I am an attorney in a public-sector position not associated with EA, although I cannot provide legal advice to anyone. My involvement with EA so far has been mostly limited so far to writing checks to GiveWell and other effective charities in the Global Health space, as well as some independent reading. I have occasionally read the forum and was looking for ideas for year-end giving when the whole FTX business exploded . . .
Jason
The FTX and Alameda estates have filed an adversary complaint against the FTX Foundation, SBF, Ross Rheingans-Yoo, Nick Beckstead, and some biosciences firms, available here. I should emphasize that anyone can sue over anything, and allege anything in a complaint (although I take complaints signed by Sullivan & Cromwell attorneys significantly more seriously than I take the median complaint). I would caution against drawing any adverse inferences from a defendant’s silence in response to the complaint.
The complaint concerns a $3.25MM “philantrophic gift” made to a biosciences firm (PLS), and almost $70MM in non-donation payments (investments, advance royalties, etc.) -- most of which were also to PLS. The only count against Beckstead relates to the donation. The non-donation payments were associated with Latona, which according to the complaint “purports to be a non-profit, limited liability company organized under the laws of the Bahamas[,] incorporated in May 2022 for the purported purpose of investing in life sciences companies [which] held itself out as being part of the FTX Foundation.”
The complaint does not allege that either Beckstead or Rheingans-Yoo knew of the fraud at the core of FTX and Alameda.
It does, however, allege (para. 46) that the “transfers were nominally made on behalf of the
FTX Foundation and Latona, but actually were made for the benefit of Bankman-Fried.” This is a weak spot in the complaint for me. There’s a quote from SBF about needing to do some biosecurity work for PR and political reasons (para. 47), but corporations do charitable stuff for PR and political reasons all the time. There are quotations about wanting to de-emphasize the profit motive / potential in public communications (para. 49-50), but if the profits flowed to a non-profit it’s unclear how that would personally enrich SBF.Quoting Beckstead, the complaint alleges that the investments were “ill-advised and not well-evaluated” (para. 51). It further alleges that there was no, or very little due dilligence (para. 51), such as a lack of any valuation analysis and in most cases lack of access to a data room. As a result, Latona “often paid far more than fair or reasonably equivalent value for the
investments” (para. 52). As for the $3.25MM gift, it was “was made in a similarly slapdash fashion” as the Foundation agreed to it without even knowing whether the recipient was a non-profit (para. 53). There are also allegations about how that gift came to be (para. 53-64).Paragraph 68 is troubling in terms of Latona’s lack of internal controls:
On May 23, 2022, Rheingans-Yoo sent a Slack message to the Head of
Operations for FTX Ventures, asking her to wire $50 million to PLS on behalf of Latona,
because “Latona doesn’t have a bank account yet, and we’d like to move these funds as soon as possible.” After the Head of Operations inquired why they were wiring $50 million when the SAFE agreement was only for $35 million, Rheingans-Yoo said there was a “separate purchase agreement [that] had a $15mln cash advance.” When the Head of Operations asked for the purchase agreement for $15 million, Rheingans-Yoo replied, “I have an email, no formal agreement,” but “if that’s not sufficient, we can send the 35 first and get the purchase more formally papered.
There apparently wasn’t an attempt to formally paper the $15MM until September, and that paper was woefully vague and inadequate if the complaint is credible (para. 70). Likewise, Rheingans-Yoo allegedly offered to send over $3MM to another firm, Lumen, “on a handshake basis” without any paperwork “while we hammer out the full details” (para. 75). However, the funding was not actually sent until an agreement was signed (para. 78-79). A third investment was made on Rheingans-Yoo’s recommendation despite Beckstead describing it as “unattractive” for various reasons (para. 82-84).
The “donate, then invest” approach seemed to also be in play with a fourth firm called Riboscience. Paragraph 93 doesn’t sound great: “On June 29, 2022, Glenn emailed Rheingans-Yoo that “sufficient time ha[d] passed since the (most generous) donation to the Glenn Labs, that we can now proceed with your desired investment in Riboscience.”
Counts One to Five are similar to what I would expect most clawback complaints would look like. Note that they do not allege any misconduct by the recipients as part of the cause of action, as no such misconduct is necessary for a fraudulent conveyance action. You can also see the bankruptcy power to reach beyond the initial transferee to subsequent transferees under 11 USC 550 at play here. Most of the prefactory material is doubtless there in an attempt to cut off a defense from the defendant biosciences firms that they gave something of reasonably equivalent value in exchange for the investments.
In Count Eleven, the complaint alleges that “Rheingans-Yoo knew that the transactions with the Lifesciences Defendants did not provide and had virtually no prospect of providing Alameda with reasonably equivalent value, and that Bankman-Fried personally benefited from the transactions. Rheingans-Yoo thus knowingly assisted in and/or failed to prevent Bankman-Fried’s breaches of fiduciary duty to Alameda.” (para. 169). This allegedly harmed Alameda to the tune of $68.3MM. Elsewhere, the complaint alleges that ” [u]pon information and belief, Bankman-Fried and Rheingans-Yoo intended to benefit personally from any profits generated by any of
these companies if they turned out to be successful and/or developed a successful product.” (para. 5). However, “upon information and belief” is lawyer-speak for “we’re speculating, or at least don’t have a clear factual basis for this allegation yet.”In Count Twelve, the complaint alleges that “Beckstead and Rheingans-Yoo knew that the transfer to PLS funded by FTX did not provide and had virtually no prospect of providing FTX with reasonably equivalent value, and that Bankman-Fried personally benefited from the transaction. Beckstead and Rheingans-Yoo thus aided and abetted Bankman-Fried’s breaches of fiduciary duty to FTX.” (para. 174). This allegedly harmed FTX to the tune of $3.25MM.
In the end, the complaint doesn’t exactly make me think highly of anyone involved with FTX or the FTX Foundation. However, from my non-specialist eyes, I’m not seeing a slam dunk case for critical assertions about Beckstead and Rheingans-Yoo’s knowledge in paras. 169 and 174.
Moving Toward More Concrete Proposals for Reform
The annual report suggests there are 45 to 65 statutory inquiries a year, link below (on mobile / lunch break, sorry!). So maybe a half to slightly less seem to end up as public reports.
I skimmed the oldest ten very quickly and it looks like four subjects were wound up / dissolved, and four more had trustee-related actions like appointment of new trustees by a Commission-appointed Interim Manager, disqualification from being a trustee, etc. One organization had some poor governance not rising to misconduct/misadministration (but some trustees resigned), one had Official Warnings issued to trustees, one got an action plan.
Pending more careful and complete review, most inquires that result in public reports do seem to find substantial mismanagement and result in significant regulatory action.
Strong upvote for both James and Joel for modeling a productive way to do this kind of post—show the organization a draft of the post first, and give them time to offer comments on the draft + prepare a comment for your post that can go up shortly after the post does.
- 26 Feb 2023 18:41 UTC; 18 points) 's comment on Why I don’t agree with HLI’s estimate of household spillovers from therapy by (
I’m struggling to see how releasing information already provided to the investigation would obstruct it. A self-initiated investigation is not a criminal, or even a civil, legal process—I am much less inclined to accept it as an adequate justification for a significant delay, especially where potentially implicated people have not been put on full leaves of absence.
- 25 Mar 2023 1:48 UTC; 42 points) 's comment on Holden Karnofsky’s recent comments on FTX by (
While Dawn claims it is “important” that Singer filed a demurrer rather than contesting factual allegations in court, no one should update on that legal strategy. Almost any rational litigant would have done the same thing given the procedural posture.
For background at the 10,000 foot level, at the very early stages of litigation, you can file a demurrer (“motion to dismiss for failure to state a claim” in federal court) claiming that even if everything in the complaint is true, it doesn’t give rise to liability. You generally cannot ask the court to dismiss the case at that point because the alleged facts aren’t true. The reason is that if the plaintiff’s legal theory is sound, she should ordinarily have an opportunity to develop the facts through discovery (document production, depositions, etc.) before the court addresses factual issues.
Discovery is time consuming and expensive, so if you have an argument that “even if everything you say is true, there’s no liability here” and an argument that “what you say isn’t true,” it is almost always better to present only the former argument at the demurrer stage. If you start disputing alleged facts, you’re implicitly telling the court that there needs to be discovery of the facts before the legal sufficiency of the complaint can be determined. Most people would rather not pay tens to hundreds of thousands of dollars, have to turn over documents, have to sit for a deposition, etc. Thus, they would file the demurrer even if they also think they could prove factual allegations in the complaint to be untrue.
At least MacAskill and Beckstead need to resign from the EVF board, take a leave of absence, publicly recuse from responding to the FTX situation, or submit a detailed explanation of the facts and circumstances that render none of these actions appropriate. They are just too intertwined in the events that happened to be able to manage the conflict of interest and risks to impartiality (or at least the appearance of the same).
That’s not me saying that I think they committed misconduct, but I think the circumstances would easily “cause a reasonable person with knowledge of the relevant facts to question [their] impartiality in the matter,” and that’s enough. Cf. 5 CFR § 2635.502(a) (Standards of Conduct for Employees of the Executive Branch [of the U.S. Government]). Although I’m not going to submit that EA officials should always follow government ethics rules, this is a really clear-cut case.
Hopefully they have recused and this just isn’t being stated due to PR concerns (because some people might misinterpret recusal as an admission of wrongdoing rather than as respect for a basic principle of good governance).
Obviously this is very breaking news, but depending on the ultimate facts, I would be nervous about the risk of a clawback action if I were an organization that had received funding from an FTX-aligned source in the past few years. It’s been a while since I took bankruptcy law, but the trustee can have pretty significant clawback powers when the debtor was actually insolvent at the time of transfer and the transfer was not for value. Of course, we do not know at this juncture whether the insolvency is of recent origin or existed for a while before this week.
I would also consider deferring any sizable donations to an organization I thought might be at risk for a crippling clawback, stick those monies in a DAF or similar entity for the time being, and ask the DAF to slowly regrant to the at-risk organization over time depending on the circumstances until it became clear there was no clawback risk. If a charitable organization is subject to a large clawback, it might be more efficient to move the charity’s operations to a new charity (paying FMV for any assets, of course). In that case, it would be better to have not given money to the exposed charity as that money would end up in the hands of FTX’s creditors. For instance, a number of charities had to pay clawbacks in the Madoff scandal despite not having committed any wrongdoing—despite the name, there does not have to be any evil intent to have been involved in a fraudulent conveyance.
None of this is intended to be in the least bit authoritative—it is merely a suggestion to stop and assess risk before taking certain significant actions in the short run.
- 12 Nov 2022 11:41 UTC; 18 points) 's comment on IMPCO, don’t injure yourself by returning FTXFF money for services you already provided by (
I don’t understand the importance of CEA saying anything to the community about this particular matter. We can all read Bostrom’s statement and draw our own conclusions; CEA has—to my knowledge—no special knowledge about or insight into this situation. The “PR speak” seems designed to ensure that each potentially quotable sentence includes a clear rejection of the racist language in question.
The government’s sentencing memorandum for SBF is here; it is seeking a sentence of 40-50 years.
As typical for DOJ in high-profile cases, it is well-written and well-done. I’m not just saying that because it makes many of the same points I identified in my earlier writeup of SBF’s memorandum. E.g., p. 8 (“doubling down” rather than walking away from the fraud); p. 43 (“paid in full” claim is highly misleading) [page cites to numbers at bottom of page, not to PDF page #].
EA-adjacent material: There’s a snarky reference to SBF’s charitable donations “(for which he still takes credit)” (p. 2) in the intro, and the expected hammering of SBF’s memo for taking credit for attempting to take credit for donations paid with customer money (p. 95). There’s a reference to SBF’s “idiosyncratic . . . beliefs around altruism, utilitarianism, and expected value” (pp. 88-89). This leads to the one surprise theme (for me): the need to incapacitate SBF from committing additional crimes (pp. 87, 90). Per the feds, “the defendant believed and appears still to believe that it is rational and necessary for him to take great risks including imposing those risks on others, if he determines that it will serve what he personally deems a worthy project or goal,” which contributes to his future dangerousness (p. 89).
For predictors: Looking at sentences where the loss was > $100MM and the method was Ponzi/misappropriation/embezzlement, there’s a 20-year, two 30-years, a bunch of 40-years, three 50-years, and three 100+-years (pp. 96-97).
Interesting item: The government has gotten about $3.45MM back from political orgs, and the estate has gotten back ~$280K (pp. 108-09). The proposed forfeiture order lists recipients, and seems to tell us which ones returned monies to the government (Proposed Forfeiture Order, pp. 24-43).
Life Pro Tip: If you are arrested by the feds, do not subsequently write things in Google Docs that you don’t want the feds to bring up at your sentencing. Jotting down the idea that “SBF died for our sins” as some sort of PR idea (p. 88; source here) is particularly ill-advised.
My Take: In Judge Kaplan’s shoes, I would probably sentence at the high end of the government’s proposed range. Where the actual loss will likely be several billion, and the loss would have been even greater under many circumstances, I don’t think a consequence of less than two decades’ actual time in prison would provide adequate general deterrence—even where the balance of other factors was significantly mitigating. That would imply a sentence of ~25 years after a prompt guilty plea. Backsolving, that gets us a sentence of ~35 years without credit for a guilty plea.
But the balance of other factors is aggravating, not mitigating. Stealing from lots of ordinary people is worse than stealing from sophisticated investors. Outright stealing by someone in a fiduciary role is worse than accounting fraud to manipulate stock prices. We also need to adjust upward for SBF’s post-arrest conduct, including trying to hide money from the bankruptcy process, multiple attempts at witness tampering, and perjury on the stand. Stacking those factors would probably take me over 50 years, but like the government I don’t think a likely-death-in-prison sentence is necessary here.
You touched on something here that I am coming to see as the key issue: whether there should be a justice system within the EA/Rationality community and whether Lightcone can self-appoint into the role of community police.
Pretty much every community has norms and means of enforcing those norms (“social control” to the sociologists). Those means may be more or less formal, but I don’t think communities are very viable without some means of norm enforcement. I think “justice system” implies something significantly different than what has happened here: e.g., the US justice system can throw me in a dungeon and take away all my money. To use a private example, if I were Catholic, the Catholic justice system could excommunicate me, defrock me as a priest, etc. A campus justice system can expel or fire me.
What happened here feels more like gossip on steroids. Lightcone said bad things about Nonlinear, which had the effect of decreasing community opinion of Nonlinear. That might in turn have concrete adverse effects on Nonlinear. But as far as I know: Lightcone did not, and could not, directly impose consequences on Nonlinear unmediated by the actions of the community.
Likewise, I don’t think “community police” is right, at least in the frame of modern Western societies. The police and prosecutors (collectively “police”) have the exclusive ability to charge people with crimes. They have the exclusive ability to use certain investigative tools, like search warrants. Although everyone has the ability to investigate possible wrongdoing in some capacity (e.g., journalists), the police are generally recognized as having a preeminent role. For example, if there is a police investigation into certain conduct, other processes like private organizations and civil investigations are usually expected to step aside to avoid disrupting the police investigation.
Rather, I think Lightcone’s role here is more akin to the prosecutorial role in classical Athens. Any citizen could bring a criminal prosecution. In fact, there were no public prosecutors. There was a preliminary stage, but then the matter was tried before a large jury of citizens (e.g., 500 for the most famous trial, that of Socrates). It’s not clear to me that Lightcone has claimed any role in norm enforcement that is superior to the role they would believe appropriate for any community member. True, it championed the cause of A and C, but that was a result of a voluntary decision among A, C, and Lightcone. Similar things happened in classical Athens, and Lightcone hasn’t appointed itself sole champion of people with grievances. What we are seeing on the Forum is vaguely like the deliberations of the citizen jury.
I’m not seeing a clear alternative to “private” prosecution of norm violations in such a decentralized community. Who is in a place to be the public prosecutor? One could argue for CHSP, but it is merely another private actor with no real democratic legitimacy and some serious conflicts of interests (e.g., due to being part of CEA, due to so much of CEA’s funding coming from Open Phil). I do not like the alternative of their being no norm enforcement except what is available through the legal system.
Although I can envision alternatives to adjudication by the whole community, I don’t think we can criticize the abstract idea of bringing disputes to the whole community for adjudication at this time. In such a decentralized structure, sanctions for norm violations are imposed by community actors in their individual capacities. In other words, everyone who hears of Lightcone’s charges has to decide for themselves whether they will change the way they interact with Nonlinear (and/or Lightcone) as a result. Likewise, there is no adjudicatory authority who would publicly warn everyone else of the risks of associating with an organization that has committed serious norm violations. Whatever its flaws, public adjudication seems to be the only real option at present for a certain class of matters.
Emerson talking about suing for libel—his right—was seen as defection from the norms which that Lightcone employee thinks should apply to the whole EA/rationality community. When did Emerson opt into following these norms, being judged by these norms? Did any of us?
Fair enough, but the flipside is that Lightcone didn’t opt in to following your proposed norm of not criticizing people for threatening to file defamation suits. Lightcone criticizing Emerson for his speech is Lightcone’s right. It is your right (and mine, and everyone else’s) to decide not to associate with Lightcone, Nonlinear, both, or neither based on your assessment of their various actions.
If you disagree with the priorities of major EA donors (notably, if you want CEA to be less GCR-focused than OP’s GCRCB team), you may wish to donate to CEA to preserve funder diversity.
In most scenarios, the benefits of a marginal increase in funder diversity seem rather modest in comparison to the opportunity costs. At present, OP would be ~ $22.6MM of a $28.2MM budget, with other established donors at ~ $2MM and a funding gap of ~ $3.6MM. In other words, depending on how much of the gap was filled, OP would be funding somewhere between 80 and 92% of the budget. Where the actual percentage is within that range has ~ zero impact on my assessment of COI risk, and not too much more impact on my assessment of its potential influence on CEA more generally. (Of course, some of the “other” donors are also GCR-focused, and so do not contribute to funder diversity from a cause-area perspective.)
Stated metaphorically, this feels like rearranging deck chairs on the Titanic. I think the effect of funder diversity may be very roughly exponential between 0% and ~50%,[1] but in the sense that the effect may be “almost imperceptable” at 8% and “very modest” at 20%.
Moreover, it’s not clear that if I hypothetically had $1MM to donate this year (I don’t!), that doing so while identifying strongly as a non-GCR donor would generate any appreciable shift in the focus of CEA’s work as a result of marginally increasing funder diversity. I’m sure hypothetical me could earmark those funds, but money is fungible. I’m not sure I am getting much “vote” for my $1MM.
On the other hand—given that non-GCR meta seems much more thinly funded than GCR and GCR-leaning meta -- $1MM could have a pretty significant effect if the donor figured out a way to counterfactually increase non-GCR meta spending by other organizations by most of $1MM. So the opportunity cost of improving CEA’s funder diversity to a donor interested in non-GCR meta seems fairly high. This may be even more so at the $100K point, which would still be a pretty significant donation to (e.g.) Probably Good.
- ^
50% is a stand-in for the point at which the organization could perceive itself as substantially continuing to exist without the predominant donor’s support. I’m hazy on what substantially means here; obviously, it would involve a lot of downsizing but would be significantly more than the minimum viable organization.
- 30 Nov 2023 17:41 UTC; 10 points) 's comment on CEA is fundraising, and funding constrained by (
- ^
This is highly subjective, but it seems to me the overall level of centralisation within EA is currently similar to fairly decentralised religious groups, and many social movements.
It can also be helpful to break down “centralisation” into sub-dimensions, such as:
[ . . . .]
On these dimensions, it seems to me that EA is currently fairly decentralised on group membership and information flow, very decentralised on ownership, and in the middle on decision-making power [5] and culture.I think there’s an important difference to be made between “level of centralization” in general and “level of power centralization.” When people are saying “EA is too centralized,” I think they are predominately referring to the latter concept.
Moreover, to the extent that the text above is breaking down centralization into sub-dimensions, and then impliedly taking something like the mean score of sub-domains to generate an overall centralization score, I don’t think that would be correct. Rather, I think the overall centralization measure is strongly influenced by the sub-dimension with the highest centralization score, especially where that dimension is decision-making / control of resources.
As an example, imagine analytic philosophy (AP) except that its meta organizations and individual practitioners (or at least university departments) were dependent on a funding ecosystem of short-term grants. Moreover, in this hypothetical, 70% of them are made by a single grantmaker and the overwhelming majority by a few dozen grantmakers who do not rubber-stamp grant renewals. Based on Will’s rating of EA, this would seem to move AP’s decision-making concentration to “in the middle.” If one determines centralization by some sort of averaging-like mechanism, this wouldn’t move AP’s overall centralization average that much. But I suspect that funding structure would have a huge practical impact on AP’s centralization (or that of any other academic field).
SBF’s sentencing memorandum is here.
On the first page of the intro, we get some quotes about SBF’s philanthropy. On the next, we are told he “lived a very modest life” and that any reports of extravagance are fabrications. [N.B.: All page citations are to the typed numbers at the bottom, not to the PDF page number.]
For the forecasters: based on Guidelines calculations in the pre-sentence report (PSR) by Probation, the Guidelines range is 110 years (would be life, but is capped at the statutory max). Probation recommended 100 years. PSRs are sealed, so we’ll never see the full rationale on that. The average fraud defendant with a maxed-out offense level, no criminal history, and no special cooperation credit receives a sentence of 283 months.
The first part of the memo is about the (now advisory) Sentencing Guidelines used in federal court. The major argument is that there should be no upward adjustment for loss because everyone is probably getting all their money back. Courts have traditionally looked at the greater of actual or “intended” loss, but the memo argues that isn’t correct after a recent Supreme Court decision.
As a factual matter, I’m skeptical that the actual loss is $0, especially where much of the improvement is due to increases in the crypto market that customers would have otherwise benefitted from directly. Plus everyone getting money back (including investors who were defrauded) is far from a certain outcome, the appellate courts have been deferential to best-guess loss calculations, and the final Guidelines range would not materially change if the loss amount were (say) $25MM. If I’m the district judge here, I’d probably include some specific statements and findings in my sentencing monologue in an attempt to insulate this issue from appeal. Such as: I’d impose the same sentence no matter what the Guidelines said, because $0 dramatically understates the offense severity and $10B overstates it.
There are a few places in which the argument ventures into tone-deaf waters. The argument that SBF wasn’t in a position of public or private trust (p. 25-26) seems awfully weak and ill-advised to my eyes. The discussion of possible equity-holder losses (pp. 20-21) also strikes me as dismissive at points. No, equity holders don’t get a money-back guarantee, but they are entitled to not be lied to when deciding where to invest.
The second half involves a discussion of the so-called 3553 factors that a court must consider in determining a sentence that is sufficient, but not greater than necessary. Pages 41-42 discuss Peter Singer and earning to give, while pages 46-50 continue on about SBF’s philanthropy (including a specific reference to GWWC and link to the pledger list on page 46).
Throughout the memo, the defense asserts that FTX was different from various other schemes that were fraudulent from day one (e.g., p. 56). My understanding is that the misuse of customer funds started pretty early in FTX’s history, so I don’t give this much weight. The memo asserts that SBF was less culpable than various comparators, ranging from Madoff himself (150 years) to Elizabeth Holmes (135 months) (pp. 73-80). The bottom-line request is for a sentence of 63-78 months, which is the Guidelines range if one accepts the loss amount as $0 (p. 89).
There are 29 letters in SBF’s support by family members, his psychiatrist, Ross Rheingans-Yoo, Kat Woods, and a bunch of names I don’t recognize.
[Caution: The remainder of this post contains more opinion-laden commentary than what has preceded it!]
I generally find the 3553 discussion unpersuasive. The section on “remorse” (pp. 55-56) rings hollow to me, although this is an unavoidable consequence of SBF’s trial litigation choices. There is “remorse” that people were injured and impliedly that SBF made various errors in judgment, but there isn’t any acknowledgment of wrongdoing. One sentence of note to this audience: “Sam is simply devastated that the advice, mentorship, and funding that he has given to the animal welfare, global poverty, and pandemic prevention movements does not begin to counteract the damage done to them by virtue of their association with him.” (p. 55).
I find the discussion of the FTX Foundation to be jarring, such as “Ultimately, the FTX Foundation donated roughly $150 million to charities working on issues such as pandemic prevention, animal welfare, and funding anti-malarial mosquito netting in Africa.” (p. 57). Attempting to take credit for sending some of the money you took from customers to charity takes a lot of nerve!
Although the memo asserts that SBF’s neurodiversity makes him “uniquely vulnerable” in prison (p. 58), the unfortunate truth is that many convicted criminals have characteristics that make successfully adapting to prison life more difficult than for the average person (e.g., severe mental illness, unusually low intelligence). So I’m not convinced by the memo that he would face an atypical burden that would warrant serious consideration in sentencing.
Although I certainly can’t fault counsel for pointing to SBF’s positive characteristics, I’m sure Judge Kaplan knows that many of his opportunities to legibly display these characteristics have been enabled by privilege that most people being sentenced in federal court do not have.
I’m also not generally convinced by the arguments about general deterrence. In abbreviated form, the argument is that running SBF through the ringer and exposing him to public disgrace is strong enough that a lower sentence (and the inevitable lifetime public stigma) suffices to deter other would-be fraudsters. See pp. 66-67. And there’s good evidence that severity of punishment is relatively less important in deterrence.
However, if a tough sentence is otherwise just, I don’t think we need a high probability of deterrent effect for an extremely serious offense for extended incarceration to be worth it. Crypto scams are common, and as a practical matter it is difficult to increase certainty and speed of punishment because so much of the problematic conduct happens outside the U.S. So severity is the lever the government has. Moreover, discounts for offenders who have a lot to lose (because they are wealthy already) and/or are seen as having more future productive value seem backward as far as moral desert.
Finally, I think there’s potentially value in severity deterrence of someone already committing fraud; if the punishment level is basically maxed out at the $500M customer money you’ve already put at risk, there is no reason (other than an increased risk of detection) not to put $5B at risk. As the saying goes, “might as well be hanged for a sheep as for a lamb” as the penalty for ovine theft was death either way.
Defense recommendations on sentencing are generally unrealistic in cases without specific types of plea deal. This one is no different. Also, the sentencing discussion will sound extremely harsh to at least non-US readers . . . but that’s the situation in the US and especially in the federal system.
I’d note that SBF’s post-arrest decisions will likely have triggered a substantial portion of his sentence. Much has been written about the US “trial penalty,” and it is often a problem. However, I don’t think a discount of ~25-33% for a prompt guilty plea, as implied by the Guidelines for most offenses (also by the guidelines used in England and Wales) is penologically unjustified or coercive. Instead of that, SBF’s sentence is likely to be higher because of multiple attempts at witness tampering and evasive, incredible testimony on the stand. So he could be looking at ~double the sentence he would be facing if he had pled guilty.
He likely could not have gotten the kind of credit for cooperation his co-conspirators received (a “5K.1”) because there was no other big fish to rat out. Providing 5K.1 cooperation often reduces sentences by a lot, in my opinion often too much. Given the 5K.1 cooperation and the lesser role, one must exercise caution in using co-conspirator sentences to estimate what SBF would have received if he had promptly accepted responsibility.
Finally, I’d view any sentence over ~60 years as de facto life and chosen more for symbolic purposes than to actually increase punishment. Currently, one can receive a ~15% discount for decent behavior in prison, and can potentially serve ~25-33% of the sentence in a halfway house or the like for participating in programs under the First Step Act. It’s hard to predict what the next few decades will bring as far as sentencing policy, but the recent trend has been toward expanding the possibilities for early release. So I’d estimate that SBF will actually serve ~75% of his sentence, and probably some portion of it outside of a prison.
Threats of lawsuits are fundamentally different to other threats. They are, as @Nathan Young put it, bets that the other party is so wrong you’re willing to expend both of your time and money to demonstrate it.
I don’t think that’s quite right. Threats of lawsuits are extremely cheap—it takes ten seconds max to type “I’ll sue you!” They are also commonly made without a reasonable basis in law or fact, and without any intent to actually follow through. They are often used to inappropriately silence speech and truthseeking.
Hiring a lawyer and filing a complaint gets closer to a bet, as you are putting some significant money and time on the line. However, people definitely file defamation suits as a form of PR management without any real intent to see them to trial, expecting that observers will see the filing as a marker of earnestness and/or a signal to withhold judgment until legal proceedings are complete. By the time the case is withdrawn or settled on confidential terms, public interest has moved on and nobody pays much attention. [1]
So evaluating whether it was an appropriate to threaten litigation requires an assessment of whether the threatening party would have a reasonable basis in fact and law, as well as a reasonable basis under the community’s norms,[2] for initiating and completing litigation. Bluffing to intimidate a would-be publisher, or filing litigation without good cause, should be strongly condemned. I also think that suing, or threatening to sue, individual employees in their personal capacities should generally be off the table where their conduct was in the scope of their employment.
- ^
For example, in recent defamation litigation involving cheating accusations made by the then-world champion, I think the plaintiff’s attorneys may have brought certain claims and made procedural missteps to draw a motion to dismiss that would stall the case out for months and allow the plaintiff to characterize the court’s adverse ruling as on procedural matters rather than the merits. After that, the case confidentially settled with no indication that money changed hands. A less interesting possibility is that plaintiff’s lawyers were just not very good.
- ^
In my tentative view, those norms should match the US constitutional norms where individuals of little prominence are concerned (mere negligence required), and where the speaker is a limited-purpose public figure (“actual malice” required). However, between those poles, I would often apply a sliding scale rather than a binary. For individuals/organizations of some prominence, where the subject matter is one of legitimate community concern, we probably want to err to some extent on not suppressing speech. So my standard there might be ~ “gross negligence OR refusal to retract when actually presented with pretty clear evidence of falsity.”
- ^
Also, if you had asked for growth predictions in September 2022 -- and told the forecasters that 2023 would be a year in which public awareness of / interest in the nearness and risks of powerful AI greatly increased—one would likely factor this information into the prediction in addition to historical trends.
By rough analogy, if the stock market is unexpectedly up 25 percent over last year, your sector is up 35 percent, but your company’s stock is flat . . . the market seems to be saying something negative about your company. The flat stock value may well mean that positive macroeconomic trends and sector trends have been cancelled out by negative outlook on your company.
Here’s my (working) model. I’m not taking a position on how to classify HLI’s past mistakes or whether applying the model to HLI is warranted, but I think it’s helpful to try to get what seems to be happening out in the open.
Caveat: Some of the paragraphs relie more heavily on my assumptions, extrapolations, suggestions about the “epistemic probation” concept rather than my read of the comments on this and other threads. And of course that concept should be seen mostly as a metaphor.
Some people think HLI made some mistakes that impact their assessment of HLI’s epistemic quality (e.g., some combination of not catching clear-cut model errors that were favorable to its recommended intervention, a series of modeling choices that while defensible were as a whole rather favorable to the same, some overconfident public statements).
Much of the concern here seems to be that HLI may be engaged in motivated reasoning (which could be 100% unconscious!) on the theory that its continued viability as an organization is dependent on producing some actionable results within the first few years of its existence.
These mistakes have updated the people’s assessment of HLI’s epistemic quality to change their view of HLI from “standard” to “on epistemic probation”—I made that term up, it is fleshed out below.
An organization on epistemic probation should expect greater scrutiny of its statements and analyses, and should not expect the same degree of grace / benefit of the doubt that organizations in standard status will get. These effects would seem to logically follow from the downgrade in priors about epistemic quality referenced in (1).
Whie on probation, an organization will be judged more strictly for mild-to-moderate epistemic faults. Here, that would include (e.g.) the statement James expressed concern about.
Practically, that means that the organization should err on the side of being conservative in its assertions, should devote extra resources toward red-teaming its reports, etc. While these steps may slow impact, they are necessary to demonstrate the organization’s good epistemics and to restore community confidence in its outputs.
An organization can exit epistemic probation by demonstrating that its current epistemics are solid over a sufficient period of time, and that it has controls in place to prevent a recurrence of whatever led to its placement on probation in the first place. In other words, subsequent actions need to justify a re-updating of priors to place the organization back into the “standard” zone of confidence in epistemic soundness. An apology will usually be necessary but not sufficient.
For HLI, the exit plan probably includes producing a new transparent, solid CEA of StrongMinds that stands up to external scrutiny. (Withdrawing that CEA might also work.)
It probably also includes a showing that sufficient internal or external controls are now in place to minimize the risk of recurrence. This could be a commitment to external peer review of the revised StrongMinds CEA as well as other new major recommendations and the reports on which they are based, a commitment to offer bounties for catching mistakes in major CEAs (with a third-party adjudicator), etc., etc.
Finally, the exit plan probably includes a period of consistently not making statements on the Forum, its website, and other arenas that seem to be a stretch based on the underlying evidence.
Of course, HLI’s funding position makes it more challenging for it to meet some of these steps to exit probation. Conditional on HLI having properly been placed on probation, I don’t know to what extent the existence of financial constraints should alter the quantum of evidence necessary to remove it from probation.
I think the concept of epistemic probation is probably useful. It is important to police this sort of thing. Epistemic probation gives the organization a chance to correct the perceived problem, and gives the community an action to take in response to problems it deems significant that isn’t excluding the organization from the community.
For better and for worse, each of us have to decide for ourselves whether an organization is on epistemic probation in our eyes. This poses a problem, because the organization may not realize a number of people have placed it on epistemic probation. So while I don’t like the tone or some of the contents of certain comments, I think it’s critical that the community provides feedback to organizations that puts them on notice of their probationary status in the eyes of many people. If many people silently place an organization on probation, and the organization fails probation (perhaps due to not knowing it was in hot water), then those people are going to treat the organization as excluded for its epistemic failures. That’s a bad outcome for all involved.[1]
One other point, which is also more challenging due to decentralization: The end goal of probation is restoration to good standing, and so it needs to be clear to the organization what it needs to do (and avoid doing) in order to exit probation. I tried to model this in points 6(a) to 6(c) above [conditioned on my assumptions about why people have HLI on probation], as well as in the example to my comment to Greg about whether HLI has been “maintain[ing]” its position after errors were pointed out. Of course, different people who have placed HLI on probation would have different opinions on what is necessary for HLI to exit that status.
- ^
Some people may have already decided to treat HLI as excluded, but my hunch is that these people are fairly small in number compared to the number who have HLI on probation.
A “normie” board would have at least put Will and Nick on a leave of absence. Putting people on a leave of absence when something happens that significantly affects the community’s trust and it is not possible to presently address that concern due to legal/other reasons is standard operating procedure for most larger organizations. It does not imply misconduct or maladministration by those who are placed on leave.
Or maybe not. Part of the problem is that the board was too small already, and I assume it was poorly positioned to sideline 2 of the 5 members while dealing with the reputational and other upheavels caused by FTX crisis, the (now-realized) possibility of Charity Commission involvement, and the management of massive clawback litigation.
Could you say a bit more about what you mean by “should not have maintained once they were made aware of them” in point 2? As you characterize below, this is an org “making a funding request in a financially precarious position,” and in that context I think it’s even more important than usual to be clear about HLI has “maintained” its “mistakes” “once they were made aware of them.” Furthermore, I think the claim that HLI has “maintained” is an important crux for your final point.
Example: I do not like that HLI’s main donor advice page lists the 77 WELLBY per $1,000 estimate with only a very brief and neutral statement that “Note: we plan to update our analysis of StrongMinds by the end of 2023.” There is a known substantial, near-typographical error underlying that analysis:
The first thing worth acknowledging is that he pointed out a mistake that substantially changes our results. [ . . . .] He pointed out that Kemp et al., (2009) finds a negative effect, while we recorded its effect as positive — meaning we coded the study as having the wrong sign.
[ . . . .]
This correction would reduce the spillover effect from 53% to 38% and reduce the cost-effectiveness comparison from 9.5 to 7.5x, a clear downwards correction.
While I’m sympathetic to HLI’s small size and desire to produce a more comprehensive updated analysis, I don’t think it’s appropriate to be quoting numbers from an unpatched version of the CEA over four months after the error was discovered. (I’d be somewhat more flexible if this were based on new information rather than HLI’s coding error, and/or if the difference didn’t flip the recommendation for a decent percentage of would-be donors: deprivationists who believe the neutral point is less than 1.56 or so).
First off, thank you to everyone who worked on this post. Although I don’t agree with everything in it, I really admire the passion and dedication that went into this work—and I regret that the authors feel the need to remain anonymous for fear of adverse consequences.
For background: I consider myself a moderate EA reformer—I actually have a draft post I’ve been working on that argues that the community should democratically hire people to write moderately concrete reform proposals. I don’t have a ton of the “Sam” characteristics, and the only thing of value I’ve accepted from EA is one free book (so I feel free to say whatever I think). I am not a longtermist and know very little about AI alignment (there, I’ve made sure I’d never get hired if I wanted to leave my non-EA law career?).
Even though I agree with some of the suggested reforms here, my main reaction to this post is to affirm that my views are toward incremental/moderate—and not more rapid/extensive—reform. I’m firmly in the Global Health camp myself, and that probably colors my reaction to a proposal that may have been designed more with longtermism in mind. There is too much in the post for anyone to fully react to without several days of thinking and writing time, so I’ll hit on a few high points.
1. I think it’s critical to look at EA activity in the broader context of other actors working in the same cause area.[1] I suggest that much of EA’s value in spaces where it is a low-percentage funder is in taking a distinctive approach that zeroes in on the blind spots of bigger fish. [2] In other words, an approach that maximizes diversity of perspective within EA may not maximize the effective diversity of perspective in the cause area as a whole. Also, the important question is not necessarily how good EA’s epistemic tools would work in a vacuum with no one else in a given space.
In spaces where EA is a niche player, I am concerned that movements in the direction of looking like other actors may well be counterproductive. In addition to GH&D, I believe that EA remains a relatively small player in animal advocacy and even in fields like pandemic prevention compared to the total amount of resources in those areas.
2. I feel that the proposal is holding EA up to a much higher standard at certain points than comparable movements. World Vision doesn’t (to my knowledge) host a message board where everyone goes to discuss various decisions its leadership has made. I doubt the Gates Foundation devotes a percentage of its spend on hiring opposition researchers. And most major charitable funders don’t crowdsource major funding decisions. Other than certain meta spending (which raises some conflict-of-interest flags for me), I don’t see anything that justifies making demands of EA unless one is simultaneously making demands of similar outfits.
Given that a large portion of the authors’ critique is about learning from others outside EA, I think that the lack of many of their proposed reforms in many similarly-sized, mature charitable movements is a significant data point. Although I believe in more process, consultation, and “bureaucracy” than I think the median EA does, I think there has to be a recognition that these things incur significant costs as well.
3. Portions of this reform package sound to my ears like the dismantling of EA and its replacement with a new movement, Democratic Altruism (“DA”). It seems unlikely that much of classic EA would be left after at least radical democratization—there are likely to be a flood of incoming people, many with prior commitments, attracted by the ability to vote on how to spend $500MM of Uncle (Open) Phil’s money every year. Whoever controls the voter registration function would ultimately control the money.
Now, I think DA is a very interesting idea, and if I had a magic wand to slice off a tiny slice of each Western charitable spend and route it to a DA movement, I think that would more likely than not be net positive. I’m just not clear on why EA should feel obliged to be the ashes from which DA arises—or why EA’s funders should feel obliged to fund DA while all the other big-money interests get to keep their handpicked people making their funding-allocation decisions.
4. As I noted in a comment elsewhere on this thread, I don’t think the community has much leverage over its funders. Unfortunately, it is much easier to come up with interesting ideas than people who want to and can fund them. Especially Grade-A funders—the proposal suggests a rejection, or at least minimization, of various classes of less-desirable donors.
As a recent post here reminds us, “[o]nly the young and the saints are uncompromised.” There’s rarely a realistic, easy way to get large sums of money for one’s cause without becoming compromised to some extent in the process. There’s the traditional way of cultivating an army of small/mid-size donors, but that takes many years, and you end up spending lots of resources and energy on care and feeding of the donors instead of on getting stuff done. I suspect most movements will spend a lot of time waiting to launch, and seeking funding, if they will only take Grade-A donor money. That’s a massive tradeoff—I really value my bednets! -- and it’s not one I am personally desirous of making.
One final, more broadly conciliatory point: EA can be too focused on what happens under the EA brand name and can seem relatively less interested in empowering people to do good effectively outside the brand. It doesn’t have a monopoly on either effectiveness or altruism, and I’ve questioned (without getting much in the way of upvotes) whether it makes sense to have a unified EA movement at this point.
I like the idea of providing different options for people where they can do good as effectively as possible in light of their unique skills, passions, and interests. For some people, that’s going to be classic GiveWell-style EA (my own likely best fit), for others it is going to be something like the current meta, for yet others it’s going to be something like what is in this proposal, and there are doubtless many other potential flavors I haven’t thought about. Some people in the community are happy with the status quo; some people are not. The ideal might be to have spaces where everyone would be locally happy and effective, rather than try to preserve or reform the entire ecosystem into something one personally likes (but isn’t conducive to others).
For example, in Global Health & Development, you have a number of NGOs [e.g., World Vision at $1.2B is several times EA’s entire spend on GH&D; see generally here for a list of big US charities] plus the truly big fish like the Gates Foundation [$6.7B, although not all GH&D] and various governments. So the vast majority of this money is being moved through democratic processes, Gates-type subject-matter experts, and traditional charities—not through EA.
It’s scandalous to me that some of the opportunities GiveWell has found were not quickly swallowed up by the big fish.