An Open Letter to Vitalik Buterin
I hope this letter finds you in good health. Unfortunately, the topic of this letter is grim. Your $4.4million dollar donation to the Machine Intelligence Research Institute is being misused, and you are now among a class of persons being defrauded and mislead by the Machine Intelligence Research Institute and the Center for Applied Rationality.
The Machine Intelligence Research Institute (MIRI) in Berkeley and its sister organization, the Center for Applied Rationality (CFAR), claim to be organizations dedicated to AGI safety and the art of human rationality. However, these organizations are not what they make themselves out to be, and MIRI is in fact defrauding its donors through misleading promises and an ongoing cover-up of statutory rape, blackmail, and fraud.
MIRI was founded as the Singularity Institute in 2000 by Eliezer Yudkowsky, a prominent thought leader in the Less Wrong “rationalist” community. Less Wrong is a community blog (also founded by Yudkowsky,) that claims to be dedicated to the art of human rationality. Yudkowsky has written over 300 blog posts, several books, and a popular fanfiction that gained him the status of a minor celebrity, especially among fellow “rationalists.” The “rationalist” community in the Bay Area, which is fairly tight-knit, features group housing, where members live together and sometimes work together.
Yudkowsky’s ideas had members of the Less Wrong community convinced that Yudkowsky would bring about singularity and save the world. Yudkowsky has positioned himself as one of a small class of people focusing on global catastrophic risks, while teaching his followers that the ethical thing to do is to either work directly on AGI, or get a high-paying job and donate to MIRI. Many of the people in the Less Wrong community took Yudkowsky and MIRI very seriously. Until recently, most of MIRI’s donations came from within that community, with some members donating tens of thousands of dollars to the cause. The original team at MIRI was composed partially of top contributors to the community’s blogs, like Nate Soares, Luke Meuhlhauser, and Anna Salamon.
While these bloggers weren’t as popular and admired as Yudkowsky, they were widely trusted by the community. And at events like the Workshop on AI Safety Strategy (WAISS) run by CFAR, people would suggest ideas like taking out life insurance that would pay out to MIRI, and then committing suicide to further the cause.
Working at MIRI conferred a kind of special respect as someone who Yudkowsky and other prominent members thought were worthy to save the world. To young teenagers enamored with the Less Wrong community and its stated ideals, it’s easy to see how kids could’ve been coerced into having sex with adults working at MIRI.
Testimony from community members alleges that several underage young teenagers were having sex with the 30- and 40- year old researchers working at MIRI, especially between 2007 and 2014. The list of people accused includes (but is not limited to) the director and senior research fellow, and a former executive director who is now working at the Open Philanthropy Project, a major donor to MIRI.
Perhaps relatedly, this was going on around the time MIRI was accepting significant financial support to the tune of $50,000 from Jeffrey Epstein, even after his conviction for prostituting children. Yes, that Epstein.
MIRI’s entanglement in statutory rape was one of the worst kept secret in the Bay. Many rationalists are keen to say that the age of consent is an arbitrary number, and favor treating young people as fully-grown adults. So, it wasn’t much of a secret that older individuals in their 30s and 40s were having sex with teenagers. And perhaps that’s why when nearly a hundred people in the Bay Area Rationalist Community Safety Discussion group (BARCSD) saw insiders discussing whether the perpetrators had successfully evaded the statute of limitations for statutory rape in reference to MIRI, the members of BARCSD collectively shrugged. Additionally, one of the former underage teenagers was in the group, and admitted to having sex with much older adults, though this failed to interest the BARCSD.
Through their actions, the leaders at MIRI revealed that their first priority was to protect the 35- and 40-year olds who were having sex with multiple underage teenagers ranging in ages from 14 to 17. Leaders acted in the defense of abusers at the expense of their organization’s mission. And when the community found out, the people involved admitted to less obscene crimes (“only one 17 year old, and it was consensual”) to cover for worse behavior. Aside from criminality, there’s a very large age and experience gap between the young teenagers and the staff at MIRI. And as has been discussed, these weren’t ordinary 35-year-olds, but highly respected members of the community who many thought were literally going to save the world.
Unfortunately, that’s not where this story ends.
Cover-up and Donor Fraud
In 2014, Louie Helm, the former deputy director of MIRI and a professional poker player, successfully extorted MIRI using this knowledge of statutory rape. Helm had allegedly left MIRI for greener pastures, but his plans hadn’t worked out and he wanted MIRI to pay him for some transitional work he did for them, or give him his job back. When MIRI refused, Helm published a repulsive website about MIRI, and accused Yudkowsky, Meulhauser, and Marcello Herreshoff of statutory rape, in addition to mocking people for their supposed STIs. He then escalated by reporting people at MIRI to the government for immigration fraud, tax fraud, and even filed false police reports. He threatened to circulate the website widely throughout the community, and to otherwise blacken MIRI’s name. During this period, Helm reportedly exhausted all his blackmail material. So it’s interesting that MIRI chose to use a legal settlement as cover to pay Helm for his silence.
MIRI could have easily refused, and sued him for his illegal and frankly obscene behavior. The website, miricult, mixed petty personal health issues with serious accusations in a way that made the serious accusations look less believable. However, were the serious accusations of statutory rape false, then it would’ve been straightforward to simply publicize Helm’s behavior. Outrage would’ve surely followed, as the rationalist community has been smeared before, and typically defends its heroes, like the community did when Cade Metz doxxed Scott Alexander. But in the world where the accusations of statutory rape are true and just shy of common knowledge, MIRI needed to do everything it could to keep it quiet. And that’s exactly what they tried to do.
However, MIRI’s stance on extortion and blackmail, critical to their published research, argues that one should never give in to blackmail or extortion. Giving in only encourages more blackmailers, and gives measure to the strategy of extortion, much like how successful ransomware attacks beget more ransomware attacks elsewhere. MIRI made such a point of it that members of LessWrong were confident that MIRI would never pay out:
In their paper on Functional Decision Theory, MIRI devotes a whole 4 pages to the XOR blackmail problem, specifically stating that it’s worse to pay out when someone can model your behavior better. But proponents of MIRI are now saying that the reason the payout wasn’t bad was because Louie Helm could model the behavior of MIRI staff so well, in direct contradiction with their published decision theory:
alyssa also said that because the blackmailer spent a long time getting to know miri/cfar and their psychologies that this made it /a better idea/ to pay out because it was less likely to happen again. instead of a worse idea because of increased subjunctive dependence.
Their willingness to pay out was especially relevant because of the scale MIRI claims to operate on: MIRI’s research, if they’re to be believed, is of international significance. AGI is the new nuclear arms race, and governments of the world want that technology for their own ends. Needless to say, were any particular government to successfully develop an AGI, the results would be catastrophic. Therefore, MIRI’s ability to withstand pressure from those interested in subverting their mission is central to their ability to carry out their mission. But it seems MIRI couldn’t withstand harassment from a single disgruntled ex-coworker. That alone should have sunk MIRI, however, few knew that MIRI had an extortionate relationship with Louie Helm, or that donor funds were used to pay him off.
It was only in November 2019 these allegations were brought to light by the blogger and activist Ziz, who detailed her experiences with both MIRI and its sister organization, CFAR. She alleged that CFAR engaged in transphobic hiring practices, and additionally blew the whistle on MIRI’s foray into statutory rape & donor fraud, after two board members confided in her that there’d been a blackmail payout. Moreover, in a conversation between Ziz and Anna Salamon, Ziz alleges Salamon expressed a relationship with her subordinates where she filtered on willingness to collude to keep the extortion and fraud secret, to the detriment of rationalists, donors, and the wider community that is outside Salamon’s circle of coordination.
I asked concretely what kind of tail outcome were they worried about. They said some they were afraid I’d do something that was bad for the rationality community. I asked for more details. They said some kind of drama thing. (I forget if it was during this conversation or elsewhere that they mentioned Alice Monday as an example of someone they thought was negative, and seemed worried when I said I had sort of been her friend and pupil. She linked me to the Gervais Principle.) I asked what scale of drama thing. I think the answered something big. I asked “like miricult.com“? (Unfortunately, I have not been able to find the final version of the website before it was taken down.) They said yes, like that.
I said I was pretty sure miricult was false. I think I said 98 or 95% sure. In a very very tentative, cautious voice they asked, “…what if it wasn’t false?”
What if I found about this, was asked to join in the cover-up? I said I’d turn him in. Like hopefully we could figure out a way for him to work on AI alignment research from prison. They asked, in more tones I should have paid attention to, what if you were pretty sure you could actually keep it a secret? I said if it was reaching me it wasn’t a secret. I said if Eliezer had chosen sex at such high cost to saving the world once, he’d do it again. But I wouldn’t drag down everyone else with him. I think they also asked something like, what if Eliezer didn’t think it was wrong, didn’t think anyone else would see it as wrong, and said he wouldn’t do it again. I said the consequences of that are clear enough.
Anna Salamon is the president and co-founder of CFAR. She’s been a member of LessWrong for 12 years, and has been on MIRI’s board since 2015. She claims to do a lot of work for MIRI, in spite of not being on the payroll. This wouldn’t be surprising, given that MIRI and CFAR share office space, and have joint staff meetings. They are functionally the same organization, with CFAR workshops being the primary hiring funnel for MIRI. Depending on who you ask, Salamon has a direct role in the hiring process for both MIRI and CFAR:
In 2016, Harmanas was interviewed for MSFP by Anna Salamon. He did not pass the relevant tests / math aptitude checks. That said, he did reveal some level of general-problem-solving or research aptitude (or something like this), to the point where Anna offered him an internship opportunity, working as a research understudy of hers. He was directly brought into CFAR’s inner culture and later became employed. Anna, from what I distinctly remember, did not consult anyone at CFAR about her decision before bringing him in as an intern. CFAR did not, as far as I know, do a background check or check references (but I also don’t think this would have helped at all).
Lauren Lee, the author of the second quote, was also an employee at CFAR. Together, these contradictory quotes suggest Salamon is involved in the hiring process for both CFAR and MIRI. And this matters because if she’s a transphobe, that provides a current source for MIRI and CFAR’s discriminatory hiring practices. CFAR has employed no trans woman ever, in spite of the higher than average number of mathematically gifted trans women interested in working for them.
On Anna Salamon: I remember her seeing me as a transwoman who exclusively liked men; Blanchard has a whole “theory” about this. And apparently that made me a “real” transwoman (wtf), unlike the rest of the transwomen in LW.
During a meeting with Ziz, Salamon asked invasive personal questions about Ziz’s sexuality and gender under the pretense of determining whether Ziz would be a good fit to work on AGI. Salamon’s questions were consistent with the kinds of questions someone might ask if they believed trans women were really perverted men, and were entirely unrelated to whether or not Ziz would make a suitable AGI researcher:
They [Anna] asked about my sexuality, I said I was bi. They asked if I had any fetishes. […] I was very uncomfortable sharing this stuff. But I saw it as a weighing on the scales of my personal privacy vs some impact on the fate of the world. So I did anyway.
Salamon holds that trans women aren’t really women, and that people believing they are the gender they’re assigned a birth is a fundamental and important test of whether someone is capable of doing epistemics:
Anna reacted like I was personally attacking her, and said, “but I need my gender test!”. She said gender was “really interesting and important and the first thing to understanding anything to humans”, and it was how you could tell if someone was going to be able to do epistemics, if they could overcome their personal biases.
[…] Filtering out trans women who believe they are women is, by the normal definition, filtering out trans women. How many hiring decisions has Anna had a say in? CFAR has had almost 30 employees, and no trans women. That is drastically below base rates in the Berkeley rationality community. Anna framed the gender test as a rationality test. But, it’s a submission test, a constraint-by-social-web test. E.g., see her opinions as rationality as agency vs social constraint here.
A History of Abuse and Transphobia
It should come as no surprise that MIRI and CFAR attract transphobic staff, given who shaped MIRI in its early days. One of MIRI’s original supporters who helped shape their organization was Peter Thiel. He is recognized on MIRI’s website as having donated $1,627,000 to MIRI through the Thiel Foundation. Thiel was the primary supporter of MIRI (then known as “The Singularity Institute,”) through the ’00s and half of the ’10s, and was not merely a removed donor, but had an active role in supporting MIRI. Thiel was a member of the advisory board between 2007 and 2014, after the organization moved to the Bay Area. He also appears to have been associated with MIRI prior to 2007, as he helped arrange the first Singularity Summit in conjunction with Stanford in 2006.
Peter Thiel claims to be a social libertarian & fiscal conservative, but he has a long history of supporting viciously anti-LGBT political candidates and PACs. And this isn’t limited to his recent support of the very anti-trans Trump administration. The candidates he supports through donations to the Club for Growth super-PAC have long histories of voting against LGBT rights. Thiel is their largest donor, and in 2018 gave $1 million dollars to Club for Growth.
Additionally, he’s donated $850,000 to the Freedom Forever PAC, which backs Kris Kobach’s campaign. Kris Kobach has compared marriage equality to drug addiction, claims LGBT parents are bad for the children, and says that the HRC promotes “homosexual pedophilia.” Kris Kobach also seems to have a special hatred for transgender people, saying, “trans athletes destroyed the dreams of female athletes who trained their whole lives for the honor of winning that championship only to have it snatched away by a biological male.”
In spite of his own sexuality, it seems Peter Thiel and those he funds take a hard stance against the LGBT community. He’s not the only transphobe associated with MIRI, either. The former board director & CEO, Michael Vassar, is also a noted transphobe, as is Jack Gallagher, a former researcher at MIRI.
…They spent 8 hours shouting at me, gaslighting me, trying to use me to get to “Emma”, Jack talking about how he hated trans women, especially hated me and my friends we were the most cringe, wanted us dead. Vassar kept telling me I needed to compromise with Jack and like the good parts of him that weren’t that. Said it was good Jack was screaming hate at me for most of those hours, because it showed that I was in bad faith…
This pattern of organization-wide transphobia is met with an equally concerning pattern of excusing abuse, and defending abusers. Anna Salamon let Robert Lecnik run local Bay Area rationalist meetups after she learned he’d sexually assaulted someone, giving him the opportunity to later sexually assault more people. Perhaps his behavior was excused because Lecnik’s husband is one of MIRICFAR’s top donors. Brent Dill, a rapist, was allowed to predate on the community from a position of authority for years, with CFAR staff like Tim Telleen-Lawton and Elizabeth Garrett eager to defend him, ignoring warning signs and reports of bad behavior. And this is the same Elizabeth Garrett who once scolded a workshop participant for referring to their physically and sexually abusive parents, “assholes.” Garrett recommended they go back to their abusive family and use “nonviolent communication” (NVC) to better get along with them. That participant did in fact go back to their abusive family for a year, and attempted to do NVC with the parents who had sexually and physically abused them, because they trusted that CFAR and Elizabeth Garrett had correctly identified what was wrong in that relationship.
We believe that Brent is fundamentally oriented towards helping people grow to be the best versions of themselves. In this way he is aligned with CFAR’s goals and strategy and should be seen as an ally.
Yudkowsky & Salamon’s participation in the cover-up of statutory rape, the lack of belief in his own decision theory, the fraud, Salamon’s transphobia and tolerance of abusive individuals, along with the suicides, psychotic breaks, and other downstream effects of a broken community culminated in the desire to see justice done.
Ziz and a few others, including former donors to MIRI, arranged a small peaceful protest against both organizations on the night of CFAR’s alumni reunion, at Camp Meeker’s Westminster Woods in Sonoma County. Leaders at CFAR and MIRI immediately moved to discredit them, painting them as “violent psychotic men.”
On the day of the protest, the protesters arrived two hours ahead of the reunion. They had planned to set up a station with posters, pamphlets, and seating inside the campgrounds. But before the protesters could even set up their posters, nineteen SWAT teams surrounded them. The four protesters were then brutally assaulted by the police and tortured for days until they could make bail.
All of this happened because someone deliberately and repeatedly lied to the Sonoma County police to convince them that the protesters had weapons or a bomb. CFAR staff, prior to the alumni gathering, informed staff at the campground to expect trouble from Ziz and her friends:
CFAR staff have talked about the protesters implying they’re “dangerous men”, so I wouldn’t be surprised if they told the campground that “violent deranged men” were planning to attack the event, scaring the campground staff into seeing physical threats where there were none. It’s unclear to me whether CFAR staff themselves or the campground called in the false report.
This is a fairly standard example of a phenomenon called “swatting” where someone deceives emergency services into sending police based on false reporting of a serious law enforcement emergency, such as a bomb threat or active shooter event. Swatting is a punishable offence in the state of California.
The Sonoma County police have so far been unwilling to admit that they were sent to Camp Meeker on false pretenses, and the state has filed trumped up charges against Ziz and the other protesters, in an effort to justify their misuse of tax dollars. After all, the police sent a helicopter, bearcat, 19 SWAT teams, and a bomb removal squad to terrorize a four-person peaceful protest.
After the allegations were revealed and the protest was foiled, a number of rationalists rushed to the defense of MIRI & CFAR. Most didn’t challenge the claim that community leaders were sleeping with teenagers, but did attempt to argue that paying out to blackmail in this instance wasn’t “that big of a deal.”
This has sparked ongoing controversy, with some members of the rationalist community arguing that MIRI’s researchers, by paying out, revealed that they didn’t believe in their own decision theory at all:
[MIRI’s decision] to handle future attempts to blackmail [them] by strictly controlling information is the strategy that Julian Assange predicted governments of the world would fall back to with the advent of wikileaks. And that was exactly the strategy Assange wanted them to implement to their detriment, because compartmentalization, paranoid distrust, and strict control over information often cripples organizations. There is a legitimate need to control information so it cannot be used by enemies, but it’s within an org’s best interests to keep that kind information to a minimum, so that its members can make better decisions, considering more available information. The same logic of defence-against-enemies-of-the-cause doesn’t apply to protecting its leaders from sex scandals, which does not advance the goal of AI safety.
Control of information is also a strategy closely associated with the premise that your organization needs to do things that would upset, scare, or otherwise repulse your supporters to survive. (See drives for transparency in government, etc.) MIRI’s representatives in that group are de facto declaring such an intent (or are at least proving themselves insensitive to it) by not correcting the OP[Alyssa Vance] when she said [she heard from MIRI that] MIRI intended to better control the flow of information in the context of controlling information of miricult’s accusations.
Figuring out how to be more sophisticated at making unethical decisions is not an improvement on the situation.
If you do things that are likely to upset, scare, or otherwise repulse your supporters, you’re creating an incentive to blackmail you. Sometimes it may be necessary to do upsetting things, but keep in mind the specific context we’re talking about, here. If you fear blackmail for wrongdoings like accusations like rape and fraud, then you fear anyone who knows those secrets, and fear people discovering those secrets. This fear has downstream effects, like worrying that new people brought aboard will blow the whistle on your behaviour. So then you end up associating with people willing to tolerate increasingly bad behaviour. And as time goes on, you may attract opportunistic abusers who believe they’ll be right at home in your organization. And people like that who are in positions of power often engage in behaviour that will result in more scandals, opening you up to even more blackmail. Inner rings aren’t uncommon or even necessarily bad by themselves, but what secret you base your ring around is critical, because the ethics of the people you hire will shape the character of your organization and even your community.
Organizations like MIRI are reliant on whistleblowers and transparency to prevent collusion. But MIRI and CFAR have taken pains to discredit their whistleblowers, and shield themselves from scrutiny, violating both US law and their own whistleblower policy. And without that scrutiny, it is impossible for donors to discern whether or not MIRI’s researchers actually believe in the research that donors are funding.
Aside from being banned from MIRI and CFAR, whistleblowers who talk about MIRI’s involvement in the cover-up of statutory rape and fraud have been banned from slatestarcodex meetups, banned from LessWrong itself, the MIRIx Discord server, and have been harassed by community members.
Some comments on CFAR’s “AMA” were deleted, and my account was banned. Same for Gwen’s comments. Oliver Habryka, affiliated with CFAR, the same man who threatened frivolous defamation lawsuits to deter me from relating what Vassar said, said we had “a sufficient track record of being aggressive both online and offline”. I suppose he’s referring the protest where we got swatted (19 SWAT units if I remember off the top of my head, armored cars, a helicopter, a bomb squad, because of repeated deliberate false reports of us as active shooters), sexually assaulted by cops, imprisoned, bruised, starved, deprived of sleep, withheld water, withheld clothes, withheld medication, legal counsel, set $200k total bail on felony charges by a prosecutor admitting we didn’t commit any violence, but pleading with the magistrate about the cost of the operation to swat us…)
One community member went so far as to call in additional false police reports on the whistleblowers, and wrote a false smear site on Ziz to discredit her, in an attempt to punish her for not wanting to sleep with him. Meanwhile, others threatened them with various kinds of legal action.Through the sum of the community’s behavior, it’s apparent that there is a worm at the heart of the rationalist community that has spoiled the bunch. It will not stand.
Community support of MIRI has dwindled following the November 2019 protest, but a few major donors, including you, the Open Philanthropy Project, Jaan Tallinn, and an anonymous donor are still supplying MIRI with the funds to build a new compound somewhere in the Northeastern US. Until now, it has remained a question whether you have continued to donate out of indifference or ignorance.
This letter asks that you stop supporting MIRI, CFAR, and the organizations that continue to donate to them. Their current plan to get a “demon in a box” and yank the wheel away from UFAI is doomed to failure, and they’ve given up on developing FAI, themselves. Moreover, Salamon’s filter specifically selects for people who are incapable or unwilling to engage with with ideas seriously. They are no longer a legitimate institution.