Why AI Safety?

AI systems have a potential to massively benefit humanity. However, the development of these systems poses equally incredible risks. The risks posed are existential for the human species.

Ensuring any AI system’s goals are aligned with ours is a non-trivial task. Most notably, we only get one try to get this right.

What we do

We think that the best way to make progress on the problem of AI safety is getting more people to focus their careers on AI safety research, in fields like mechanistic interpretability, adversarial robustness, or AI governance.

All of our curriculum is available online.

Click here for fellowships, and click here for the upskilling tracks.

If you’d like see some projects we’ve worked on in the past, click here.

Resources

As one of the forefront AI clubs at UCLA, we adminisitrate the largest GPU server of any UCLA club. Members get access to train models on our GPUs (NVIDIA 3090s). Check below to see a live feed of our GPU utilization (updates every 10 minutes).