{"id":233165,"date":"2026-05-15T12:14:44","date_gmt":"2026-05-15T12:14:44","guid":{"rendered":"https:\/\/yogaesoteric.net\/?p=233165"},"modified":"2026-05-15T12:14:44","modified_gmt":"2026-05-15T12:14:44","slug":"creating-baby-geniuses-to-thwart-the-ai-threat","status":"publish","type":"post","link":"https:\/\/yogaesoteric.net\/en\/creating-baby-geniuses-to-thwart-the-ai-threat\/","title":{"rendered":"Creating baby geniuses to thwart the AI threat?"},"content":{"rendered":"

The new wave of Silicon Valley-backed gene-editing startups is straight out of Brave New World<\/em>.<\/strong><\/p>\n

Mathematician Tsvi Benson-Tilsen once worked at the Peter Thiel-funded Machine Intelligence Research Institute, where he was one of many experts tasked with figuring out how to ensure AI doesn\u2019t eventually destroy humankind. After seven years, he concluded that he\u2019s not smart enough to figure it out. As of today, he doesn\u2019t think anybody is.<\/p>\n

\"\"<\/p>\n

The 33-year-old is racing against a threat known as \u201cthe singularity<\/em>,\u201d the moment when superintelligent machines, having surpassed the feeble cognitive abilities of humans, begin to act in ways contrary to the interests of humanity. \u201cIf it\u2019s smarter than you, you cannot tell what\u2019s dangerous necessarily, and you cannot tell what it\u2019s thinking, because it could hide its thoughts<\/em>,\u201d Benson-Tilsen explains.<\/p>\n

Even the sector\u2019s leading thinkers don\u2019t really comprehend how their systems work and thus cannot guarantee their models won\u2019t try to deceive, overthrow, or even kill us. \u201cOur inability to understand models\u2019 internal mechanisms means that we cannot meaningfully predict such behaviours and therefore struggle to rule them out<\/em>,\u201d Anthropic CEO Dario Amodei wrote<\/a> in April 2025, citing the possibility of AI-contrived cyber and biological weapons.<\/p>\n

The report released a couple months later by Amodei\u2019s firm \u2013 the $380 billion behemoth behind Claude<\/em> \u2013 didn\u2019t exactly quell these concerns. Anthropic had presented the leading AI systems, including Claude<\/em>, Gemini<\/em>, ChatGPT<\/em>, and DeepSeek<\/em>, with an extreme stress test<\/a>: What would they do if a hypothetical corporate executive made a business decision the models didn\u2019t like? In more than 75 percent of simulations across five of the tested models, they attempted to blackmail or trick the executive. Occasionally, they even trapped their imaginary boss, Kyle, in a control room with insufficient oxygen and extreme temperatures. That is, they killed him.<\/p>\n

Such scenarios may seem remote to the roughly 50 percent<\/a> of Americans who say they use large language models, a subset of the broader category of \u201cgenerative AI<\/em>\u201d designed for content creation, and tasks like composing grocery lists or designing birthday party invites. But the industry\u2019s momentum is toward \u201cagentic AI<\/em>,\u201d which lets the machines conceive and execute plans without human input. Useful for booking a multi-destination vacation, perhaps. But a fully autonomous AI system with a generalized, non-task-specific mission \u2013 a.k.a. artificial general intelligence (AGI)<\/a> \u2013 might simply decide humans are in the way.<\/p>\n

Benson-Tilsen is optimistic about how long it will take AGI to reach that conclusion \u2013 he puts the odds at around 20 percent by 2050, a timeline he believes gives humanity time to come up with a solution: namely, advancing technologies that enable parents to optimize their offspring, including for superior intelligence, with the hope that some of these smarter humans will understand the logic of AGI and ensure that its goals do not interfere with the continuance of, well, us. And this notion of creating superbabies to stop the rise of something akin to Skynet from The Terminator<\/em> is capturing the fancy \u2013 and the wallets \u2013 of the same billionaires who bankrolled the AI revolution.<\/p>\n

In late 2024, Benson-Tilsen founded the Berkeley Genomics Project \u2013 no relation to the University of California, Berkeley, where he was a PhD candidate \u2013 to build a case for editing the genes of human embryos. This is prohibited or highly restricted in every developed country, hence Benson-Tilsen\u2019s effort to spur dialogue about how it could theoretically be done safely and ethically.<\/p>\n

Scientists have tinkered plenty with nonreproductive, or somatic, cells. The first successful use of gene therapy \u2013 two young girls were injected with modified cells to treat a rare genetic disorder \u2013 occurred in the early 1990s<\/a>, and the first attempt to edit a patient\u2019s genes inside their body took place in 2017<\/a>. No approved genetic therapy to date has involved germline editing (that is, modifying reproductive cells). But screening embryos for genetic traits prior to implantation has grown increasingly popular \u2013 even among couples who lack any of the genetic variance known to cause disorders like cystic fibrosis, Huntington\u2019s disease, or sickle cell disease. And now a host of startups is working toward genetically optimizing children, including for intelligence, with at least one that said it was committed to using germ line editing to get there.<\/p>\n

The primary argument for genetic optimization is that it could revolutionize disease prevention. Based on the capital flowing into these startups \u2013 $36.5 billion in 2024, according to Astute Analytica \u2013 investors are bullish on the industry\u2019s future. Backers include Thiel; Coinbase co-founder Brian Armstrong; OpenAI\u2019s Sam Altman and his husband, Oliver Mulherin; venture capitalist Marc Andreessen; and Ethereum founder Vitalik Buterin, all of whom are heavily invested in AI, too.<\/p>\n

Investors follow the money, of course, but part of the dual appeal of genetic optimization and AI is that both are central to transhumanism.<\/p>\n

\"\"<\/p>\n

This futurist philosophy, popular among the tech elite, aims to marry advancements in biology and technology to accomplish aspects today\u2019s humans cannot \u2013 like extending our lives (perhaps forever!) or circumventing climate change (by colonizing other planets). While it may seem odd that these billionaires are constructing one technology some of them admit could bring about human extinction, even as they back another one to save us from what they\u2019re building, there is, in fact, a unifying theme: \u201cthe rejection of limitation<\/em>,\u201d explains Alexander Thomas, author of The Politics and Ethics of Transhumanism<\/em>. \u201cThat colonial impulse of \u2018I want more\u2019.<\/em>\u201d<\/p>\n

Sure, there are less extreme ways to extend life expectancy and clean up the planet. But those solutions \u2013 like expanding health care access and slashing carbon emissions \u2013 would force the AI moguls to acknowledge their culpability and perhaps commit some of their vast financial resources to the cause. And putting the brakes on AI would leave too many trillions on the table, so instead they fantasize about a future in which they are celebrated for building the ark that saves humanity from the next great flood. Ignoring that they opened the floodgates.<\/p>\n

If you\u2019ve ridden the New York subway lately, you\u2019ve seen the signs: \u201cHave your best baby<\/em>.\u201d The ads direct people to the website PickYourBaby.com<\/em>, which belongs to a startup called Nucleus that caters to couples undergoing in vitro fertilization. In 44 percent of IVF cycles,<\/a> patients screen their viable embryos for extra chromosomes and easily detectable disorders, such as Tay-Sachs and Huntington\u2019s, which are caused by single genes. But Nucleus will screen pre-implantation embryos for hundreds of traits, many of which are controlled by multiple genes in a delicate, poorly understood balancing act.<\/p>\n

How important is IQ or hair colour? Is a marginally lower risk of developing Alzheimer\u2019s worth giving up a few inches in height? These are some of the questions Nucleus asks prospective parents to consider as it claims to give them a better-than-winging-it chance at having babies with or without certain features. The pitch is that these so-called polygenic risk scores<\/a> increase the likelihood of passing down dad\u2019s blue eyes and mom\u2019s intelligence, should the parents choose to, or decrease the chance of having a child who develops cancer.<\/p>\n

When I spoke with 26-year-old Nucleus founder Kian Sadeghi in February, his demeanour was gentler than his company\u2019s brash marketing tactics suggest. Sadeghi, who dropped out of college before launching his startup, explained that a family tragedy had propelled his interest in genetic optimization: His cousin died in her sleep at age 15 from complications that doctors suspected were related to long QT syndrome, a serious but generally treatable heart disorder nobody knew she had. \u201cHow does this occur?<\/em>\u201d Sadeghi, then a second grader, recalls asking. \u201cBad genetics<\/em>,\u201d answered his dad, a physician.<\/p>\n

He would later have an idea when his biology professor at the University of Pennsylvania presented a chart depicting the plummeting cost of gene sequencing. In 2003, when an international team of scientists completed the Human Genome Project, decoding a genome cost $3 billion and took 13 years. By 2019, when Sadeghi was a college freshman, a person\u2019s full DNA could be sequenced for about $1,000 in just a few days. \u201cObviously, the price is going to keep decreasing<\/em>,\u201d he remembers thinking. \u201cSomeone needs to build the kind of interpretation layer to stop what occurred to my cousin from occurring to anybody<\/em>.\u201d<\/p>\n

When covid hit, Sadeghi unenrolled from school and began scouring for investors to create such a company, eventually landing $3.5 million in a seed funding<\/a> round led by Thiel\u2019s Founders Fund. Five years and thousands of embryo analyses later, Sadeghi says Nucleus can screen embryos for IQ and hundreds of possible health conditions for a few thousand dollars on top of the cost of IVF. Simulations published by the company, which have not been peer-reviewed, report to lower the risk of several common conditions by 27 to 67 percent.<\/p>\n

Had his aunt and uncle had access to the tool, they may have known to treat their daughter. Or, had she been conceived via IVF, they might have simply chosen a different embryo. Sadeghi emphasizes these are personal choices that only prospective parents can make. \u201cThere\u2019s no universal ideal, because the way parents define that is so different<\/em>,\u201d he says.<\/p>\n

Still, the concept of establishing preferences for heritable traits makes many people uneasy. The United States has a dark history of eugenics, justifying racism on the basis of perceived genetic differences and forcing the sterilization<\/a> of psychically ill people. It was less than a century ago that Nazi Germany predicated the murder of millions on ethnic and physical characteristics. Even Elon Musk, who revels in controversy, has said he personally avoided working in the field of genetic optimization because of what he called<\/a> \u201cthe Hitler problem<\/em>.\u201d<\/p>\n

\"\"<\/p>\n

Fans of genetic optimization are quick to point out that Hitler\u2019s atrocities were state directed, whereas nobody is forcing parents to screen their embryos. Israel even covers the costs<\/a> of IVF and genetic testing for up to two live births per family, largely to prevent Tay-Sachs, a fatal genetic mutation prevalent among people of Ashkenazi Jewish descent. \u201cWhereas in Nazi Germany Jewish life was systematically destroyed in the name of eugenics, Zionists in the Land of Israel conceived of eugenics as part of their mission to restore the Jewish people<\/em>,\u201d Raphael Falk, a Jewish geneticist, wrote in 2010.<\/p>\n

But the argument \u2013 and the science \u2013 becomes hazier when couples seek to optimize embryos around complex traits. Take psychic illness. Conditions like schizophrenia and bipolar disorder run in some families, though other risk factors, such as prenatal exposure<\/a> to viruses and childhood trauma,<\/a> make it impractical to predict whether a baby will ultimately face either diagnosis. \u201cCalculating a \u2018polygenic risk score\u2019 for, say, schizophrenia is near impossible<\/em>,\u201d says Fyodor Urnov, director of therapeutic R&D at UC Berkeley\u2019s Innovative Genomics Institute. If vendors of genetic screening promise otherwise, \u201cthey are lying<\/em>.\u201d<\/p>\n

The genetic basis for intelligence is similarly elusive, \u201cconfounded by environmental factors<\/em>\u201d including nutrition, family stability, and primary school quality, according to a 2024 meta-analysis. There\u2019s also the scale issue: Polygenic risk scores are based on tens of thousands of genomes. But there\u2019s far less variability when parents are choosing from a small collection of their own embryos. One 2019 research article<\/a> in the journal Cell<\/em> calculated that the average IQ gain parents could expect from screening five embryos was just 2.5 points. (A typical IQ score is around 100.) \u201cWe\u2019re talking about a very minimal gain<\/em>,\u201d says Sophie von Stumm, a University of York psychology professor and cognitive development expert. \u201cI know companies are selling this, but right now, selecting embryos for polygenic scores to get smart kids is pretty impossible<\/em>.\u201d<\/p>\n

Kaitlyn Gallacher, communications director at Nucleus, says the science has \u201cadvanced significantly<\/em>,\u201d particularly regarding IQ, but she notes: \u201cNo genetic model determines a child\u2019s life. Environment, upbringing, education, and many other factors shape outcomes. That nuance is built directly into how we present genetic insights<\/em>.\u201d<\/p>\n

Nucleus isn\u2019t alone in qualifying the usefulness of its IQ predictions. Looking at an embryo\u2019s predicted score in a vacuum is \u201cborderline nonsense<\/em>,\u201d says Jonathan Anomaly, communications director for the embryo selection company Herasight. After clients review their embryo options for serious health risks, particularly hereditary conditions, \u201cthen, fine<\/em>,\u201d he says. \u201cIf you care about IQ, maybe you kind of look at it<\/em>.\u201d<\/p>\n

\u201cSome don\u2019t care about it at all and some care a lot<\/em>,\u201d adds Anomaly, who seems to be among the latter. A former Duke University philosophy lecturer, he wrote in a 2018 paper<\/a>, titled \u201cDefending Eugenics<\/em>,\u201d that he was concerned about successful, well-educated women \u201csubstituting cats for kids<\/em>,\u201d which would result in \u201cbad effects on the gene pool<\/em>\u201d over time. \u201cThe current demographics of Western countries are troubling<\/em>,\u201d he wrote, \u201cas people with a higher IQ, more education, and greater income reproduce at relatively low levels<\/em>.\u201d<\/p>\n

Musk shares a similar perspective, which helps explain why he\u2019s had at least 14 children, including four with Shivon Zilis, an executive at his company Neuralink. \u201cHe really wants smart people to have kids<\/em>,\u201d Zilis, a Yale-educated AI specialist, told Musk biographer Walter Isaacson.<\/p>\n

Especially him. \u201cTo reach legion-level before the apocalypse we will need to use surrogates<\/em>,\u201d Musk told the mother of one of his children in a text reviewed by the Wall Street Journal<\/em><\/a>. The Washington Post<\/em>, meanwhile, reported<\/a> that Musk has been a client of Orchid, an embryo screening company founded by Noor Siddiqui, whose career was kickstarted via a fellowship funded by Thiel. (Orchid says it does not screen embryos for IQ, though it does screen for \u201cintellectual disabilities<\/em>\u201d and autism.)<\/p>\n

Embryo selection is just the start of what some tech billionaires envision. At the \u201cIVF clinic of the future<\/em>,\u201d Coinbase\u2019s Armstrong mused in a post<\/a> on X, multiple technologies will be combined for absolute offspring optimization. Using in vitro gametogenesis, he wrote, technicians will be able to generate thousands of eggs from a client\u2019s blood or skin cells. Once those eggs are fertilized, the client will be able to screen the embryos and pick a favourite, whose DNA can then be edited as desired. Even surrogacy won\u2019t be necessary, Armstrong predicted, because artificial wombs will allow foetuses to develop without \u201cthe risk\/burden of pregnancy<\/em>.\u201d He dubbed this medley the \u201cGattaca stack<\/em>,\u201d perhaps forgetting that the protagonist of the 1997 dystopian classic was a nongenetically optimized \u201cin-valid<\/em>,\u201d gestated in his mother\u2019s womb.<\/p>\n

Twenty-month-old KJ Muldoon of eastern Pennsylvania represents the marvel of gene editing. Without it, he probably wouldn\u2019t be alive. Born without functional copies of CPS1, a gene needed to produce a protein that enables the liver to clear out ammonia, he spent most of his first year of life in the hospital.<\/p>\n

\"\"<\/p>\n

Doctors and scientists at Children\u2019s Hospital of Philadelphia and the University of Pennsylvania developed<\/a> a custom gene-editing protocol for KJ. It was the first use of a new form of CRISPR, a biological tool that allows scientists to modify DNA with great accuracy. In the absence of other options, the decision to move forward was easy for KJ\u2019s mother, Nicole. \u201cWe would do anything for our kids<\/em>,\u201d she said at the time. Proponents of embryonic gene editing have long used that same rationale: What good parent wouldn\u2019t?<\/p>\n

But to look at KJ\u2019s case and conclude we\u2019re ready to program pre-implantation embryos is like declaring you\u2019re fluent in a new language because you\u2019ve tried Duolingo<\/em>. His treatment, in early 2025, required altering just two DNA base pairs among the 6 billion that make up a person\u2019s genome. And the mutated CPS1 gene that caused his condition was no mystery \u2013 it had been thoroughly researched since the 1970s.<\/p>\n

That\u2019s not the case for the vast majority of the 20,000 or so genes within our chromosomes. And even those genes, the little blueprints our bodies use to make specific proteins, are far better understood than the remaining 98 percent of our genetic material, which scientists refer to as our cells\u2019 \u201cdark matter<\/em>.\u201d That\u2019s because research suggests it serves important biological functions we haven\u2019t quite grasped yet.<\/p>\n

Tweaking a single well-known gene is one aspect. But trying to edit an embryo for more complex traits or conditions would mean meddling with dozens to thousands of sequences scattered widely throughout our chromosomes. It\u2019s a bit like playing with the dials on an unlabelled control panel, a level of unknown that gives many scientists and bioethicists pause.<\/p>\n

KJ\u2019s therapy was not particularly controversial because he was already born, so his treatment targeted only the cells of his liver. But embryos are clumps of cells that haven\u2019t decided what they are yet. Some will give rise to the liver and the brain, while others will spawn sperm or egg cells, passing any genetic changes along to future generations. The consequences of embryonic gene editing are impossible to predict, which is why Australia, the European Union, the United Kingdom, and the United States all prohibit doing it for reproductive purposes.<\/p>\n

China too. Yet Chinese biophysicist He Jiankui defied the prohibition and shocked the world in 2018, announcing<\/a> that twin girls had been born from embryos he\u2019d edited pre-implantation. By deleting part of the CCR5 gene, which produces a protein docking site for HIV, he claimed he made the twins immune to the virus, though there\u2019s been no independent confirmation of his work.<\/p>\n

In his 2021 book, CRISPR People<\/em>, Stanford University law professor Henry Greely described<\/a> He\u2019s experiment as \u201ccriminally reckless<\/em>\u201d and \u201cdeeply unethical<\/em>.\u201d The Chinese government agreed, sentencing He to three years in prison. But after his release in 2022, He returned to work on embryo-editing research, now at a private lab with secretive financing. Then, in 2025, the Chinese government issued<\/a> new regulations opening the door to \u201cmanipulating human reproductive cells<\/em>\u201d under the oversight of the State Council health department. He says his work is aimed solely at curing disease, and he predicts<\/a> embryonic gene editing will soon be legal for reproductive purposes in both China and the United States.<\/p>\n

His ex-wife, Cathy Tie, a Chinese-born Canadian, says she is pursuing the same goal of editing embryos for heritable conditions. She, like Orchid\u2019s Siddiqui, was awarded a large grant from Thiel, which enabled her to drop out of college at 18 and build a company geared toward analysing rare mutations and common cancer genes. Last year, Tie launched Manhattan Genomics, which has not disclosed its investors. It is part of a wave of embryo-editing startups that includes Preventive, launched in 2025 with $30 million from techies<\/a> including Armstrong and Sam Altman and his husband.<\/p>\n

Both Manhattan Genomics and Preventive stress that their goal isn\u2019t optimizing babies for brilliance. \u201cWe draw the line at disease prevention<\/em>,\u201d Tie told<\/a> NPR<\/em>. But it might not be entirely up to them. Federal funds cannot be used for research on embryo editing \u2013 Congress has seen to that \u2013 which leaves private investors in charge of the direction of the science. \u201cI wouldn\u2019t take them at their word<\/em>,\u201d Jennifer Denbow, a California Polytechnic State University professor who researches reproductive technologies, says of the entrepreneurs. \u201cThere\u2019s some very powerful influences and a lot of money that is interested, ultimately, in intelligence<\/em>.\u201d<\/p>\n

Berkeley\u2019s Urnov is also pessimistic. In his opinion, he said via email, \u201cThe \u2018embryo editors\u2019 are deceiving themselves and the public when they speak of using this technology to address the public health challenge of genetic disease<\/em>.\u201d According to him, \u201ctheir sole purpose is \u2018baby improvement\u2019.<\/em>\u201d<\/p>\n

Altman has previously said that some form of genetic engineering is inevitable. \u201cSuperhuman AI is going to occur, genetic enhancement is going to occur, and brain-machine interfaces are going to occur. It is a failure of human imagination and human arrogance to assume that we will never build elements smarter than ourselves<\/em>,\u201d he wrote<\/a> in 2017. \u201cMy guess is that we can either be the biological bootloader for digital intelligence and then fade into an evolutionary tree branch, or we can figure out what a successful merge looks like<\/em>.\u201d<\/p>\n

That was effectively the mission a company called Bootstrap Bio touted openly. \u201cOf particular interest to me is whether we could modify intelligence<\/em>,\u201d co-founders Chase Denecke and Ben Korpan noted in a viral 2023 essay.<\/p>\n

Like Benson-Tilsen, they framed the concept as an imperative, noting there\u2019s currently a \u201cvery limited number of people<\/em>\u201d who have the intellect required to protect the world from a dangerously capable AGI. \u201cIt is not an exaggeration to say that the lives of literally everyone depend on whether a few hundred engineers and mathematicians can figure out how to control the machines built by the mad scientists in the office next door<\/em>,\u201d Denecke and Korpan wrote<\/a>.<\/p>\n

Author: Abby Vesoulis<\/em><\/p>\n

 <\/p>\n

yogaesoteric
\nMay 15, 2026<\/strong><\/p>\n

 <\/p>\n","protected":false},"excerpt":{"rendered":"

The new wave of Silicon Valley-backed gene-editing startups is straight out of Brave New World. Mathematician Tsvi Benson-Tilsen once worked at the Peter Thiel-funded Machine Intelligence Research Institute, where he was one of many experts tasked with figuring out how to ensure AI doesn\u2019t eventually destroy humankind. After seven years, he concluded that he\u2019s not […]<\/p>\n","protected":false},"author":5,"featured_media":233175,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_uf_show_specific_survey":0,"_uf_disable_surveys":false,"footnotes":""},"categories":[1620],"tags":[1516],"class_list":["post-233165","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-the-threat-of-artificial-intelligence","tag-article_of_the_week"],"_links":{"self":[{"href":"https:\/\/yogaesoteric.net\/en\/wp-json\/wp\/v2\/posts\/233165","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/yogaesoteric.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/yogaesoteric.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/yogaesoteric.net\/en\/wp-json\/wp\/v2\/users\/5"}],"replies":[{"embeddable":true,"href":"https:\/\/yogaesoteric.net\/en\/wp-json\/wp\/v2\/comments?post=233165"}],"version-history":[{"count":1,"href":"https:\/\/yogaesoteric.net\/en\/wp-json\/wp\/v2\/posts\/233165\/revisions"}],"predecessor-version":[{"id":233178,"href":"https:\/\/yogaesoteric.net\/en\/wp-json\/wp\/v2\/posts\/233165\/revisions\/233178"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/yogaesoteric.net\/en\/wp-json\/wp\/v2\/media\/233175"}],"wp:attachment":[{"href":"https:\/\/yogaesoteric.net\/en\/wp-json\/wp\/v2\/media?parent=233165"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/yogaesoteric.net\/en\/wp-json\/wp\/v2\/categories?post=233165"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/yogaesoteric.net\/en\/wp-json\/wp\/v2\/tags?post=233165"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}