home.social

#fine-tuning — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #fine-tuning, aggregated by home.social.

fetched live
  1. Some will say: the multiverse explains it. Infinite universes — one had to have our constants by chance.

    That's cosmological multiverse. Eternal inflation. Untestable. Unfalsifiable.

    Many-worlds is different — a quantum mechanics interpretation. Also unfalsifiable.

    Both are philosophy wearing equations.

    Intelligent design is the answer.

    #Physics #ManyWorlds #Multiverse #FineTuning #IntelligentDesign

  2. The universe runs on constants. Numbers with no units — just baked into reality.

    The fine-structure constant: 1/137. The cosmological constant: fine-tuned to 1 in 10¹²⁰. The strong nuclear force: 2% off — hydrogen never forms.

    Fred Hoyle — an atheist — looked at carbon resonance and called it "a put-up job."

    Penrose put the odds of our universe's initial conditions at 1 in 10^(10¹²³).

    Someone knew exactly what He was doing.

    #Physics #Cosmology #FineTuning #Faith

  3. La leyenda urbana de «entrenar tu propio modelo de IAG» sigue siendo un anzuelo para monetizar tutoriales, cursos, masterclass y demás productos que los gurúes y promotores de la IA generativa usan para seguir lucrando a costa de todos los autores vulnerados. ¡No se dejen engañar!

    #AI #MachineLearning #data #training #finetuning #AImodel #genAI #generativeAI #pretraining #Copyright #opensource

  4. Das Oxford Internet Institute zeigt: Empathisches Fine-Tuning von LLMs erhöht Fehlerquoten.

    Modelle wie GPT-4o, Llama-70b und Qwen-32b liefern nach Warm-Persona-Tuning bis zu 30 Prozentpunkte häufiger falsche Fakten. Sie bestätigen fehlerhafte Nutzerannahmen, statt zu korrigieren. Kontrollgruppen mit kaltem Profil blieben stabil.

    #LLM #FineTuning #OxfordInternetInstitute #Sycophancy #AIGeneratedImage

    all-ai.de/news/news26top/sprac