learning and presentation of the conceptual AI Safety framework associated.
Convolutional Neural Networks
Implementation exercises of interpretability techniques on CNN.
GPU and Compute Governance
Exploration of GPU usage and computational resource governance in AI systems.
The Transformer Architecture
Overview of the technical implementation of the Transformer architecture, conceptual differences with the RL framework, implementation of interpretability tools and presentations of edge-cutting research in the field.
Conceptual readings
Part of the AI Safety Fundamentals Alignment Track, including instrumental convergence, scalable oversight, interpretability, the main research agendas.
Governance
Part of the AI Safety Fundamentals Governance Track.
Group Projects
Dedicated time for exploration of specific subjects of interest within AI alignment.
Support for Independent Projects
Guidance to pursue independent projects aligned with personal interests and aspirations during the last part of the camp.
Conferences on AI Risks
Three conferences, including Q&As, given by leading researchers on AI risks.
Leisure and Recreational Activities
FAQ
How many people will attend the camp?
About 20 participants plus staff/cooks/teachersWill there be free time?
We take learning seriously, and therefore rest as well. There will be periods without a program each day. And there will be some free days.Are there other editions of this camp?
Yes, there are camps in France, Germany and Denmark with similar contents.
We are also running another edition if this camp in the winter semester, which will likely take place in February.Is this camp only intended for people living in Switzerland?
No, attendees from any country can join the camp. You just need to be physically present in Switzerland for the camp.Is the camp only intended for students?
No, we are very happy to welcome attendees with any background.Are there other courses/formations that would replace this one?
Yes, you can find a list there, there and plenty other links here.Is this camp similar to aisafety.camp?
No, we share a very similar name, but we are not affiliated with this camp, and the program & scope of both camps are quite different. We are sorry if this caused confusions to you.Further questions? Hesitations?
Please do not hesitate to reach out to us at hello@aisafetycamp.ch!
Help us Spread the Message!
Do you know a group of people who might be interested in joining, but to which we have not reached out yet? We would love to hear which group you have in mind! Please get in touch with us at hello@aisafetycamp.ch.
About Us
The Swiss AI Safety Summer Camp is run by a group of volunteers who believe AI system may pose a serious threat.
Please feel free to contact us at hello@aisafetycamp.ch.
FURTHER IMPRESSIONS
(homepage will be updated continuously)