Posts Note Bayesian Win in AI Debate Goes Unnoticed
Researchers on social media highlight separate Bayesian groups and an overlooked success.
Pedro Ortega posted that a respected source told him the Bayesian view won the AI debate yet almost no one noticed. Separate posts share a diagram titled The Laplacian Ecumene that divides Bayesian statistics into groups including Helmholtzians, de Finettians, Jaynesians, Yudkowskians, Pearlites, Gelmanians, and Fristonians. One reply adds the cult of Bayesian cognitive scientists distinct from active inference researchers. Thomas Dietterich replied that LLMs prevailed through maximum likelihood pre-training while Bayesian methods lost ground on uncertainty and anomaly detection.
Combined views
10.6K
7 posts, first seen 5d ago
Reactions from ranked influencers
7 posts
are you East Bayesian or West Bayesian. because the bridge is a long way, after all
It's kind of wild that there are not one but two cults built around Bayesian inference and they are almost entirely disjoint.
Someone I respect a lot recently told me: "In the end, the Bayesian view won the AI debate, but almost no one seems to have noticed." 🙂
It's kind of wild that there are not one but two cults built around Bayesian inference and they are almost entirely disjoint.
@pfau There's also the cult of Bayesian cognitive scientists -- not to be confused with the active inference / predictive processing cult.
It's kind of wild that there are not one but two cults built around Bayesian inference and they are almost entirely disjoint.
It's kind of wild that there are not one but two cults built around Bayesian inference and they are almost entirely disjoint.
@CsabaSzepesvari But I respect dozens :)