(3:50 Spark Session, Heidelberg Laureate Forum)
new AI trained by current AI
es: GPT4 data for training GPT5
AI self-improvement (and why math is well suited for it).
open-source automated formal proof generation
large language model (LLM), state-of-the-art (SOTA) performance
A conversation with Terry Tao interessante
rob: non so se questa startup sopravvivera'.
yt t=391 Emily Riehl — The future of mathematics | Math, Inc.
I mean formalization has completely changed my view of what mathematics is. ...
principalmente dovuta a usare come fondamenti della matematica la teoria dei tipi dipendenti piuttosto che la teoria degli insiemi.
leanprover.zulipchat.PrimeNumberTheorem/Fejér's theorem
Given xy² = y³x and yx² = x³y, derive x y^{2n} x⁻¹ = y^{3n}.
x y² x⁻¹ = y³ moltiplicando membro a membro
xy²x⁻¹ xy²x⁻¹ = y³ y³ sviluppando
x y⁴ x⁻¹ = y⁶ ripetendo il procedimanto
x y⁶ x⁻¹ = y⁹ yx⁶y⁻¹ = x⁹
For n=3: y⁹ = e.
Matching orders of y² ~ y³ ⇒ y = e.
Then x³ = x² ⇒ x = e.Group is trivial.
The paper - https://arxiv.org/pdf/2508.03685
You can check from The reasoning traces summary : -
https://chatgpt.com/.../68bc7446-ca74-8007-9646-955c4e7e6daa