Tag
Briefings and ideas about Training data from AI podcast discussions and the research they cite, with links to the original sources.
On Moonshots, Emad Mostaque said AI could be aligned from inception by designing training data like a child's curriculum, and Salim Ismail welcomed the idea.
Updated Oct 17, 2026 IdeaThe AI monoculture argument holds that models converging on similar answers and styles may share blind spots.
Updated Sep 29, 2026 IdeaPublishers could be paid each source's estimated Shapley value for AI agent use: roughly the cost of matching answer quality without it.
Updated Aug 25, 2026Journalist Karen Hao argues that AI companies behave like empires. In her view, their warnings of catastrophe and promises of abundance work together to justify keeping AI under their control. She made the case in a replayed Diary of a CEO clip.
Modal CEO Erik Bernhardsson says coding agents trained on old examples keep writing outdated Modal syntax, so it retires features more cautiously. Modal also tests whether agents recommend it and clarifies pricing content they misread.
On Moonshots, a Musk clip urging hopeful AI stories set off a split: Dave Blundin and Chamath Palihapitiya argued for inspecting what labs build in, while Alex Wissner-Gross bet on competing AIs and human-machine merger.
Tenex estimates its AI token bill at one to two times its employee costs. Alex Lieberman says falling prices make runaway spending less of a worry than six to eight months ago, and make it harder for enterprises to justify fine-tuning.
Biohub's Alex Rives expects pretraining, elicitation and RLVR to carry over from language models to biology, but says verification must happen in the lab. He is unsure whether biology and language models will share one latent space.
Biohub science head Alex Rives argues that no lab animal can ever be a perfect stand-in for people. He says an AI model trained across many species, as protein models were trained across evolution, could predict human biology far better.
Biohub's head of science Alex Rives argues that cell models are held back by what biology can measure. His examples are whole-zebrafish imaging and a laser phase plate for electron microscopes, so far tested only on isolated proteins.
Biohub's head of science, Alex Rives, says virtual cell models fall short above all because biology can't yet measure what matters. He expects an information-level cell model within about five years.
Josh Gao, CEO of robot-data supplier Mecka AI, says labs train on about a million hours of human data, tens of thousands of gripper hours and a few teleop hours. He argues scaling that alone won't yield a robot that learns any task by watching.
Mecka CEO Josh Gao says robotics is still orders of magnitude short of Tesla's driving data. He expects constrained, goal-oriented jobs such as industrial cleaning to come first. He says homes are where the field must take the most care, and that data collected on site can close the reliability gap.
Mecka CEO Josh Gao described how his company makes robot training data, with camera caps, force-sensing gloves and checks against fraud. He said contributor pay matters less to its margin than verifiably high-quality data at scale.
Peter Diamandis cited a fall from $2.36 to $0.38 per robot subtask between July and September and projected near-zero cost within a year. RoboCurve's published test instead compares two models at $1.14 and $0.90 per attempt.