I am writing this at a turning point in my life, a moment of great distress, pain and anguish. In the hope of clearing up a dark chapter of US history, the personal emails of the notorious financier Jeffrey Epstein were opened to the public. We did not get much more clarity so far, but anyone ever exchanging private emails with Epstein entered the spotlight, including myself. While my friends and coworkers understand the context of these messages, many people in the general public do not, and I find myself attacked by strangers, even subjected to death threats, and many people who take an interest on my work on the philosophy of cognition, consciousness and artificial intelligence feel taken aback, shocked and distressed.
How I met Jeffrey Epstein
As many people know, during the years of 2013 to 2017, a part of my work at the MIT Media Lab and the Harvard Program for Evolutionary Dynamics was funded by Jeffrey Epstein. I met him through several well-known scientists from the fields of AI and psychology, whom I deeply respect and admire for their contributions to the field, and whom I consider to have very high integrity: Ben Goertzel, Stephen Kosslyn, and Roger Schank. They recommended to him that he should fund some of my work, and to me, that I accept his funding, as did other brilliant researchers I met at MIT and Harvard. (Epstein never influenced the topics or goals of my research, and never tried to do so.) Several years earlier, Epstein had been convicted for the trafficking of minors and had spent a year in prison. While Epstein’s public reputation was forever destroyed, many people whose judgement I trusted assured me that he had reformed himself and was committed to staying on the good side of the law. I have not met a single person from his network of academics who was aware of any instance of him breaking the law after his conviction, or who witnessed or condoned any illegal or questionable activity by Epstein.
Epstein had an unusual psychology. He was high strung, intensely curious, and utterly devoid of fear, guilt or shame. In the absence of these basic human traits, he was a high functioning sociopath, bound only by rules and loyalties he discovered for himself. He was incredibly well connected, with friends in all political camps, including existing and former heads of state in the US and all over the world. His circle of acquaintances and friends also included a number of the most famous mathematicians, physicists, cognitive scientists, AI researchers, economists and biologists of our time. Epstein was not a scientist himself, and regarded the consensus mechanism of peer-reviewed science with skepticism. Instead, he generated his own models of the world, and bounced them against the best minds he could find.
From time to time, Epstein would invite a circle of top tier scientists and high ranking politicians into an office at Harvard, discussing science and politics, with radical openness and disregard for ideological camps and public taboos. During some of my time in Cambridge, Epstein sent frequently short, dyslexic emails with random thoughts in my direction. I tried to probe and understand his world view, which was highly unusual and often darker and more radical than anyone else I’ve ever talked to.
The context of my argument with Jeffrey Epstein
During an in-person discussion with the linguist Noam Chomsky in 2017, we discussed what makes the human species so profoundly different from other apes and animals. Chomsky argued that humans possess a unique mutation that gives us universal compositional grammar. While this is a great hypothesis (that Chomsky is well known for), I disagreed, and argued that human and the brains of other great apes may lie on a continuum, with the difference being the length of childhood neuroplasticity. Chomsky dismissed my argument. In an email to Epstein, I tried to explain my thinking.
Humans have an extraordinarily long childhood, spending more time in early infancy, as a toddler, as an adolescent etc. Each cognitive layer takes a considerable time to train, before neuroplasticity subsides and the next layer is trained. Great apes reach intellectual maturity within a much shorter period of time than humans. I argued that from the perspective of AI, apes process less training data at every stage. If that hypothesis is true, it creates a counterintuitive, testable prediction: When individuals show slower development, they generally don’t end up at the same level of cognitive development as others in the same population (because slower development signals an organic problem). However, when groups show slower development, it may indicate a genetic switch that prolongs neuroplasticity, thus leading to development of more abstract/symbolic cognition.
This idea mirrors how the world of AI has changed its view on intelligence: before the widespread success of LLMs, most of the field thought of intelligence as a particular set of specific mechanisms, similar to Chomsky’s universal grammar mutation. Today, most of the AI field sees intelligence as the result of scaling the amount of training data and computation.
I got my idea of the “human scaling hypothesis” from watching a documentary called “Babies”, which tracks the first year of infants in several different countries. The movie shows how babies in Namibia have much faster development (motor and social) than children in Japan and the USA, despite being less stimulated by their parents. That documentary is not scientific evidence (it just observes a few babies without commentary), but there is data from preschoolers in the US that suggests a similar inverse relationship between speed of development and cognitive attainment between African American and white/Asian children. (https://pubmed.ncbi.nlm.nih.gov/3293732/, https://academic.oup.com/aler/article-abstract/8/2/249/149636)
I am extremely uncomfortable discussing topics that touch on heritability of intelligence in public, because scientific nuance is lost when it becomes instrumentalized by racists.
Race does not cause differences in behavior and in cognitive attainment. It is also not predictive of individual traits, and the census categories of race are not even aligned with ancestry or ethnicity. Nevertheless, race is heavily weaponized by voices on the far right, and used to construct differences in human value. As a German, I am especially aware of the consequences of racism and ethno-fascism. I have seen the horrors of Nazi extermination camps while visiting them as a child. The belief that one race is somehow superior to another is vile and a source of great historical evil.
Determining degrees of heritability is also very complicated, and I have no special expertise to cut through the data and possible political motives of people who present them (e.g. traditional twin studies paradigms put heritability estimates far higher than the most recent estimates from molecular genetics; see (Astral Codex Ten: Missing Heritability) for some of that debate).
A similar difficulty exists around the nuances of sex differences in interests and cognitive traits. While scientific studies have long observed statistical differences, it is very hard to identify and separate cultural and genetic components, and individual performance is not caused by sex or gender. At the same time, arguments about sex differences get invariably instrumentalized in political debates between the left and the right, which makes public discourse almost impossible.
Portraits in my living room in Berlin, 2013: Karl Marx (as a student), Noam Chomsky, Boris Vian
I think it’s despicable to judge people not on their behavior but on their ancestry, skin color, sex or gender, and I don’t think that the value of a human being is constituted by traits like mathematical or verbal intelligence, or social skills. I am not interested in supporting racists and sexists in any way. My interest lies in understanding how minds work.
What it is like to get caught in a media firestorm
A number of people who read the conversation between me and Jeffrey Epstein have taken them as an endorsement of racist, even eugenicist ideas. I found myself accused by social media activist of wanting to change the genetics of black children (which would not just be an evil idea, but is also scientifically absurd, a cartoonish supervillain fantasy out of a 1970s Bond movie). A climate reporter from the Boston Globe headlined her article with calling me racist and sexist.
The consequences of this firestorm have been devastating. While I have the unconditional support of my friends, who know my heart and soul, I receive death threats on social media. Presentations, a seminar and public talks have been cancelled. Some business partners severed their ties, with harsh financial and social consequences for me and my family.
The experience of standing in the middle of this storm is traumatic. It is hard to feel my body. At times, I find it nearly impossible to breathe. I understand why targets of this sort of media maelstrom do not always survive, and need a long time to heal. At the same time, my family and friends need me to go on, and my work is too important to give up on it. It has just gotten harder for a while.
I feel pain and despair, but I don’t feel any anger. Not at myself, because I know that I tried to do the right thing. I am not angry at the people who hound me, because this is just how public discourse works today. I can also see and understand the pain of people who feel hurt by words directed to interact with the worldview of a sociopath, ideas they interpret as racist and cruel. I experience the stress my friends and coworkers are being put under.
The difference between normative (good/bad) and descriptive (true/false) arguments is falling by the wayside when it comes to matters of ideology and politics. There is a good reason why many discussions about politically fraught subjects can only be conducted in private. A liberal society requires freedom of thought, and not just freedom of speech, but it also also a right to speak in private, to discuss strange ideas, even with disagreeable people and sociopathic billionaires, to go where our quest for truth leads us.