Welcome to my personal blog. The main purpose of this blog is to serve as a location for me to clarify my thoughts. I think it’s also a good way of connecting with my friends, and sharing ideas that I find important or helpful.
The posts approximately fall into three categories: (1) technical notes, (2) life-optimization, (3) philosophy. Below I suggest “entry points” into exploring this blog.
0stack: This is a collection of serialized posts. This is the best place to go to understand “what has Alek been thinking about lately”, besides actually talking to me, which is probably a better method.
goodness(universe): What is goodness? What do I want my life, and the universe at large to look like?
Some examples of technical notes:
backdoors and deceptive alignment, algorithm class notes
AI alignment: Over the last decade, machine learning has made immense progress on a wide array of tasks. What happens if this progress continues into the future? Many AI experts predict (and there are scaling models to back up this prediction) that we are only 5-10 years away from having AI agents that could, e.g., automate a substantial portion of all human labor. However, I don’t think humanity is ready to create a new, more intelligent, species than ourselves. In particular, I think that the current trajectory of AI development poses an existential risk to humanity — by default, in creating more powerful entities than ourselves, we relinquish our control over the future to the more powerful entities, and we don’t have good techniques right now for ensuring that these entities care about making a good future. In this sequence of posts I’ll write about some potential risks from advanced AI agents, and talk about some ideas for mitigating these risks.