The Weight of Knowing: AI, Genetic Prediction, and What We Choose to Become
Imagine you have a doctor's appointment to find out your first child’s gender—whether it will be a boy or a girl—you’re excited and nervous. You wait for the results. "It's a girl!" the doctor informs you, and tears of joy leave your eyes, until you hear the change in the doctor’s voice; the smile drops slowly, bracing for what comes your way, and suddenly you hear something that feels like a mirage in your ears: “Your baby girl will be disabled. She has Spina Bifida”— a condition that very likely means she won’t be able to walk. You have to choose: a baby girl you dreamed of, or saving your daughter's life from misery that hasn't even started yet, leaving them to stare at the ultrasound machine as the slow echo of the baby’s heartbeat fills the room.
The advancement of genetic tests not only revealed the gender but also brought another piece of news the family wasn’t expecting, the kind of dilemma Scientific American has called under-appreciated even as prenatal screening becomes routine. This information right there forces the parents to stand at the precipice of choices they never imagined making. History not only advanced towards a greater good, but it also brought the consequences of those advancements. From just checking the baby’s heartbeat to see how healthy it is to actually being able to tell its gender and locate disabilities at an early onset through amniocentesis, or non-invasive prenatal testing (Zaami et al., 2021), science made it all possible, but decided not to stop yet.
Now imagine you go to the hospital, and instead of telling the doctor, “We are pregnant,” you say, “We want a baby.” Crazy until it’s not. Science has advanced despite those unmade decisions and has presented AI-based polygenic testing (Raz et al., 2025). Not only can you decide your baby's gender, but you can also screen for chromosomal disorders, a routine part of embryo testing for years, and now use AI-driven polygenic risk scoring to predict your future baby's traits. Design it however you like, though predictive accuracy for these trait scores is still low and not considered reliable for real-world use (Polyakov et al., 2022).
“The results are foolproof; it's science.” “This new genetic technology is our future!" We see the headlines, but it’s their present, right now. The results of one genetic test declared one pregnancy paralysed. Her lower body? Not functional; she'll be born, no complications at birth, they said. But her life after that? Complicated. Whereas the other daughter, pre-conceived, has everything figured out before even coming into existence. AI predicts our whole life before we even get to live it. Knowing and choosing used to be separated by time; now AI collapses them into the same moment.
So how does any of this actually happen? A decade ago, understanding a single protein's structure—what it looks like and how it functions inside the body—required years of painstaking lab work using X-ray crystallography, one protein at a time. Now, Artificial Intelligence has changed that timeline entirely. Using tools like DeepMind's AlphaFold, scientists can predict a protein's 3D structure directly from its genetic sequence alone, and use that prediction to estimate its likely function in the body, all without ever stepping into a lab (Jumper et al., 2021). What used to take years now takes a computation. And that shift matters beyond a single protein: the same pattern-recognition ability that lets AI read one gene's structure also lets it read patterns across an entire genome at once—the exact leap that turns single-disorder testing into the kind of whole-embryo, whole-genome prediction sitting behind the scan in that hospital room. The second parents above weren't handed a guess. They were handed a rendering, built the same way, of a shape that hadn't been born yet.
Now, we have two schools of thought following both our families—or, as I like to call them, camps.
The first is the Mercy Camp. Bioethicists call this principle procreative beneficence, first named by philosopher Savulescu, J. (2001): the idea that if we can know, and that knowledge can reduce suffering, medicine has always had reason to act on it. AI extends this logic further than genetics ever could alone, mapping patterns across thousands of cases to flag not just risk, but likely outcome, giving doctors and parents something closer to a preview than a guess. What parent doesn't want the best for their kid? If understanding a disorder can spare a child suffering, this camp argues understanding a trait—height, intelligence, temperament—isn't so different: it's the same hope for a good life, just reached a little earlier than tutoring or nutrition ever could. In a case like the second family, this camp would argue the screening isn't designing a child; it's handing parents the clearest possible picture of which possibility to choose.
The second is the Legibility Camp, grounded in the disability rights and expressivist critique of prenatal technology (Zaami et al., 2021; Scully, 2023). This camp isn't against AI or against healthcare innovation. It's naming a cost: that turning a body into a forecast can quietly narrow what counts as a "normal" life before that life has even started. A diagnosis stops being something a doctor explains to a person, and starts being something a model assigns to a probability. The second family raises a sharper version of the same problem. Once a model can flag not just disorder but desirable traits—height, intelligence, temperament—someone decides what "desirable" means, and this camp calls that decision what it actually is: eugenics, no longer state-mandated, just privatized and offered as a choice (Rahim, 2024). Nor is it equal: screening like this isn't free, so only some families get to choose, sorting the next generation by budget as much as by nature. So we have to ask: once a model can predict a trait, does it also predict a person's worth?
Look past these two families, and the field itself keeps moving, faster than either of them had time to notice. Each year brings a new headline, a new capability, a breakthrough folded into the ones before it. Right now, somewhere, a model is learning to read a single letter of DNA and predict what it will do, a resolution of sight that didn't exist a few years ago—DeepMind's AlphaGenome, built on the same lineage as AlphaFold (Avsec et al., 2026). Embryo screening made MIT Technology Review's list of 2026's ten breakthrough technologies, and it won't stay imprecise for long.
Somewhere in that speed, something quieter is also happening. The more a machine can tell us in advance, the less room is left for a life to simply arrive, unrehearsed, the way lives always used to. That not-knowing was never just a gap in our data; it was part of what it meant to be human, to love someone before you had proof they'd turn out a certain way. AI can map a genome. It cannot sit in a waiting room, cannot hold the weight of a decision, cannot feel the years after. It was built to assist that moment, not to stand in for it, and how much of that moment we hand over is still, entirely, our own choice to make.
Even as prediction gets sharper and traits get easier to cherry-pick before a child exists, being human, uncertain, complicated, still worth defending on its own terms, has to stay the priority, even as the technology around it keeps moving faster than we do. Nature has always run on balance, not certainty, and maybe that's worth protecting on purpose, even now that we finally can.
But the world moving on doesn't mean the moment moves with it. Because beneath all of it, the parents from earlier are still standing in that hospital room. AI didn't take away their decision, but it did hand it to them earlier, faster, and with more certainty than any generation before had to carry. The decision in front of them, untouched.
Sources
Rahim, H. (2024, March 11). Designer babies? the ethical and regulatory implications of polygenic embryo screening - Petrie-Flom Center. https://petrieflom.law.harvard.edu/. https://petrieflom.law.harvard.edu/2024/03/11/designer-babies-the-ethical-and-regulatory-implications-of-polygenic-embryo-screening/
Manoj, S. (2026, February 2). How alphafold revolutionized protein structure prediction | BioInnovation Group at UC Davis. https://big.ucdavis.edu/blog/how-alphafold-revolutionised-protein-structure-prediction
Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., Tunyasuvunakool, K., Bates, R., Žídek, A., Potapenko, A., Bridgland, A., Meyer, C., Kohl, S. a. A., Ballard, A. J., Cowie, A., Romera-Paredes, B., Nikolov, S., Jain, R., Adler, J., . . . Hassabis, D. (2021). Highly accurate protein structure prediction with AlphaFold. Nature, 596(7873), 583–589. https://doi.org/10.1038/s41586-021-03819-2
Avsec, Ž., Latysheva, N., Cheng, J., Novati, G., Taylor, K. R., Ward, T., Bycroft, C., Nicolaisen, L., Arvaniti, E., Pan, J., Thomas, R., Dutordoir, V., Perino, M., De, S., Karollus, A., Gayoso, A., Sargeant, T., Mottram, A., Wong, L. H., . . . Kohli, P. (2026). Advancing regulatory variant effect prediction with AlphaGenome. Nature, 649(8099), 1206–1218. https://doi.org/10.1038/s41586-025-10014-0
Savulescu, J. (2001). Procreative Beneficence: Why we should select the best children. Bioethics, 15(5–6), 413–426. https://doi.org/10.1111/1467-8519.00251
Lupas, A. N., Pereira, J., Alva, V., Merino, F., Coles, M., & Hartmann, M. D. (2021). The breakthrough in protein structure prediction. Biochemical Journal, 478(10), 1885–1890. https://doi.org/10.1042/bcj20200963
Zaami, S., Orrico, A., Signore, F., Cavaliere, A. F., Mazzi, M., & Marinelli, E. (2021). Ethical, Legal and Social Issues (ELSI) Associated with Non-Invasive Prenatal Testing: Reflections on the Evolution of Prenatal Diagnosis and Procreative Choices. Genes, 12(2), 204. https://doi.org/10.3390/genes12020204
Navon, D. (2024, February 20). New prenatal genetic screens pose underappreciated ethical dilemmas. Scientific American. https://www.scientificamerican.com/article/new-prenatal-genetic-screens-pose-underappreciated-ethical-dilemmas/
Raz, A., Halsband, A., Langner, R., & Shkedi-Rafid, S. (2025). The new frontier in assisted reproduction. EMBO Reports, 27(2), 265–268. https://doi.org/10.1038/s44319-025-00668-2
Scully, J. L. (2023, February 28). Prenatal genetic testing and disability: The ethical minefield. Disability Innovation Institute. https://www.disabilityinnovation.unsw.edu.au/prenatal-genetic-testing-and-disability-ethical-minefield
Polyakov, A., Amor, D. J., Savulescu, J., Gyngell, C., Georgiou, E. X., Ross, V., Mizrachi, Y., & Rozen, G. (2022). Polygenic risk score for embryo selection—not ready for prime time. Human Reproduction, 37(10), 2229–2236. https://doi.org/10.1093/humrep/deac159
Black, J. (2026, February 11). Embryo scoring: 10 Breakthrough Technologies 2026. MIT Technology Review. https://www.technologyreview.com/2026/01/12/1130011/embryo-scoring-genetic-testing-2026-breakthrough-technology/
Podgursky, B., & Katz, M. (2026, June 29). Polygenic embryo screening and your family. Orchid. https://www.orchidhealth.com/guides/polygenic-embryo-screening-and-your-family
