Ethical behavior doesn’t take much philosophy. Be good to your family. Be good to your friends. Be good to your community. Be good to people who you for some ideological reason think that you should be good to or serve (here philosophy reintroduces itself). Be decent to people who are reasonable.
It’s much simpler, intellectually speaking, to figure out if you are on the right track by using interactions as cues. Following through might be more difficult.
That’s easy. More challenging is being good to everyone, always. People you don’t like. People who push your buttons or aggravate you. Your estranged family member. The foolish man. The obnoxious bore. The guy who has polar opposite political beliefs.
Try it for a day.
EAs have long since internalized this idea, and do actually look at the reality on the ground. For example, here's a citation-laden summary of the empirical evidence that those malaria bed nets are helping, both on net and on the margin:
https://www.givewell.org/international/technical/programs/in...
AFAIK - one reason for this shift is because A) people who (want to) work in ML/AI want jobs B) billionaires like to fund shit that strokes their ego (AI > mosquito nets) and C) people who work in the EA field need money too and turns out mosquitos nets aren't super sexy to fund. So, inherently, people in EA will stroke the ego of billionaires with rapture-like projects because they get funding from those projects and they'll probably say stuff that is a bit extreme (along with some psychos just being thrown into any crowd).
I know people who work in EA. It's complicated and difficult to balance. You have to realize - people (billionaires) contribute often just to stroke their own ego... not because they only want to help the world. At best - you can get them to do a split investment on their pet AI project and then a bit on real EA stuff. Otherwise, they'll just not contribute anything - and then you have the same capitalistic issue that you'll always have... too much wealth being concentrated into the hands of a few. The people I know in EA are often like, "Well if I take any money out of these guys - it's better than them having it. They have so much and no one person should have that much. It's inefficiently used by their hoarding."
I feel exactly the same way - but having been present through a lot of it, perhaps I can add some context.
Back in ~2009, Toby Ord was at 'The Oxford Uehiro Centre for Practical Ethics' and started 'Giving What We Can' which is exactly what you remember: Demonstrably cost-effective interventions like mosquito nets and deworming tablets. Measuring charity impact by quality-adjusted life years. Membership by pledging to donate 10% of your income to such causes. Approving of earning to give.
The thing is, GWWC was basically complete when it launched. Their recommended most efficient charities today are the same ones they launched with.
Oh, there were still some things to discuss - is it even more efficient to spend money on lobbying government to spend the foreign aid budget on those highly efficient causes? - but not enough to create a thriving discussion community or keep philosophy professors employed.
But GWWC wasn't the only thing happening at Oxford's philosophy department at the time. GWWC, The Oxford Uehiro Centre, the Centre for Effective Altruism and the Future of Humanity Institute were all in the same place at the same time - and in some senses EA is a broader version of GWWC, sharing its spreadsheet-and-QALYs-based approach. Toby Ord's focus moved from global poverty to existential risks, and GWWC and EA just sort of got rolled up together.
Thus creating this strange organisation - where one branch of the organisation would say the benefits of donating to medical research aren't quantifiable enough; while another branch works on things that are a hundred times more speculative than that.
And as the 'existential risk' branch has a lot more to argue about - they're a much more visible part of the community.
But one of the main focuses of EA is arguing about "exactly how good they are", and in particular the assumption that a cause whose general good was "fully decided years ago" shouldn't be seen as a good destination for philanthropic dollars without the organization focusing on providing metrics and RCTs to back up the idea that it's more beneficial than other definitely-has-some-positive-impact causes
Applying more scrutiny to aid organizations is fine and some of their scepticism of (e.g) microfinance is completely justified, but saying "but you can't prove that this money is spent as efficiently as they could have been" to sending stuff to poor people in South Asia or "you can't prove this will ever work" to funding African entrepreneurs isn't particularly compatible with recommending money is sent to more opaque organisations paying expensive developers to pursue passion projects on because it's theoretically possible the organization will have a non-zero and non-negative impact on a possible future harm which might theoretically be big.
Obviously not all EAs do this and many have very strong and consistent donation guidelines or are more interested in the maximise my personal potential to contribute side of EA than the demand more from aid agencies side, but it was the founder of GiveWell that recommended $30m of philanthropic cash was spent on buying a board seat at the non-profit subsidiary of extremly opaque and extremely well funded corporation OpenAI, which clashes quite hard with his earlier body of work. EAs who are interested in AI safety obviously aren't at all unusual in seeing moonshot research as a valid use of philanthropic dollars, but most people funding universities aren't simultaneously questioning the value of organizations for feeding people for not providing enough metrics to prove they're feeding the right people.
(It cuts both ways too: the arguments for investing in "AI safety" - diversification in the face of uncertainty - are arguments against the EA principle of maximising the efficiency of philanthropic dollar based on rigorous analysis of different organizations' return)
But I will say that this doesn't seem unusual to me; there seems to be a "missing middle" in many situations. For instance, in business, it's easy to get capital for things that are a sure thing - "for the past decade, every dollar we have put into this initiative has generated three dollars of revenue" will get funded every time - and also often pretty easy to get funding for high risk but high reward initiatives - "this is unlikely to work but if it does, every dollar invested will return one hundred dollars" - but it's pretty hard to get funding for medium risk / medium reward initiatives - "this will probably work, and if it does it will be a stable small business".
I think the same thing is going on here, and maybe it even makes sense.
I find longtermism to be dishonest because it has nothing to do with long term thinking. The long term nonsense is just there crank up the utility score.
When I checked back a few years later, the community seemed to be all about surviving the Rapture, I mean “Singularity”, summoning God, I mean the “Friendly AI”, and maximizing the happiness of the expected 10 to the trillion lives in Heaven, I mean the “Simulation”. I found all of that a lot less convincing.
There are several social reasons why this shift happened, but the root problem, to me, is utilitarianism taken to a dangerous extreme. Effective Altruism was arguably always rooted in some form of utilitarianism as a guiding principle. The problem is, the concept seems to have attracted some very logical, abstract thinkers who somehow didn’t realize when they started to lose contact with reality.
Note that all of this is just my opinion, but here’s my point. I find it always problematic to assign a numerical goodness value (“utilons”) to a human life, though I concede that it can work as a heuristic in small, non-extreme cases. What definitely doesn’t work is to take that number, multiply it by population sizes and claim to have solved ethics. No matter how much you wish it was, ethics is simply not a subbranch a mathematics. To put it very bluntly: That road has been traveled before, and it leads to genocide.
To be clear, this isn’t hyperbole. I won’t link any sources to limit exposure, but you can find articles by self-proclaimed altruists who argue that it’s better to lose a billion lives to climate change than to delay AI research by even one year. The suggested reason being that delaying the Ascent into Heaven — I mean the “Upload into the Simulation” — costs humanity a collective 10 to the gazillion utilons, compared to the mere billions of utilons potentially lost to the climate catastrophe. Just think about the implications.