Research update: Vacation and CAIF

[Ram Rachum]

Hi everyone!

I'm back from my vacation! I spent one week in London and then two weeks in New York. I had a great time. I did lots of the usual tourist stuff, especially visiting art museums. (Loved this one.) I met some friends in both cities. In London I met Edgar and Georg for the first time, which was very exciting.

I landed back in Tel Aviv a week ago. The Post-Vacation Vacation Syndrome is real... It's hard to get back into my routine and be productive again. The mountain of errands that's accumulated over three weeks certainly doesn't help.

Retrospective on goals for April

In the April update I outlined a list of goals. Here's my update on these goals:

1. Have fun on my vacation: ✅ Done

2. Experiment with existing RL frameworks: ↷ Postponed I couldn't find time and energy to continue learning Gym, so I'm going to do it in May instead.

Stuff I've done beyond the goals

In the March update I was talking about trying to get funding for my research from various foundations, and in the April update I postponed that task, because I'd rather focus on advancing my research some more, and think about money in a few months from now. But in this last month, I've started spending more time looking into the Cooperative AI Foundation. From their website:

As announced in a recent Nature commentary, the Cooperative AI Foundation (CAIF) is a new charitable entity, backed by an initial philanthropic commitment from the Center on Emerging Risk Research of $15 million. CAIF's mission is to support research that will improve the cooperative intelligence of advanced AI for the benefit of all humanity. For updates and funding opportunities, please sign up to the Cooperative AI mailing list.

(CAIF is pronounced as "safe", which is a cute way to remind us that they're focusing on creating safe AI.)

There's a lot I can get from CAIF besides funding. They hold a seminar series which I'm now attending regularly. Experts from different universities and corporate labs are giving talks about their research and inviting discussion and questions. Many of them are in MARL or close to it. If this sounds interesting to you, consider attending! You can see the list of past seminars and their videos. Sign up to get notifications of new seminars at the bottom of this page.

I think that could be a great opportunity for me to meet MARL researchers, learn about their research and get a better calibration on my research direction. I want to get more involved with CAIF, so this is going to be a goal for May. More details about it below.

My goals for May

1. Experiment with existing RL frameworks. This is the same old goal from the April update.

2. Get more involved with CAIF. In the section above I explained about CAIF and why it's a good opportunity for me. I want to apply for a grant from them, and I want to give a talk in their seminar series. However, I feel like I need a few more months to prepare before I can be a serious candidate... I'm conflicted here, because on one hand I want to come prepared, but on the other hand I have FOMO. I feel that things are already happening without me and it's scary. I need to think about it I guess.

   I'm also concerned that the fact I'm not an academic will make me less attractive to CAIF. Everyone there are experienced researchers with PhDs. I'm still a beginner. I'll have to work harder to show that I can make a meaningful contribution.

   Here are a few important points I should work on:

   • I should read their paper Open Problems in Cooperative AI so I could have an intelligent conversation around what's known and what isn't known. This paper has like a hundred citations, so I'll have to be judicious in how deep I want to go down that rabbit hole.

   • I should keep attending the seminar series.

   • I should try to talk to a few people one-on-one, so I could get feedback and calibrate myself.

3. Possibly get interviewed for an Israeli podcast. I've been keeping silent about this one because I don't want to jinx it. It's possible that in May I'll be interviewed about my research for a popular Israeli podcast. These interviews sometimes get cancelled or postponed, so I'm keeping expectations low until after it happens. In the meanwhile, I'm working on a list of topics of interesting things I could say.

That's it for now. See you in June!

Ram.