Kestrel🔸
Bio
I work as a researcher in statistical anomaly detection in live data streams. I work at Lancaster University and my research is funded by the Detection of Anomalous Structure in Streaming Settings group, which is funded by a combination of industrial funding and the Engineering and Physical Sciences Research Council (ultimately the UK Government).
There's a very critical research problem that's surprisingly open - if you are monitoring a noisy system for a change of state, how do you ensure that you find any change as soon as possible, while keeping your monitoring costs as low as possible?
By "low", I really do mean low - I am interested in methods that take far less power than (for example) modern AI tools. If the computational cost of monitoring is high, the monitoring just won't get done, and then something will go wrong and cause a lot of problems before we realise and try to fix things.
This has applications in a lot of areas and is valued by a lot of people. I work with a large number of industrial, scientific and government partners.
Improving the underlying mathematical tooling behind figuring out when complex systems start to show problems reduces existential risk. If for some reason we all die, it'll be because something somewhere started going very wrong and we didn't do anything about it in time. If my research has anything to say about it, "the monitoring system cost us too much power so we turned it off" won't be on the list of reasons why that happened.
I also donate to effective global health and development interventions and support growth of the effective giving movement. I believe that a better world is eminently possible, free from things like lead pollution and neglected tropical diseases, and that everyone should be doing at least something to try to genuinely build a better world.
Posts 17
Comments185
I think my point is that there's *already* a lot of bad advice out there. The difference between n and n+1 pieces of advice that's bad for you being present around you shouldn't be the thing that everything hinges on. If it is, something's seriously off with your ability to filter and prioritise.
Take the High-Impact Professionals Playbook, for example (one might call this an "official EA advice source for exceptionally talented people"). It's got sections on
- Career
- Donations
- Workplace initiatives
- Trusteeship
- Mentoring and Advisory
- Network and Influence
Probably, any given person will fit in on one or maybe two of these, and so all the other bits are "bad advice". However EA assumes you're competent at basic seeking methodology because how else would you have got interested in EA - or alternatively, tries to teach you basic seeking methodology as a core part of you becoming an EA - so by the time you've ended up buying the book, you probably should be capable of dipping in and out of the relevant bits of a book.
If someone is exceptionally talented in capacity, but their seeking methodology is so poor they'd be swayed by whatever advice happens to be immediately around them at any given time, EA probably *doesn't* want them doing a bunch of the "highest-impact" stuff (particularly in AI) because they've got a very substantial possibility of making things worse.
The obvious first thing to ask is have they considered donating to the operations costs of an effective giving fundraiser team such as One for the World ( https://1fortheworld.org/ ) as a means of raising more money for the GiveWell Top Charities than they themselves would be able to donate?
Getting people to cross-cause prioritise is difficult. There's much that can be obtained by considered within-cause prioritisation.
I disagree voted, but I am not disagreeing about this happening - I am disagreeing about it being a good idea.
People who are exceptionally talented should be exceptionally capable of ignoring all the other bad advice they get, and the presence of more advice that is bad for them shouldn't hurt them. Also they should be capable of understanding EA as a project about "the most good you can do" and be on board with solid, actionable advice for the average person being easily accessible around the place.
One of EA's biggest failings is its community structures actively causing harm to people interested in EA, and more grounded advice would go a long way towards addressing that.
Solidly great practical questions here, I think I'll incorporate some of them into things I run. In particular I agree with you that EtG is extremely underprioritised. It is both the genuinely most impactful life path for most people, and also comes with a huge degree of personal flexibility and benefits.
I don't think it's in any way shameful to spend some of your time doing things that aren't the "best" use of your time. In fact, I do a lot of this. I have a life I want to lead, and I won't apologise for doing things that make me happy.
But in any case, community discussion of where best to give is usually positive from a util standpoint. It tends to teach you and other people around you structured and evaluated ways to prioritise your giving better, that are grounded in practical experience. And assuming you will be giving a whole bunch of money over the rest of your life, learning how to do it better is a super good use of time. There's a reason Giving What We Can used to run "where to allocate £1000" sessions back when it did more community outreach.
Yes, learning how to recognize (and fund) good stewardship is a hard skill.
But I feel that (some) genuinely committed, longer-term EAs are exactly the kind of people who may actually be able to sit down with each other to do that.
I am speaking with experience of being a Quaker (the other seeking-focused moral ambition cult I hang out in), and the answer the Quakers have is to fund administrators, not pastors. That is, the people who take salaries for community building essentially take direction from the volunteer community organisers and get all the niggly administrative bits sorted out to ensure that the "job" of a voluntary community organiser remains fun and meaningful rather than overly stressful. It also means it costs a lot less - you don't need that many administrators.
I could see this model working fairly well for a longer-term EA group. And it's basically how the EA Forum works: paid administrators keep the platform going and enforce basic discipline standards, "volunteer" EAs post whatever they're working on. And one could argue most of EAG works like this too: CEA hosts the space and food and Swapcard and enforces basic conduct standards for 1:1s, and leaves the EAs to all get on with EA-ing.
It shouldn't be such a hard sell. Why, when well stewarded meta-EA is such a ridiculously high-impact giving opportunity, and EA is full of givers searching for impact, do so few of us give to EA infrastructure?
Ultimately I think this is going to have to come from individual pledgers / EtGers who take control of their local groups via donations and reshape them into something that works for people in EA for the long haul.
Ending global extreme poverty via evidence-backed targeted cash transfers would be roughly $300 billion a year https://cega.berkeley.edu/end-of-poverty/
I don't think that's "every conceivable project that will make the world better", that's basically via GiveDirectly which a lot of EAs contribute to and is consistently estimated to be within the EA GHD funding bar sphere and does joint projects with GiveWell, part of "the main portfolio" if you will. Not to even start on the funding gaps in other bits of GHD, in animal welfare, in governance improvement, etc.
I understand that $50 billion sounds like a lot, and I'm really happy EA will be allocating more money effectively, but relative to both the scale of critical global need and the scope of possible global funding (shown via defense budgets) it's still tiny. We are a fraction of a shard of the size we "need to be".
For those reading this post: PauseAI UK is currently fundraising for operational expansion. If you want to donate to PauseAI UK, you can see their donor prospectus and make donations at https://pauseai.uk/donate