8648 karmaJoined May 2016


Topic Contributions

This could depend on your population ethics and indirect considerations. I'll assume some kind of expectational utilitarianism.

The strongest case for extinction and existential risk reduction is on a (relatively) symmetric total view. On such a view, it all seems dominated by far future moral patients, especially artificial minds, in expectation. Farmed animal welfare might tell us something about whether artificial minds are likely to have net positive or net negative aggregate welfare, and moral weights for animals can inform moral weights for different artificial minds and especially those with limited agency. But it's relatively weak evidence. If you expect future welfare to be positive, then extinction risk reduction looks good and (far) better in expectation even with very low probabilities of making a difference, but could be Pascalian, especially for an individual (https://globalprioritiesinstitute.org/christian-tarsney-the-epistemic-challenge-to-longtermism/). The Pascalian concerns could also apply to other population ethics.

If you have narrow person-affecting views, then cost-effective farmed animal interventions don’t generally help animals alive now, so won't do much good. If death is also bad on such views, extinction risk reduction would be better, but not necessarily better than GiveWell recommendations. If death isn't bad, then you'd pick work to improve human welfare, which could include saving the lives of children for the benefit of the parents and other family, not the children saved.

If you have asymmetric or wide person-affecting views, then animal welfare could look better than extinction risk reduction depending on human vs nonhuman moral weights and expected current lives saved by x-risk reduction, but worse than far future quality improvements or s-risk reduction (e.g. https://onlinelibrary.wiley.com/doi/full/10.1111/phpr.12927, but maybe animal welfare work counts for those, too, and either may be Pascalian). Still, on some asymmetric or wide views, extinction risk reduction could look better than animal welfare, in case good lives offset the bad ones (https://onlinelibrary.wiley.com/doi/full/10.1111/phpr.12927). Also, maybe extinction risk reduction could look better for indirect reasons, e.g. replacing alien descendants with our happier ones, or because the work also improves the quality of the far future conditional on not going extinct.

EDIT: Or, if the people alive today aren't killed (whether through a catastrophic event or anything else, like malaria), there's a chance they'll live very very long lives through technological advancement, and so saving them could at least beat the near-term effects of animal welfare if dying earlier is worse on a given person-affecting view.

That being said, all the above variants of expectational utilitarianism are irrational, because unbounded utility functions are irrational (e.g. can be money pumped, https://onlinelibrary.wiley.com/doi/abs/10.1111/phpr.12704), so the standard x-risk argument seems based on jointly irrational premises. And x-risk reduction might not follow from stochastic dominance or expected utility maximization on all bounded increasing utility functions of total welfare (https://globalprioritiesinstitute.org/christian-tarsney-the-epistemic-challenge-to-longtermism/ and https://arxiv.org/abs/1807.10895; the argument for riskier bets here also depends on wide background value uncertainty, which would be lower with lower moral weights for nonhuman animals; stochastic dominance is equivalent to higher expected utility on all bounded increasing utility functions consistent with the (pre)order in deterministic cases).

EDIT: Looks like Ariel beat me to this point by a few minutes.

(Not speaking on behalf of RP. I don't work there now.)

FWIW, corporate chicken welfare campaigns have looked better than GiveWell recommendations on their direct welfare impacts if you weigh chicken welfare per year up to ~5,000x less than human welfare per year. Quoting Fischer, Shriver and myself, 2022 citing others:

Open Philanthropy once estimated that, “if you value chicken life-years equally to human life-years… [then] corporate campaigns do about 10,000x as much good per dollar as top [global health] charities.” Two more recent estimates—which we haven’t investigated and aren’t necessarily endorsing—agree that corporate campaigns are much better. If we assign equal weights to human and chicken welfare in the model that Grilo, 2022 uses, corporate campaigns are roughly 5,000x better than the best global health charities. If we do the same thing in the model that Clare and Goth, 2020 employ, corporate campaigns are 30,000 to 45,000x better.[8]

About Open Phil's own estimate in that 2016 piece, Holden wrote in a footnote:

Bayesian adjustments should attenuate this difference to some degree, though it’s unclear how much, if you believe – as I do – that both estimates are fairly informed and reasonable though far from precise or reliable. I will put this consideration aside here.

My understanding is that they've continued to be very cost-effective since then. See this comment by Saulius, who estimated their impact, and this section of Grilo, 2022.

Duffy, 2023 for RP also recently found a handful of US ballot initiatives from 2008-2018 for farm animal welfare to be similarly cost-effective, making, in my view, relatively conservative assumptions.

See the meat-eater problem tag and the posts tagged with it. That being said, wild animal effects can complicate things.

Globally, there are around 20 billion farmed chickens alive at any moment, mostly factory farmed, so about 3 per human alive, higher in high-income countries and lower in low-income countries. There are also probably over 100 billion fish being farmed at any moment, so over 12 per human alive. See Šimčikas, 2020 for estimates.

If it's just a cold, or you're testing negative for COVID but still have mild symptoms, I think it should be okay to attend wearing a mask indoors and distanced outside, and eating outside or alone. I did this once for an EAG under the advice of the team with only and multiple negative tests and mild symptoms. I also checked with each of my 1-on-1s if they were still okay meeting and how, and (maybe excessively, and probably not what the team expected) skipped almost all group events and talks I had originally planned to attend. Part of the reason I skipped group events and talks was because I wouldn't be able to check with everyone if they were comfortable with me attending.

That being said, I felt pretty self-conscious attending, which was unpleasant, but I also had good 1-on-1s, as well as good interactions outside of the formal events.

As a one-liner, would it be accurate to say that the (vast) majority of individual animals (directly) used/exploited by humans per year probably become shrimp paste?

I'm guessing we aren't "using" nematodes, worms or arthropods not included above in larger numbers, at least not directly. Obviously they're affected by and involved in agriculture, and may be affected by fishing.

And I guess it's also useful to distinguish time spent exploited/used, because an individual wild-caught animal won't be under direct human control nearly as long as the average farmed animal. Farmed insects would plausibly already dominate if you weigh by time under direct human control while alive, and insect farming is growing quickly.

On the last question at the end, I don’t think the fact that LLMs take their own outputs as inputs means we would have to interpret the text space as the global workspace itself or everything in the global workspace in order to think of LLMs as having feedback/recurrence.

It could be more like talking (or thinking out loud), which also allows the LLM to guide its own attention. Each output text token would (also) be like an attention command for the next run of the network. The global workspace can still be internal, and the LLM's outputs are just sensed by the LLM, just like hearing your own voice when you speak.

Using outputs this way also doesn't necessarily mean the LLM is outputting everything it's thinking (if it thinks at all), just like we don't. It could have different internal thoughts, reflecting the high-dimensional stuff going on before just before output (or even earlier).

Its text outputs could even be coded commands to help it generate hidden internal thoughts over time, summarising what things it's done internally or what to do next without saying so to us. I don't think LLMs are doing this now, but maybe a superintelligent LLM could.

All this being said, the contents of an LLM's global workspace could change pretty dramatically from one pass over the text to the next pass of the same text+an extra token. Maybe unusually abruptly compared to animals.

Thanks Nick!

I'd guess it's actually often worth it to do both in the same places and actually doubly protect many people, despite the trickiness of accounting for both. You can use or check with more conservative assumptions to avoid overestimating impact. Otherwise you might miss out on a lot of impact by avoiding overlap.

Are these submissions available online to read?

FWIW, I'm not sure if I'd say "Unfamiliar" was the "top" reason for not trying the Impossible option in Malan's PhD thesis, because "Other" had a lot of specific answers indicating respondents preferred real meat (Appendix. XI.) and people might also use "Unwilling to spend swipe" to mean that without filling it in explicitly. Both options were popular. (Also, the answers sum past 100% because people could select multiple.)

I also wouldn't necessarily take the views of taste from people who hadn't even bothered to try it as good evidence against taste parity. Only 5% "Tried it elsewhere, didn't like" according to Table 23, although that's compatible with the rest trying it elsewhere and liking it, just less than real meat.

Of those that did try it, 90% agreed/somewhat agreed that it was delicious, and 85% agreed/somewhat agreed that it was a satisfying alternative to meat, according to Table 20. Maybe "somewhat agree" isn't a strong enough endorsement, though. And, of course, one-time consumers, who made up about 29% of interviewed ever-consumers of the Impossible option, tended to agree less with both statements. Furthermore, we can see that East Asians were more likely to try it than other race/ethnicities (Table 18), so we can possibly use race/ethnicity as a variable for openness or food neophobia, but among respondents who did try it, breaking down by race/ethnicity, East Asians found it the least satisfying as an alternative to meat, and were kind of middling on how delicious they thought it was (Table 26).

I think it’s possible it did reach taste parity for some people, but not others. (Tasting as good, not necessarily tasting the same, although people differ in how well they can distinguish tastes.) Taste is subjective!

Load more