My lectures on AI risk and long-term AI safety

This week in Europe has been the darkest in a very long time. My thoughts go primarily to the brave Ukrainian people with wishes for success in their defence against Russia's criminal assault, and secondarily to those barve anti-war protestors who take to the streets in cities all over Russia.

In the shadow of all this, I have given a series of three two-hour lectures at CHAIR (Chalmers AI Research Centre) about AI risk and long-term AI safety, organized as follows:

1. How and why things might go wrong

2. Timelines, natural language processors and oracle AI

3. Research directions in AI alignment

Here are video recordings of the three lectures: