
What we’re about
Join us for a variety of events on technical AI safety, governance in a world of advanced AI, and more.
Hosted by Trajectory Labs, a nonprofit coworking and events space catalyzing Toronto's role in steering AI progress toward a future of human flourishing.
Is there a topic you'd love to see us cover at a future event? Submit your suggestion here.
Upcoming events (4+)
See all- Hackathon: Alignment Faking Model Organisms30 Adelaide East, Industrious Office 12th Floor Common Area, Toronto, ON
Important registration information: To participate in this event, please join the discord link before registering.
Many safety and governance measures rely on AI models showing us their true colours. "Alignment faking" is the phenomenon of a model hiding misaligned behaviour when it believes it's being observed.
In this hackathon, we will be constructing model organisms of alignment faking: realistic, experimentally-verified pathways under which alignment faking can occur. We'll be test-driving a new framework for alignment faking experiments. The environment, monitoring and scoring are already set up - all we need to do is supply the models! These can be fine-tunes of open source models or simple prompt engineering.
Trajectory Labs, the jamsite, provides a comfortable and spacious coworking space along with coffee, tea, and other refreshments (meals not provided, but there are many nearby options). Other locations will also be taking part!
Bring a laptop (beefy GPUs are not necessary, we'll provide credits for API-based finetuning of open source models so you don't need to run them locally).
*More details and resources to come, including some useful background reading on model organisms and alignment faking.
- AI Policy Tuesdays: Does AI diffusion undermine the US-China race?30 Adelaide East, Industrious Office 12th Floor Common Area, Toronto, ON
Registration Instructions (IMPORTANT)
This is a paid event ($5 general admission, free for students) with limited tickets - you must RSVP on Luma to secure your spot.If you can't make it in person, feel free to join the live stream at 6:30 pm, via this link.
Description
Pushing to develop transformative AI as fast as possible risks catastrophe—a concern often met with the objection that “if we don’t do it, China will.”This “but China” objection often doesn’t give much consideration to diffusion: the speed and way in which AI is integrated into the military and economy. Wim Howson Creutzberg will explore how diffusion might change the stakes of racing towards transformative AI.
Timeline
6:00 to 6:30 - Food & Networking
6:30 to 7:30 - Main Presentation & Questions
7:30 to 8:00 - Discussion - AI Safety Thursdays: Can we make LLMs forget? An intro to machine unlearning30 Adelaide East, Industrious Office 12th Floor Common Area, Toronto, ON
Registration Instructions (IMPORTANT)
This is a paid event ($5 general admission, free for students) with limited tickets - you must RSVP on Luma to secure your spot.If you can't make it in person, feel free to join the live stream at 6:30 pm, via this link.
Description
LLMs are pre-trained on a large fraction of the internet. As a result, they can regurgitate private, copyrighted, and potentially hazardous information, causing deployment and safety challenges.Lev McKinney will guide us through machine unlearning in LLMs—how models retain facts, methods for identifying influential training data, and techniques for suppressing predictions. Finally, we'll assess current research and its effectiveness for policy and safety concerns.
Timeline
6:00 to 6:30 - Food & Networking
6:30 to 7:30 - Main Presentation & Questions
7:30 to 8:00 - Discussion - AI Policy Tuesdays: The Effects of Synthetic Data on Power Structures30 Adelaide East, Industrious Office 12th Floor Common Area, Toronto, ON
Registration Instructions (IMPORTANT)
This is a paid event ($5 general admission, free for students) with limited tickets - you must RSVP on Luma to secure your spot.If you can't make it in person, feel free to join the live stream at 6:30 pm, via this link.
Description
Savannah Harlan explores how the rise of synthetic data is redrawing the boundaries of power at three levels: between governments and the public, between ordinary people and corporations, and between humanity and artificial intelligence itself.We'll finish by examining whether synthetic data is likely to increase x-risk as we shift more of the work of alignment onto the models themselves.
Timeline
6:00 to 6:30 - Food & Networking
6:30 to 7:30 - Main Presentation & Questions
7:30 to 8:00 - Discussion