"The TESCREAL bundle: Eugenics and the promise of utopia through artificial general intelligence" is a peer-reviewed paper by Timnit Gebru and Émile P. Torres, published in First Monday 29(4), April 2024. It is the fullest statement of the TESCREAL framework, which the wiki had previously cited only by URL.
The argument
The paper opens on what it treats as an unexamined premise: "The stated goal of many organizations in the field of artificial intelligence is to develop artificial general intelligence… Without seriously questioning whether such a system can and should be built, researchers are working to create 'safe AGI' that is 'beneficial for all of humanity.'"
Its two claims are separable, and the paper is clearest when they are read apart.
The engineering claim. "Unlike systems with specific applications which can be evaluated following standard engineering principles, undefined systems like 'AGI' cannot be appropriately tested for safety." Safety testing requires a specification of intended function; a system defined as capable of "performing any task under any circumstance" supplies none, so no test can establish its safety. This argument stands independently of the historical thesis.
The genealogical claim. Asking "what ideologies are driving the race to attempt to build AGI," the authors analyze "primary sources by leading figures investing in, advocating for, and attempting to build AGI," and "trace this goal back to the Anglo-American eugenics movement, via transhumanism." The TESCREAL bundle names the resulting genealogy — "transhumanism, Extropianism, singularitarianism, (modern) cosmism, Rationalism, Effective Altruism, and longtermism" — described as "direct descendants of first-wave eugenics" that "emerged in roughly this order," with "many… shaped or founded by the same individuals."
The connecting argument is that the discriminatory attitudes of historical eugenics did not stay behind: "many of the very same discriminatory attitudes that animated eugenicists in the past (e.g., racism, xenophobia, classism, ableism, and sexism) remain widespread within the movement to build AGI, resulting in systems that harm marginalized groups and centralize power, while using the language of 'safety' and 'benefiting humanity' to evade accountability."
The eugenics distinction
The paper distinguishes two waves, and the distinction carries its argument. First-wave eugenicists "strove to improve the 'human stock' by altering society-wide patterns of reproduction, a process that would require many generations to work." Second-wave eugenics "arose in response to new technological possibilities associated with genetic engineering and biotechnology," under which "across just one generation, parents could potentially 'design' their children."
The second wave "claims to be 'liberal,' emphasizing the freedom of parents to decide whether, and how, to produce 'enhanced' offspring" — a self-description the authors report and then contest, noting that "some philosophers have argued that in practice this new, 'liberal' eugenics" reproduces the coercive dynamics it disclaims. The TESCREAL bundle is positioned as a further radicalization along this line.
Evidence offered
The historical link is argued through funding and authorship rather than assertion. The paper notes that "the first book on AGI (Pennachin and Goertzel, 2007) was co-authored by a transhumanist, cosmist, and participant in the Extropian movement whose express aim was to create 'transhuman AGI'," and that "much of the billionaire funding for projects focused on AGI comes from wealthy individuals aligned or explicitly affiliated with one or more of these ideologies."
It also situates the AGI goal historically: after the AI winters of the 1970s and 1990s — caused in part because researchers "did not want to be associated with grandiose claims that researchers didn't deliver on" — the field "was mostly focused on building specialized systems that some call 'narrow AI'." The recent return of AGI as a stated goal is what the paper treats as requiring explanation.
Present harms. The paper's case does not rest on future risk. It catalogs documented harms already attributed to the AGI race: "worker exploitation, data theft, environmental racism, the spread of misinformation and disinformation, plagiarism, and systems that amplify hegemonic views like racism, ableism, homophobia, and classism" — citing among others Bender and Gebru et al. (2021).
Methodology and standpoint
The authors disclose their positions rather than claiming neutrality: "One of us is an electrical engineer who has worked in the tech industry for nearly two decades (with more than a decade of that in the field of AI). Another is a philosopher and historian who was a noted contributor to, and advocate of, the TESCREAL movement for nearly a decade before leaving the movement." Torres's prior participation is presented as the basis of access to primary sources and internal discourse, and is also the ground on which critics have questioned the paper's neutrality.
Recommendation
The paper's closing move is constructive rather than prohibitive: it "urg[es] researchers to work on defined tasks for which we can develop safety protocols, rather than attempting to build a presumed all-knowing system." The recommendation follows from the engineering claim rather than the genealogical one — well-specified systems admit safety protocols, and undefined ones do not.
Relationships
- depends-on: TESCREAL Framework (Gebru and Torres) — the framework's fullest statement, replacing the inline URL citation that page carried
- related: Stochastic Parrots (Bender, Gebru, McMillan-Major, Mitchell, 2021) and the Octopus Test — same lead author; the documented-harms catalog draws on it
- related: Effective Altruism, Longtermism, AGI Timelines, AI Power Concentration
- related: Timnit Gebru