Research/Hybrid @ MIT FutureTech/Monash University/Ready Research
2584 karmaJoined Dec 2015Working (6-15 years)Sydney NSW, Australia



Visiting Scientist at MIT FutureTech helping with research, communication and operations. Doing some 'fractional movement building'. 

On leave from roles as i) a behaviour change researcher at BehaviourWorks Australia at Monash University and ii) EA course development at University of Queensland.

Founder and team member at Ready Research.

Former movement builder for i) UNSW, Sydney, Australia, ii) Sydney, Australia and iii) Ireland, EA groups.

Marketing Lead for the 2019 EAGx Australia conference.

Founder and former lead for the EA Behavioral Science Newsletter.

See my LinkedIn profile for more of my work.

Leave (anonymous) feedback here.


A proposed approach for AI safety movement building


Topic Contributions

Thank you for this, Ren! I really appreciate it.

One tip I would suggest is to consider trying to find template(s) to follow for any new research project. That is, try to find one or more papers which address the same topic and/or use the same method as your paper, before you start on the paper. The best case is if the templates are published in the journal/outlet you are targeting. 

When you have templates, use them to guide and simplify the production of your paper. For instance, this might be by closely following the method, structure, style of diagrams, sections or paragraphs etc in the template. Or by rewriting an existing but similar section (e.g., the abstract or conclusion) rather than starting one from scratch. 

Often you can basically reuse arguments and references. For instance, if you are trying to claim that Animal welfare is important in context x, and a published paper has a paragraph arguing for that, then you can often just rework that paragraph and cite the same references rather than spend time trying to find your own references and construct new arguments etc.

You also reduce a lot of risk by building on something that was accepted after a lot of revision and work, rather than trying to implement your naïve best attempt at what would be acceptable etc.

Is this true GWWC? I didn't realise that sacrificed income counted.

Thank you for sharing. This felt very personally relevant. I also haven't taken the pledge. I very nearly did in 2016. I go back on forth on whether and when I should take it. There are so many considerations at play. 

How much am I willing/able to sacrifice my wealth and financial security alongside sacrifices made for my direct work? Would/will I have as much direct impact without my savings/passive income potential? I'd be more risk-averse and less likely to address 'funding market failures'. Could I give more and give better later on that now? How much is motivated reasoning because I am very averse to spending on most things... etc

I also find it a little hard to talk about with people who have taken the pledge without feeling worried that I will be judged for being selfish or just feeling selfish and bad etc.

Related to that, I would really like someone to make a decision tree, guesstimate model, or similar for people to determine when and whether someone should take the pledge or donate money.

Related to this desire for when to give analysis, I would also like some way of upvoting potential posts or project (and even funding them), perhaps hosted on or linked to the EA forum. The situation feels like a potential coincidence of wants problem. I think I would pay maybe 100usd for a good post on this topic, if easy. I suspect that a lot of people would also be willing to do this (but not sure). I think someone else would take the opportunity to write the post/make the model if they knew there was demand, but the opportunity/demand is not easily advertised/aggregated etc.

The vast majority of academic contributions get read by approximately no one

Academia has benefits, but reaching a larger audience is not one of them, as far as I can tell (there are of course exceptions and some publications are much better suited for being published in a journal than a blogpost, but by and large academia does not have a good way of actually driving readership).

I agree that most academic research is a bad ROI but I find that a lot of this sort of 'nobody reads research' commentary is equating reads with citations which seems completely wrong. By that metric most forum posts would also not be read by anyone.

I do think it is somewhat fair to say that you will miss out on audiences if you only post on a forum because a forum post probably won't reach certain audiences.

Almost nobody searches academic databases, even academics and researchers mostly use normal Google search when researching things.

This has not been my experience as an academic. For several years my job was to trying to find and synthesise evidence for industry/government decision makers. Those reports involved searching academic databases and largely omitted unpublished literature and never included any forum posts, as far as I recall.  I think that this was also true of my colleagues.

Literature reviews are also published and read a lot and these usually do not include unpublished work.

I think we should maybe consider doing some research to test the reach and credibility of different communication modalities to have a better sense of when they are worthwhile. For instance, an experiment to test if people (ideally policymakers or similar demographics) differ greatly in their assessments of content in different formats (e.g., preprint v forum post). Or to see where key decision makers get their information from and what they trust.

Otherwise, we run the risk that a lot of very detailed and helpful content on places like LW and EAF is largely overlooked by people outside those communities, including many important decision makers. 

Thank you for this. I found it very helpful, for instance, because it gave me some insight into which audiences are currently perceived as being most valuable to engage by leaders in the AI safety and governance communities.

As I mentioned in my series of posts about AI safety movement building, I would like to see a larger and more detailed version of this survey.

Without going into too much detail, I basically want more uncertainty reducing and behavior prompting information. Information that I think will help to coordinate the broader AI safety community to do things that benefit themselves and the community. Obviously a larger sample would be much better.

For instance, I would like to understand which approaches to growing the community are perceived as particularly positive and particularly negative, and why. I want people with the potential to reach and engage potentially valuable audiences to better understand the good ways/programs etc to on board new people into the community. So we get more of what organizations and leaders think are good programs or good approaches and less of the bad.

I'd like to know what number of different roles organizations plan to hire. Like is it the case that these organizations collectively expect to hire 10 information security experts, or do they think it's important and want someone else to fund that work? Someone who works in information security might be very likely to attempt a career transition if they expect job opportunities but this doesn't quite demonstrate that. I would like it to be the case that someone considering the possibility of a stressful and risky career transition into AIS has the best possible information they can have about the probability that they will get a role and be useful in the AI safety community.

In my experience many people find the AI safety opportunity landscape is extremely complex and confusing and this probably filters out a significant portion of good candidates who don't have time to figure out and be secure in pursuing options that we probably want them to take. More work like this, if effectively disseminated, could make their decisions and actions easier and better.

Yeah, I think that might be one reason it isn't done. I personally think that it is probably somewhat important for the community to understand itself better (e.g., the relative progress and growth in different interests/programs/geographies). Especially for people in the community who are community builders, recruiters, founders, etc. I also recognise that it might not be seen as priority for various reasons or risky for other reasons and I haven't thought a lot about it. 

Regardless, if people who have data about the community that they don't want to share with the community, it might be good to ensure that the community knows that i) someone is collecting this data and ii) who they can ask for access to that data.  I think this would increase confidence in the quality of community governance and reduce thought/time wasted on debates or analyses (e.g., about the topic of this post). 

My uncertainties were mainly related to questions like how FTX had affected the trajectory of the community, size of pledge programs, and growth of AI relative to other areas of EA. But also around broader community understanding, like which programs are bigger, growing faster, better to recommend people to etc.

Thanks, I didn't know about the dashboard or had forgotten about it. Very helpful.

Short of doing something like this again, a simple annual post that reminds people this dashboard exists a and summarises what it has/shows, could get a lot of the value of the bigger analysis with a lot less effort. I imagine that a lot of people don't know about the dashboard and a lot of new people won't know next year.  

Thank you for this Angelina! It is extremely informative and has given me many useful updates about the size and trajectory of EA and its programs. It has also resolved some uncertainties which helps with my motivation. I expect that many readers will have a similar experience.

I would like to see more of this sort of monitoring in the future. Do you plan to do a similar analysis next year?

Thank you for all of this work. I really appreciate it. 

The process of writing this post has only strengthened my conviction about an issue I’ve previously noted: I believe the community should assign responsibility to, and funding for, one or more people or organizations to conduct and disseminate this sort of high-level analysis of community growth metrics. I honestly find it baffling that measuring the growth of EA and reporting findings back to the community isn’t someone’s explicit job.

I completely agree with this and have made several similar comments. 

Load more