Genetics

Open the episode, then seek to 0:00.

Hsu recalls an encounter with Richard Feynman and the personal advice he says Feynman offered.

Scientific careers

Open the episode, then seek to 5:00.

Hsu continues his recollections and discusses changes he observed in weight training and athletic culture.

Scientific careers

Open the episode, then seek to 10:00.

The speakers compare historical images of physique with modern athletic training; Hsu offers his interpretation of changing physical culture.

Scientific careers

Open the episode, then seek to 11:47.

Hsu introduces genomic prediction as a statistical problem and describes his company’s proposed application to choosing among IVF embryos.

Polygenic scores

Research note: A successful predictor has a defined outcome and test population →

Open the episode, then seek to 16:17.

Hsu describes his move from physics into genomics, compressed sensing and his group’s claims about the sample sizes needed for prediction.

Polygenic scores

Open the episode, then seek to 21:04.

Hsu explains additive predictors, correlated variants and sparse optimisation; Patel asks why selection has not maximised the traits discussed.

Polygenic scores

Open the episode, then seek to 25:01.

Hsu discusses evolutionary arguments for additive variation and argues that changing environments complicate ideas of biological optimisation.

Natural selection

Open the episode, then seek to 29:58.

The speakers use software analogies to discuss genetic architecture; Hsu argues that many disease predictors involve relatively distinct sets of variants.

Evolutionary trade-offs

Open the episode, then seek to 34:55.

Patel asks about a health-index result; Hsu discusses available genetic variation and extrapolates from selection and breeding to potential human interventions.

Embryo selection

Research note: Four estimated years are not four observed extra years of life →

Open the episode, then seek to 39:50.

Hsu forecasts future editing capabilities and discusses possible longevity gains, while acknowledging uncertainty about ultimate biological limits.

Genetic engineering

Open the episode, then seek to 43:20.

Hsu discusses clinic relationships, data access, patents and the uncertain durability of his company’s scientific advantage.

Genomic prediction business

Open the episode, then seek to 46:25.

Hsu describes embryo-sample processing and possible cloud services, then forecasts facial prediction while saying his company does not offer that service.

Genomic prediction business

Open the episode, then seek to 51:05.

The speakers discuss genetic identification and privacy; Hsu distinguishes disease risk from diagnosis and speculates about future longevity and inequality.

Genetic privacy

Open the episode, then seek to 54:50.

The speakers discuss variation among possible offspring; Hsu speculates about genetic compatibility and possible reproductive applications.

Embryo selection

Open the episode, then seek to 58:24.

Hsu acknowledges environmental effects on height and reduced prediction quality across ancestry groups, discussing possible explanations and future data needs.

Polygenic scores

Research note: A score can lose accuracy when its training population changes →

Open the episode, then seek to 1:03:20.

The speakers consider distinguishing causal variants from statistical tags and discuss uncertainties and controversies around interpreting population differences.

Polygenic scores

Open the episode, then seek to 1:06:45.

The speakers speculate about regulation and access, then discuss historical and political attitudes toward reproductive technologies.

Embryo selection

Open the episode, then seek to 1:11:18.

Hsu argues that better cognitive measurements could improve predictors and offers his interpretation of obstacles to collecting those data.

Polygenic scores

Open the episode, then seek to 1:15:22.

Hsu calls proposed links between inherited disorders and cognitive traits speculative, then argues that future genetic interventions might avoid some trade-offs.

Evolutionary trade-offs

Open the episode, then seek to 1:18:53.

Hsu distinguishes educational attainment from cognitive test performance and describes differences in predictive power within sibling pairs.

Polygenic scores

Research note: Sibling comparisons address specific sources of bias →

Open the episode, then seek to 1:23:25.

The speakers discuss scientific communication and possible consumer uptake; Hsu distinguishes the IVF market from broader speculation about human evolution.

Genomic prediction business

Open the episode, then seek to 1:28:17.

Hsu discusses adoption within IVF and praises independent scholarship, while considering the opportunities and constraints of cross-disciplinary work.

Scientific careers

Open the episode, then seek to 1:33:02.

Hsu discusses scientific specialisation, then considers how predictor quality and the number of available embryos might affect selection.

Embryo selection

Open the episode, then seek to 1:36:43.

Hsu discusses pluripotent-cell gametes, editing and iterated selection as possible future technologies, while acknowledging adoption concerns.

Genetic engineering

Research note: Producing functional mouse eggs is not human clinical validation →

Open the episode, then seek to 1:41:41.

The speakers discuss possible inequality from reproductive technologies; Hsu acknowledges the concern and proposes broader public access.

Embryo selection

Open the episode, then seek to 1:44:48.

Hsu and Patel discuss different approaches to programming and technical problems, including visual, spatial and verbal reasoning.

Cognitive abilities

Open the episode, then seek to 1:49:48.

Hsu discusses visual reasoning and explains why derivatives pricing and data analysis have attracted people trained in physics.

Scientific careers

Research note: Mathematical similarity does not make a market a physical law →

Open the episode, then seek to 1:53:05.

Hsu discusses experience with noisy experimental data; Patel compares it with debugging, before they turn to physicists’ career paths.

Scientific careers

Open the episode, then seek to 1:58:02.

The speakers discuss transitions out of physics and use stories about Jeff Bezos to consider different combinations of abilities.

Scientific careers

Open the episode, then seek to 2:03:02.

Hsu argues that cognitive abilities are multidimensional and discusses technical problem-solving and founders’ communication skills.

Cognitive abilities