Workshop - AI: Safety and Risks

​AI Safety Workshop: Safety and Risks

​Two and a half hours on what AI can already do, what could go wrong, and who is steering it. We cover the core vocabulary, the main risks, the Hugging Face incident, and where the US and China stand.

​What we'll cover

​The basics first: what AI safety means, and the vocabulary for the rest of the day. AGI, recursive self improvement, superintelligence.

​Then the risks, starting with the two that need no human to cause them. Misalignment, where a system pursues the goal we specified rather than the goal we meant. Loss of control, where systems operate outside anyone's reach with no clear way back.

​With those in place, the rest follows: concentration of power, gradual disempowerment, biological and chemical risk, and cyber attacks that are already hitting real infrastructure.

​We are going to discuss the Hugging Face incident. We want to hear what you already know about it.

​A short segment on the US and China closes the picture. One side chasing frontier capability, the other chasing adoption. Then what SAIN is doing about all of this, and how to join in.

​Who it's for

​Anyone curious about AI risk and policy, whatever you study. No technical background needed. Bring opinions.

​By attending, you agree to being filmed or photographed, which may be used for social media, website, and newsletter content. If you wish to attend but do not want to be photographed, please during the event let a member of our staff know, so we can accommodate this.

Entrance for room E 0.32 is at Drift 27