All of Robi Rahman's Comments + Replies

(Even) More Early-Career EAs Should Try AI Safety Technical Research

There are tons of people vaguely considering working on alignment, and not a lot of people actually working on alignment.

1levin2d
Yes! This is basically the whole post condensed into one sentence
The Role of Individual Consumption Decisions in Animal Welfare and Climate are Analogous

Sorry, yes, didn't mean to imply Charles He was only talking about catering. I was just using that as an example of EAs following vegan diets in a way that costs more money, as opposed to costlessly. This post by Jeff Kaufman is relevant, https://www.jefftk.com/p/two-kinds-of-vegan :

"Go vegan!", you hear, "it's cheaper, more environmentally sustainable, and just as healthy and delicious!" The problem is, these aren't all true at the same time.

The Role of Individual Consumption Decisions in Animal Welfare and Climate are Analogous

The specific points being made in those quotations aren't mutually exclusive. Onni Aarne is saying you can make very inexpensive adjustments to your diet that greatly reduce animal suffering, and Charles He is saying that EA events spend extra money on catering to satisfy the constraint of making it vegan. I think both claims are correct.

4Stefan_Schubert25d
I didn't interpret Charles He as talking about EA events spending extra money on catering, but about individuals adopting vegan diets.
The Role of Individual Consumption Decisions in Animal Welfare and Climate are Analogous

The two are disanalogous from an offsetting perspective: Eating (factory farmed) animal products relatively directly results in an increase in animal suffering, and there is nothing that you can do to "undo" that suffering, even if you can "offset" it by donating to animal advocacy orgs. By contrast, if you cause some emissions and then pay for that amount of CO2 to be directly captured from the atmosphere, you've not harmed a single being.

This is only reasonable if you believe that causing x units of suffering and then preventing x units of suffering i... (read more)

4Onni Aarne25d
Yes, that's the narrowly utilitarian perspective (on the current margin). My point was that if you mix in even a little bit of common sense moral reasoning and/or moral uncertainty, causing x harm and preventing x harm is obviously more wrong than staying uninvolved. (To make this very obvious, imagine if someone beat their spouse but then donated to an anti-domestic abuse charity to offset this.) I guess I should have made it clearer that I wasn't objecting to the utilitarian logic of it. But even from a purely utilitarian perspective, this matters because it can make a real difference to the optics of the behavior.
The Role of Individual Consumption Decisions in Animal Welfare and Climate are Analogous

[Objection] from Robi Rahman: “A person choosing to eat 1kg less chicken results in 0.6 kg less expected chicken produced in the long run, which averts 20 days of chicken suffering. A comparable sacrifice would be to turn off your air conditioning for 3 days, which in expectation reduces future global warming by 10^(-14) °C and reduces suffering by zero.”

Without quibbling with the precise numbers, I think this is fundamentally a point about the importance of the two cause areas.

Actually, what I meant was fundamentally not a point about the importan... (read more)

The Role of Individual Consumption Decisions in Animal Welfare and Climate are Analogous

However, whether this means EAs are not allocating their resources appropriately requires consideration of the marginal cost-effectiveness of: [1, 2, 3, 4, 5]

Actually, it doesn't require knowing all of those! If you find that among two of those options, one is more cost-effective than the other, but resources are going to the less effective one, you already know the overall allocation is suboptimal (even though the optimal allocation is probably some entirely different option).

That's a main point of this essay, which I think is underappreciated througho... (read more)

1Matt_Sharp25d
Based solely on Gabriel's essay, how do we know this? There are some thoughtful qualitative suggestions why this may be the case, but I would find it more convincing if there were quantitative estimates which backed up these suggestions.
The Role of Individual Consumption Decisions in Animal Welfare and Climate are Analogous

I question your economic analysis of meat-eating:

your reduction in demand for meat makes them cheaper for others, which will lead some to increase in their consumption...

Following this logic the amount of consumption of any given non-essential good would never change.

You've misunderstood the line you quoted. It's only saying that other people's meat consumption will increase by some fraction of the amount you've reduced your consumption, not that people will increase their consumption by however much you reduce yours.

9Gabriel Weil25d
Yep, I'm just saying that the equilibrium change in the quantity consumed will be less than the individual's foregone consumption, not that it will be zero. How much less depends on the elasticities of supply and demand.
Who wants to be hired? (May-September 2022)

Intermediary German, Basic Spanish & Russian

Intermediate, not intermediary.

8jknowak1mo
Thanks! Shouldn't write that in a hurry - that's why I didn't add Proficient English 😅
Who wants to be hired? (May-September 2022)

Location: Boston, Massachusetts

Remote: yes

Willing to relocate: yes

Skills: computer programming, databases, machine learning, statistics

Resume: https://www.linkedin.com/in/robirahman/

Email: robirahman94@gmail.com

Introducing the ML Safety Scholars Program

Someone referred me to apply to be a TA for this program. How would you like such people to contact you - should I email you, or is there another form for that?

2ThomasWoodside1mo
Not clear right now whether we will need more TAs, but if we do, we'll make a post soon with an application. I'll reply to this if/when that happens. Thanks for your interest!
Norms and features for the Forum

Ah, you're right, I misinterpreted it since the epistemic status suggestion said time per post and that one didn't.

Norms and features for the Forum

Speculative: NLP for claim detection: the site asks you for your probabilities about the main claims. Time cost: 30 mins.

You think it'd take only 30 minutes to implement a feature that detects claims in forum posts? I'm not a web developer but that strikes me as wildly optimistic.

6Charles He2mo
I think all time costs stated are time costs to author of the post. From a product and ML implementation perspective and for the NLP component of the problem, I think in this case, it might be easy to build an 80% good solution. It’s less that the system will find and understand all arguments but more like the author might be asked questions and it’s relatively easy to see if the answers cover the same space as the post content. My guess is that manipulation won’t make sense or even enter into peoples minds (with other design choices not related to the NLP) so a useful system that say gives guide rails is much easier to implement.
Why Helping the Flynn Campaign is especially useful right now

If you think pandemic response is the key issue, Dr. Harder is a highly experienced doctor who used to run the Oregon Medical Board. Medical and policy experience: maybe you still think your guy will be better, but by how much?

The FDA has hundreds of highly -experienced doctors and still had such a disastrous response to the pandemic they probably caused millions of extra deaths. They completely blocked challenge trials and delayed vaccine deployment by six months. What matters is not whether the people in government are doctors, it's the policies on ho... (read more)

Thanks. I agree - you can debate who would be most effective on pandemic prevention! But it is debatable and I’d love for everyone here to factor that into their back of envelope effectiveness calculations.

But I also want to convince you all that your focus is way too narrow. This is not an election for pandemic czar, it’s an open seat several decades in the making and the representation for >650k Oregonians. So it rankles to see the race turned into an experiment to see if huge amounts of money can buy it for somebody who seems disinterested in most issues facing the district.

Hmm I fear there might be a cultural clash here. Many people on this forum believe that pandemic response (and especially prevention) was a massive and avoidable bipartisan failure on the part of the US, and a massive failure internationally on behalf of our institutions, experts, and governments overall  (see here for an anonymous take). Many people on the forum don't believe in the "overwhelming and avoidable failure" narrative, but at least they're sufficiently familiar with this story that this is a common starting point of debates around here.

I t... (read more)

6Linch2mo
(Note that this has to include non-US mortality as well in the metric).
Tentative Reasons You Might Be Underrating Having Kids

Mormonism is an obvious example of a religion that people join because Mormons have well-functioning families? I'm skeptical that's a main reason for the growth of Mormonism compared to their high birthrate or their amount of missionary effort.

We've found three stellar people to incubate. 🥳 More details to be announced soon.

What are the strongest arguments against working on existential risk? (EA Librarian)

Even a comparatively low pure discount rate of 1% implies most future value is concentrated in the next hundred years

This is not correct! Suppose the human population grows at a constant rate for 1000 years. If you discount the moral worth of future people by 1% per year, but the growth rate is anything above 1%, most of the value of humanity is concentrated in the last hundred years, not the first hundred years.

There's this very surprising, maybe counterintuitive moral implication of cosmopolitanism where if you think future people have moral value and yo... (read more)

6Pablo4mo
Hi Robi, The answer was assuming a constant population rather than a growing population, although (confusingly) that assumption was not made explicit. However, I hadn't appreciated the points you make in the second paragraph. That's very interesting.
Some thoughts on vegetarianism and veganism

I'm strongly in favor of 'welfaretarianism'! It's been my diet* for a few years now and I'm really glad you invented a name for it. I've been telling people for ages that I agree you shouldn't eat animals that suffer while farmed if it causes more of them to exist, but people don't really internalize the logical conclusion of this, that it's good to eat animals if it causes happy animals to exist (assuming you don't subscribe to negative utilitarianism or the person-affecting view) or existing animals to become happier. Hypothetically, if it were more prof... (read more)

EA Fundraising Through Advantage Sports Betting: A Guide ($500/Hour in Select States)

I'm guessing they went to Colorado because they were on the west coast and it was the closest state with legal sports betting.

2Sam Anschell5mo
Yep that's right. NJ and CO have nearly the same amount of bonus money on offer, but they both have much more than the next best state. If you already live in one of the twelve states listed in the post, I'd just go for it where you live. If you're traveling specifically to churn through new user bonuses like I did, NJ or CO are where it's at.
Dismantling Hedonism-inspired Moral Realism

Thanks for writing this sequence. I have gotten some weird looks in my discussion group last year for not being a moral realist - wish I'd had this link handy back then!

Free money from New York gambling websites

EV is the same, you're just reducing volatility (risk is maybe a better word?) by guaranteeing the outcome either way.

Oh, I see. Yes, this is what I thought - the EV doesn't change but you can reduce your exposure to particular outcomes.

EA Fundraising Through Advantage Sports Betting: A Guide ($500/Hour in Select States)

I'm glad you didn't see my post until now! I am a bit lazy and your writeup is a little more detailed than mine :) plus, I didn't emphasize enough that this is available to people outside of New York, although I think the recent NY legalization is why it is so especially lucrative right now.

Free money from New York gambling websites

To my knowledge, you are generally right. The situation is exceptional this month because the market was just legalized and these sites are trying to grab market share while it's brand new.

Free money from New York gambling websites

Yeah, here are some examples. I was in NYC from 5pm Saturday to 7pm Sunday.

I signed up for the BetRivers match promo. They asked me to verify my identity by uploading a photocopy of my driver's license (I used my passport, which was accepted). Upon depositing $250, they gave me $250 in matching funds. The soonest upcoming sports game was the Kansas City Chiefs vs the Buffalo Bills, which was at -110 moneyline odds (meaning the Chiefs were 55% favorites according to the sportsbook). FiveThirtyEight had them as 65% favored, so I bet on the Chiefs. They won, ... (read more)

2shrek5mo
Thank you, those examples make it much clearer.
Free money from New York gambling websites

It is not negative EV. You are either misunderstanding the offer or someone's example.

1shrek5mo
Perhaps I'm misunderstanding your example above or @martin_glusker's comment about the free bets. As I understand Martin's comment, you don't get your free bet stake back, so if you bet on something extremely likely you'll just get the small winnings but not the notional value of your wager. For instance, if you bet on something at -1000 with $1000 of free bet, you're likely to win, but you'd only get $100 of winnings, not $1000 of stake plus $100 of winnings. I can see how bets could be constructed that would recover more value using martin's link, but as shown there they mostly top out around 80% of the value. Thus in your example above in this case you'd end up with around $800 of cash (by careful construction of bets; much less if you just bet one something likely with low payout) in the much more likely losing case. I can see how the offer could be used in a way that is positive EV (e.g. bet opposite sides with another free bet offer so you win on one side and get the free bet credit on the losing side), but I think the example you posed is not positive.
Free money from New York gambling websites

I used the maximum EV you can get without risking any of your own money. If you want to have less exposure to outcomes, you can choose lower-EV bets with higher likelihood of payouts, but if you're doing this for charity, it doesn't really make sense to do that.

For example, suppose you have a $1000 free bet. It doesn't pay out the principal if you win, just returns payouts for whatever you bet on. You can bet on an outcome that is 91% likely, in which case you have a 91% chance of winning $100 and a 9% chance of winning nothing, for an EV of $91. Or you co... (read more)

Free money from New York gambling websites

Why is it optimal to size the hedge bet such that you get the same payout for either outcome? Why does that have greater EV than if the bets are skewed in either direction?

3martin_glusker5mo
EV is the same, you're just reducing volatility (risk is maybe a better word?) by guaranteeing the outcome either way. A downside is that the hedge does increase the necessary bankroll. That said EV does vary with how long the odds are.
Free money from New York gambling websites

Yes, this was something I was thinking earlier. I was in an ideal position to take advantage of the offers without getting hooked because: I hate sports, have good background knowledge of probability, don't like gambling, and don't live in NY and thus couldn't make any more bets even if I wanted to. If I'm the one taking these offers rather than a NY resident who might end up with a gambling problem, that's a very good social outcome in addition to the donation directed to charity.

Free money from New York gambling websites

Physical presence is enough. They have geolocation applets on their site and prohibit you from placing bets unless you are in an eligible location. I live in Massachusetts but went to NYC for about a day. I used my passport for identification because I didn't bring my driving license.

Free money from New York gambling websites

Correct. They might ban you and confiscate your money if you use a VPN to obfuscate your location.

Prediction Bank: A way around current prediction market regulations?

This pays far too little to the winners to make it worthwhile to have any money in this. It wouldn't have much more liquidity than a moneyless prediction book.

1Gaurav Sett5mo
That entirely depends on the return on the bank’s investment, right? I have no idea what that could be in practice. If it were similar to the stock market, say 8% annually, then I think that would be very attractive to forecasters. Being right on a poll where 50% of respondents were also right would be like doubling what you expect to earn from stocks. But obviously that’s risky investing, so probably not feasible. Or if you were already a big bank, you could afford such risk, and make it worthwhile for winners. Was that what you were thinking?
Free money from New York gambling websites

Oh, good catch, I will edit that. What is the EV from those offers, then? It seems to still be nearly +$1000 and nearly risk-free with the following approach: bet $1000 on something very unlikely (EV of the payout will be $1000, but as something like a 1% chance of $100000). If you lose, bet the $1000 of site credit on something extremely likely, so you end up with $1000 cash. Then withdraw that.

For BetMGM, repeat the same approach but do 5x very safe bets of $200 with the site credits.

That still works for $1000 expected profit, right?

6martin_glusker5mo
One thing to note is that some books have different definitions of "free bets" or "site credit." If you place a $100 free bet on a +100 market on FanDuel it pays out $200 (your $100 stake, and your $100 winnings). But on BetMGM a $100 free bet on a +100 market pays out $100, because you don't get your free bet stake back. This means that on sites like BetMGM or DraftKings (and I believe some others), "free bets" or "site credit" are not as good as cash. They're value in cash typically 70-80% of the face value of the credit because you can convert them into that amount of cash. See good markets for free-bet conversion here [https://oddsjam.com/low-hold].
4KaseyShibayama5mo
I agree with that analysis (and someone risk-neutral should bet the 2nd game on whatever game has the lowest vig [https://www.bookmakers.bet/1794/the-vig/]). Worth considering taxes, though.
Free money from New York gambling websites

I'd guess a couple more weeks. The Caesars one was just reduced from $3000 to $1500. They were most generous right at the beginning of legal online betting but are decreasing the incentives as everyone who will end up signing up has done so.

Free money from New York gambling websites

Yes, that's correct. Optimal usage of that offer is to deposit $1,500, use the free bet on an unlikely event that resolves very soon, and then withdraw your deposit quickly, plus winnings if any.

Free money from New York gambling websites

Good point, thanks! I added a note in the FAQ.

Reflections on EA Global London

I disagree with your point about participants not being cautious enough about covid. Last I heard (someone correct me if this was later updated), four attendees tested positive during or after the conference, out of about 900 participants. That is an impressively low rate, and indicates that the safety measures worked well! I want to commend the organizers for doing a great job addressing covid issues: they had lots of rapid tests available for us, gave lots of advice about travel safety, and didn't do anything excessive or unwarranted by the risk level, like cancelling the conference or social-distancing the discussions.

1PabloAMC7mo
Notice that I did not criticized the organisers, I think encouraging but not imposing masks was a good enough approach. What worried me was the possible consequences of getting covid abroad. On the other hand, 4 people was around what I expected based on the 14 days incidence. My intuition is that people behaved that way because they say people in London acting that way. And if not wearing masks at the event was the right behavior, then does that mean that we should not wear them anymore anywhere? It depends on your preferences of course, but under uncertainty I prefer to stick to masks, especially being so low cost.
The problem with person-affecting views

I thought the person-affecting view only applies to acts, not states of the world. I don't hold the PAV but my impression was that someone who does would agree that creating an additional happy person in World A is morally neutral, but wouldn't necessarily say that Worlds B and C aren't better than World A.

How many EA 2021 $s would you trade off against a 0.01% chance of existential catastrophe?

The steady-state population assumption is my biggest objection here. Everything you've written is correct yet I think that one premise is so unrealistic as to render this somewhat unhelpful as a model. (And as always, NPV of the eternal future varies a crazy amount even within a small range of reasonable discount rates, as your numbers show.)

2NunoSempere7mo
For what it's worth, I don't disagree with you, though I do think that the steady state is a lower bound of value, not an upper bound.
A Red-Team Against the Impact of Small Donations

If the financial capital is $46B and the population is 10k, the average person's career capital is worth about ~$5M of direct impact (as opposed to the money they'll donate)? I have a wide confidence interval but that seems reasonable. I'm curious to see how many people currently going into EA jobs will still be working on them 30 years later.

A Red-Team Against the Impact of Small Donations

Sorry, I didn't mean to imply that biorisk does or doesn't have "fast timelines" in the same sense as some AI forecasts. I was responding to the point about "if [EA organization] is a good use of funds, why doesn't OpenPhil fund it?" being answered with the proposition that OpenPhil is not funding much stuff in the present (disbursing 1% of their assets per year, a really small rate even if you are highly patient) because they think they will find better things to fund in the future. That seems like a wrong explanation.

A Red-Team Against the Impact of Small Donations

> At face value, [an EA organization] seems great. But at the meta-level, I still have to ask, if [organization] is a good use of funds, why doesn't OpenPhil just fund it?

Open Phil doesn’t fund it because they think they can find opportunities that are 10-100x more cost-effective in the coming years.

This is highly implausible. First of all, if it's true, it implies that instead of funding things, they should just do fundraising and sit around on their piles of cash until they can discover these opportunities.

But it also implies they have (in my opinion,... (read more)

1Charles He7mo
This has good content but I am genuinely confused (partly because this article's subject is complex and this is after several successive replies). Your point about timelines seems limited to AI risk. I don't see the connection to the point about CEPI. Maybe biorisk has similar "fast timelines" as AI risk—is this what your meaning? I hesitate to assume this is your meaning, so I write this comment instead. I really just want to understand this thread better.

I think the party line is that the well-vetted (and good) places in AI Safety aren't funding-constrained, and the non-well-vetted places in AI Safety might do more harm than good, so we're waiting for places to build enough capacity to absorb more funding.

Under that worldview, I feel much more bullish about funding constraints for longtermist work outside of AI Safety, as well as more meta work that can feed into AI Safety later.

Within AI Safety, if we want to give lots of money quickly, I'd think about:

  • funding individuals who seem promising and are somewh
... (read more)
Open Thread: Spring 2022

I noticed something at EAG London which I want to promote to someone's conscious attention. Almost no one at the conference was overweight, even though the attendees were mostly from countries with  overweight and obesity rates ranging from 50-80% and 20-40% respectively. I estimate that I interacted with 100 people, of whom 2 were overweight. Here are some possible explanations; if the last one is true, it is potentially very concerning:

1. effective altruism is most common among young people, who have lower rates of obesity than the general populatio... (read more)

7Will Bradshaw8mo
The natural first step here is to check whether EA has lower rates of overweight/obesity than the demographics from which it primarily recruits. I can't speak much to the US, but in the European countries I've lived in overweight/obesity varies massively with socioeconomic status. My classmates at university were also mostly thin, as were all the scientists I've worked with (in several groups in several countries) over the years. And it's my reasonably strong impression that many other groups of highly-educated professionals have much lower rates of obesity than the population average. In general, I've tended to be the most overweight person in most of my social and work circles – and I'd describe my fat level over the past 10 years as, at worst, a little chubby. If it is the case that EA is representative of its source demographics on this dimension, that implies that it doesn't make all that much sense to focus on getting more overweight/obese people into the movement. Obviously, as with other demographic issues, we should be very concerned if we find evidence of the movement being actively unwelcoming to these people – but their rarity per se is not strong evidence of this. (EDIT: See also Khorton's comment for similar points.)

I think there are extensions of (1) and (3) that could also be true, like "people at EA Global were particularly likely to be college-educated" and "people who successfully applied to EA Global are particularly willing to sacrifice today in order to improve the future"

EDIT: and just generally wealth leads to increased fitness I think - obesity is correlated with poverty and food insecurity in Western countries

5Linch8mo
FWIW I see a much higher percentage of overweight EAs in the Bay Area.
5Larks8mo
I'm skeptical of the comparability of your 2/100 and 50-80% numbers; being overweight as judged by BMI is consistent with looking pretty normal, especially if you have muscle. I would guess that more people would have technically counted as overweight than you'd expect using the typical informal meaning of the word. It could also be that obese people are less likely to want to do conference socializing, and hence EAG is not representative of the movement.
Managing COVID restrictions for EA Global travel: My plans + request for other examples

Additional suggestion: don't just have a photo of your vaccine card on your phone; physically bring it or scan and print a copy.

Managing COVID restrictions for EA Global travel: My plans + request for other examples

Thanks for the writeup! I'm following this process but going to the UK a few days earlier, so I'll try this out and provide results before you leave.

I ordered a 2-day covid test and received a booking reference number. My flight arrives in London on Friday, so tomorrow morning I will fill out the passenger locator form.

Edit, 2021-10-20: Submitted all my info to the UK gov website and got a passenger locator form. I'll update tomorrow when boarding the plane.

2021-10-21: will be departing from the US for the UK on Thursday evening.

2021-10-22: will be arriving in London on Friday morning.

6Robi Rahman8mo
Additional suggestion: don't just have a photo of your vaccine card on your phone; physically bring it or scan and print a copy.