Nate Soares recommends pairing up when studying, so I figured it would be useful to facilitate that.

If you are looking for a study partner, please post a top-level comment saying:


  • What you want to study
  • Your level of relevant background knowledge
  • If you have sources in mind (MOOCs, textbooks, etc), what those are
  • Your time zone


I have been studying meta-research (a la METRICS, Cambridge Handbook of Expertise, Kuhnian revolution etc.) and while I'm not looking for a study partner per se (my schedule is very sporadic) I would be interested in diffing models about this topic with anyone who has done some of their own investigation in the area.

I'm a postdoctoral scholar at METRICS and I'd be happy to talk to you about this. Get in touch by e-mail or private message. Also, I'm giving a talk about a new research idea at the METRICS internal lab meeting this coming Monday at 12:00 at Stanford. You are welcome to attend if you want to meet the METRICS group (but the professors are probably going to be busy and may not have time to talk with you)


I have two areas I'd like to study: deep learning, and anything on the MIRI research guide. Lots of material is available on both topics, but I'd like to pair up with someone to build a good learning strategy (for lack of a better expression).

I have some knowledge of algebra, probability theory, logic, game theory, machine learning (Master's Degree in Computer Science).

Regarding deep learning, I have a small collection of links, Udacity, and I'm positive learning materials abound now that the field is really popular.

Regarding MIRI's research guide, well, the guide itself provides a lot of links and pointers.

My timezone is CET (UTC+1).

Hi, I'm an AI PhD student and I just signed up for the Udacity Deep Learning course. Lets do this!

I'm going to apply for AI research related PhD this year. I want to start some research project in the near future with a goal of learning and increasing the chances of successful PhD admission. It's very likely that the domain of this research project will lie close to ML or MIRI research agenda.

I have only a bachelor degree in Engineering (CS and Software Engineering). I work as a software engineer and spend evenings by preparing for GRE, thinking and learning about FAI. Probably will do something with my job to free more time. My timezone: UTC+6.

I forgot to mention I was currently an AI PhD student. Which doesn't entail much free time ^^

So... what exactly are you interested in learning (if you want to pair up)? I'm also interested in your project, if you have an idea in mind.

I'm working through the udacity deep learning course right now, and I'm always trying to learn more things on the MIRI research guide. I'm in a fairly different timezone, but my schedule is pretty flexible. Maybe we can work something out?

I just finished Stanford's machine learning class on Coursera and I was thinking about starting Google's Udacity course.

I don't have much formal background in CS (2 classes in college and later a couple Coursera classes), but I've been working as a software engineer for a few years now.

I am in U.S. Eastern Time (UTC-4).

Machine Learning for Good is A machine learning and deep learning study group for EAs and rationalists that I'm facilitating.

It includes a study group for the current Udacity Tensorflow/Deep Learning course. I'm not going to repost further info here, one can access it through the following group:

Everyone is allowed to join. It's closed so that technical or controversial discussions are not broadcast to friends of all members who have not chosen to join.

Four meetings in, people have met, shared updates on study projects, shared updates on Kaggle competitions, talked about making study groups and Kaggle teams. People are more informed about and seem more interested in AI (safety) progress. I'm not that emotionally committed to its continuation but it seems like enough high-potential people with a shared interest are meeting that good things will eventually emerge on an AI safety front.

Is the end-game to do data-analysis on data for charity evaluation, intervention evaluation, cost-effectiveness and that kind of thing?

Or, to inform people interested in machine learning about AI safety?

It's deliberately not just for AI safety, but half of the people are interested in AI safety currently.

As well as promoting interest in these two areas among people with AI knowledge, the aim is to promote knowledge in people who care.