I am looking for work.
Vasco Grilošø
Karma: 8,893
ļDatasets that change the odds you exist
Thanks! It looks like you are also pessimistic about donating to the organisations working on invertebrate welfare I recommended. I would be curious to know why.
Thanks, Joel! Makes sense. Sorry for not having read the 2nd bullet carefully.
Thanks, Joel. I had seen that. However, you used the formula ā= 6/ā55ā to calculate the adjustment for diminishing marginal returns, which is 11 %. The 10 % in the notes above seems to be referring to a different calculation from GWWC for the same adjustment (otherwise, I would have expected you to use the formula ā= 0.1ā³).
Thanks for the good points, Ryan.
I can see the annual probability of the absolute value of the welfare of Earth-originating beings dropping to 0 becoming increasingly low, and their population increasingly large. However, I do not think this means decreasing the nearterm risk of human extinction is more cost-effective than donating to GiveWellās top charities, or organisations working on invertebrate welfare.
Longtermists often estimate the expected value of the future from EV = p*V = āprobability of reaching existential safetyā*āexpected value of the future conditional on reaching existential safetyā. Yet, I am not aware of any modelling (as opposed to pure guesses) showing how decreases in the nearterm risk of human extinction translate into increases in p. I am not even aware of any detailed quantitative modelling estimating changes in the risk of human extinction.
I think decreasing the probability of worlds where humans go extinct soon will barely change p, instead just making a little more likely nearby worlds where humans go extinct slighly later. The easier way to decrease the risk of human extinction in 2025 is postponing it until to 2026, not until humans and their descendents have colonised the accessible universe.
It could also be that increasing V by 1 % via donating to GiveWellās top charities, or organisations working on invertebrate welfare is cheaper than decreasing p by 1 %, and it would be equally valuable. I estimate the absolute value of the welfare of marine arthropods is 99.99996 % (= 1 ā 1/ā(2.50*10^6)) of that of humans and marine arthropods. So I feel like more research on improving the welfare of wild arthropods may well be a cost-effective way of increasing V.
Yes, I think a similar objection applies. However, I would still expect more people to move to neighbourhoods with a lower mean income if people cared a lot about their income relative to their neighbours. I believe peopleās behaviour is better explained by people caring much more about their income than their income relative to their neighbours.
Thanks for sharing, Toby! I very much like this sort of transparency.
This might be wrongāthe content doesnāt seem to have led to more subscribers.
I am not sure what you mean here. Only subscribers can see the newsletterās content, so how can content lead to more subscribers? Are you saying that content has not prevented āthe monthly loss of around 200-300 subscribers per newsletterā?
Thanks, Toby!
Thatās interesting! I could definitely give that a go. I think that might be a bit off though because the more relevant thing is how much money I have access to in our budget/ā could access by making the right case to OP (and what it would be spent on otherwise). Iāve also never really understood how individual willingness to pay is a good signal when people have wildly different resources available to them. Could be a useful benchmark.
You can ask the subscribers which fraction of their net annual income they would be willing to spend to get an annual subscription of the EA newsletter, multiply the mean fraction across the surveyed subscribers by the number of subscribers to get the total value of the newsletter in terms of doublings of net income, and divide this by 2 doublings of net income per DALY averted to get the impact of the newsletter in terms of averted DALYs. Then you could divide this impact by the cost to get the cost-effectiveness in DALY averted per $, and compare with the cost-effectiveness of GiveWellās top charities of around 0.01 DALY/ā$.
Another point on thatāI donāt trust myself when it comes to valuing things monetarily. Iād never pay for social media or a newsletter, because it just isnāt something Iām used to (Iām used to them being free), but if I actually calculate the time Iām spending on social media/ā newsletters I count as a valuable use of time, and use something like this to calculate the value of my time, I get surprisingly high figure.
I wonder whether people are sometimes thinking about the total instead of marginal value. I would pay for additional/āmarginal air because I can get air for free, but I still get lots of value from air, and would pay a lot to keep what I have relative to having nothing. Similarly, you may not mind losing access to a single newsletter due to the other abundant online content you can continue to check for free. The marginal value is what matters from the perspective of assessing counterfactual impact. The product between oneās net hourly rate, and the time checking a newsletter would only be a good way of estimating its additional impact if the alternative to checking the newsletter had no value, which is not the realistic counterfactual.
This is a good questionāI should prioritise comparing these. I do expect that some of those programs (and possibly the EA opportunities board) are cheaper per click, and have a higher intent audience. The EA Newsletter hasnāt been optimised for job clicks (and probably shouldnāt be). Iāve been a bit bottlenecked by what I can most easily measure.
I also expect job boards to have more clicks per $.
Thanks for sharing, Toby!
While retaining our impact-weighted-clicks/ā$[1] score of $6/āclick.
How does your 0.167 impact-weighted-click/ā$ (= 1ā6) compare with the clicks per $ regarding 80 kās, Animal Advocacy Careerās, and Probably Goodās job boards? Since you are assuming 1 click on a job ad means 1 impact-weighted-click (footnote 1), and the above is your main metric to track cost-effectiveness, how would you update on the value of the EA newsletter if other impact-focussed job boards had way more clicks per $?
Specifically, Iām working this week on figuring out a rational Willingness To Pay per Subscriber metric (so that I can assess the price of marketing campaigns), so any suggestions there would be much appreciated.
I would ask current subscribers how much they would be willing to pay in order to maintain their subscription.
Whatās the average click rate for emails from non-profits?
NeonOne claims it is 3.29%.
Mailchimp claims it is 3.27%
I trust click rates more, I think they are much easier to measure.
As such I can more confidently say that we are doing great!
Nitpick. I think it is more accurate to say the click rate has been quite good in April, but similar to that of nonprofits from October to March 2025.
Apologies but I donāt know the two-vertices word which I should be using here [instead of triangulate].
Interpolate?
Thanks for sharing, Joe. I think the absolute value of the welfare of AI systems will remain much smaller than that of wild animals over at least the next few decades. According to Ege Erdil, āin 2024 we spent ~ $100B on NVIDIA GPUs, and by 2024 Q4 the total computing power of all NVIDIA datacenter chips worldwide was only around 4e21 FLOP/āsā, equal to that of 40 M human brains (= 4*10^(27 ā 15)). So, even if the absolute value of the welfare per FLOP of AI systems was equal to that of humans, their computing power would have to become 200 (= 8*10^9/ā(40*10^6)) times as large as that of NVIDIA datacenter chips for the absolute value of the welfare of AI systems to match that of humans. From Table S1 of Bar-on et al. (2018), there are 10^20 marine arthropods, 12.5 billion (= 10^20/ā(8*10^9)) times as many as humans. If the absolute value of the welfare per second of marine arthropods is 2*10^-4 that of humans, 10 % of Rethink Prioritiesā median welfare range of silkworms, the absolute value of their welfare would be 2.50 M (= 12.5*10^9*2*10^-4) times that of humans. So I estimate the computing power of AI systems would have to become 500 M (= 200*2.50*10^6) times as large as that of NVIDIA datacenter chips for the absolute value of their welfare to match that of marine arthropods assuming the absolute value of the welfare per FLOP of AI systems was equal to that of humans. 500 M is a huge factor considering the spending on NVIDIA GPUs in 2024 was around 0.1 % (= 100*10^9/ā(100*10^12)) of the gross world product in 2024.
Hi Ramiro,
I do not think Bryan would agree with that. Studying in the United States (US) means a higher chance of being employed relative to studying in a random country, but the US has higher salaries, which plays against leaving.
Thanks for sharing, Ramiro!
Tbh, I barely care about what economists in Kochās pockets claim (something Caplan apparently doesnāt consider worth denying)
I overwhelmingly prefer to look into the arguments, and put very little weight into the funding source.
Thanks for sharing.
The update averages out to ~$10Bn/āyear for The Gates Foundation (~$6Bn/āyear increase), compared to (example) ~Ā£9.2Bn expected spending for the UK Government in 2027.
10 billion $ is 4.95 % (= 10ā202) of the global official development assistance (ODA) in 2021 of 202 billion $.
Thanks, Alex! Strongly upvoted. It would be great if you could also share a cumulative version of that chart, showing the status of the commitments achieved in all the years before and during the commitment year, by commitment year. This would be helpful to understand the progression of the number of commitments in each reporting status as a fraction of the total number of commitments. I think both this graph and the one above would be good additions to your cage-free fulfillment report. Thanks for all your work!
I do not have any particular benchmarks in mind, but I think the ones Mechanize is aiming to develop will capture economic value more closely. So you may want to look at the existing benchmarks which you think will more closely resemble those. I would also ask Mechanize about it, @Thomas Kwa.
Your top comment no longer includes the graph.
Thanks for sharing, Matrice! Very relevant, although I do not think anything there undermines the points made by Anson. Toby concludes by saying āAnd of course it is also important to know how much any of this generalises to other suites of tasksā. I expect the half-life to be shorter for broader tasks which track economic value more closely.
Thanks for sharing, Thomas! I expect benchmarks whose scores are closer to being proportional to economic output improve slower.
Thanks for the suggestion, Ryan! 50 k$ is a large fraction of my savings, so it would not work for me. Moreover, I agree software engineering will be largely automated until 2031. I am just very sceptical that this will lead to a growth explosion. I guess my probability of it being automated more than e.g. 99 % is much lower than yours, but this may be hard to operationalise. The bet does not have to resolve in time for human intervention for it to be financially positive for you if I transfer money to you now? I wonder whether you expect the annual unemployment rate, globally or in the United States (US), to be much higher in 2027. In the US, it was 4 % in 2024, and I would say the probability of it being higher than 8 % in 2027 is lower than 1ā3.
Stocks grew 5 % per year from 1900 to 2022, and I think they will grow faster over the next 5 years, maybe around 10 % per year. So I would want to use an interest rate higher than yours of 4.5 %.