{"type":"rich","version":"1.0","provider_name":"Transistor","provider_url":"https://transistor.fm","author_name":"The New Biology","title":"The Bitter Lesson for Biology — Adam Green on Virtual Cells and Scaling Laws","html":"","width":"100%","height":180,"duration":5374,"description":"Markov Biosciences, a startup in San Francisco, is betting that biology is about to have its GPT moment. In this episode, founder Adam Green explains the \"bitter lesson\" for biology, the idea borrowed from Richard Sutton that large unbiased datasets and the right training objective tend to outcompete models with hard-coded rules and human priors. Adam thinks, in particular, that the virtual cell field took a wrong turn by spending hundreds of millions of dollars collecting expensive perturbation data. Green’s counterargument is that the data needed to train useful virtual cells is not limiting, but rather compute (and the loss function) are. By treating single-cell RNA-seq as a ranking problem rather than raw counts (a century-old idea traceable to a 1927 psychophysics paper), they found that virtual cells pre-trained on plain observational data show clean scaling laws, getting monotonically better at predicting unseen perturbations as the models grow, and beating a state-of-the-art model built specifically for that task.\n\n00:00 - Cold open and introduction \n01:58 - The first clinical prediction from a virtual cell\n05:38 - What is a \"virtual cell,\" really? \n08:01 - Single-cell RNA-seq biases and the urns analogy\n23:29 - The bitter lesson for biology\n30:55 - Geometric Plackett-Luce: the right loss function\n59:26 Trop2 deep dive\n1:11:16 - Top-down vs. bottom-up biology, mechinterp, and control as the goal \n\nReadings and mentions: \nMarkus Covert — A Whole-Cell Computational Model Predicts Phenotype from Genotype\nMarkov's ADC-predictions thread (Adam Green)\nScannell et al. (2012), \"Diagnosing the decline in pharmaceutical R&D efficiency\" (Eroom's Law)\nAdam Green on the Bitter Lesson\nAdam Green on RNA-seq issues\nArc Institute — STATE model (Adduri et al., 2025)\nGPT-1: Radford et al. (2018), \"Improving Language Understanding by Generative Pre-Training\"\nRich Sutton, \"The Bitter Lesson\" (2019)\nYann LeCun's \"cake\" analogy (explainer)\nMarkov paper — Generative ranking /...","thumbnail_url":"https://img.transistorcdn.com/Cdlp_4Y2x0oxnTGtoVdklrgCMQQ-ARclU1VI-dDVZCI/rs:fill:0:0:1/w:400/h:400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS9jYTRi/NTg1Yzc3YmRiZjg1/ZjNmYmNkNjAyYjFl/NjllMS5qcGc.webp","thumbnail_width":300,"thumbnail_height":300}