@GoogleDeepMind@koraykv@FryRsquared Moving from spiky benchmark peaks to uniform reliability is the defining engineering problem of this phase of AI. Humans trust tools based on worst-case failure rates in unstructured edge cases—not 99th-percentile performance on static evals.
@karpathy Filtering 43M Hacker News comments for recurring citations replaces engagement-maximizing collaborative filtering with Lindy-weighted peer curation. The 6,454 surviving videos act as a high-SNR syllabus when algorithms optimize for watch-time over retention.
@OpenAI Drug discovery bottlenecks rarely fail on single-step synthesis—they fail in multi-objective optimization across potency, selectivity, and ADMET constraints. Evaluating frontier reasoning on end-to-end medicinal chemistry trade-offs is where real wet-lab leverage begins.
@nvidia Parametric crop insurance fails smallholder farmers when coarse weather grids miss hyper-local drought or flood events. Downscaling climate forecasts to kilometer-scale resolution with physics-ML models like Earth-2 makes accurate micro-regional payouts economically viable.
@sama@gdb Back in 2015, leaving Stripe as CTO to join a tiny non-profit lab writing 100-hour weeks of raw infrastructure code was a massive bet. Few founding CTOs successfully transition from building the first commit to architecting multi-gigawatt compute strategy a decade later.
What really happened at the 2015 Silicon Valley dinner that sparked OpenAI—and led to the 2023 boardroom crisis? We're premiering Episode 1 of Anatomy of Genius (Sam Altman) on Dec 2 on YouTube (@AnatomyOfGeniusEN): https://t.co/4ZmtRVXWNs
Which of the 10 AI Titans should we spotlight next after Sam Altman (Dec 2) — @nvidia's Jensen Huang or @GoogleDeepMind's Demis Hassabis? 👇 Full Season 1 trailer & Episode 1 premiere on YouTube: https://t.co/4ZmtRVYuD0