BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.com//pycones-2026//talk//MHXF3P
BEGIN:VTIMEZONE
TZID:Europe/Madrid
BEGIN:STANDARD
DTSTART:20251108T000000
TZNAME:CET
TZOFFSETFROM:+0100
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20260329T030000
RDATE:20270328T030000
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20261025T030000
RDATE:20271031T030000
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
SUMMARY:Questioning the Mathematical Foundations of AI - Dr Ana Rojo-Echeb
 urúa
DTSTART;TZID=Europe/Madrid:20261108T151000
DTEND;TZID=Europe/Madrid:20261108T155000
DTSTAMP:20260726T231306Z
UID:pretalx-pycones-2026-MHXF3P@pretalx.com
DESCRIPTION:Since neural networks\, and especially transformers\, became d
 ominant\, production AI has largely settled on a small set of mathematical
  assumptions and rarely revisits them. The research literature has questio
 ned these assumptions for years\, but the alternatives have not crossed ov
 er into the systems most of us actually build\, so in practice we keep pus
 hing the same architectures further rather than asking a more fundamental 
 question: are we even using the right mathematical approach in the first p
 lace?\n\nThis talk focuses on one place where that question bites hardest:
  how we represent meaning and retrieve it. Many modern AI systems built in
  Python lean heavily on embeddings and similarity metrics such as cosine s
 imilarity\, particularly in components like hybrid search. The usual setup
  treats representations as points in a flat\, Euclidean space and measures
  relatedness by angle. That choice is convenient\, but it is not neutral\,
  and it quietly limits what these systems can capture.\n\nWe will look at 
 how embeddings and similarity metrics are used in practice\, what inductiv
 e biases they introduce\, and where they start to break down.\n\nFrom ther
 e we will introduce alternative perspectives such as hyperbolic spaces and
  manifold-based representations that capture hierarchy and structure with 
 far less distortion than flat space. A third lens comes from the theory of
  moving frames.\n\nWe will keep all of this intuitive\, with a simple illu
 strative example in Python.\n\nTake-home message: the mathematical choices
  behind AI systems are not neutral. If we question the geometry and struct
 ure we assume\, we can better understand the limitations of current approa
 ches and open the door to more suitable ways of representing meaning and r
 elationships.\n\nTarget audience: Python developers\, data scientists\, an
 d engineers working with machine learning or AI systems who want to unders
 tand the foundations behind embeddings\, similarity\, and retrieval.
LOCATION:Track 02 - Nagarro
URL:https://pretalx.com/pycones-2026/talk/MHXF3P/
END:VEVENT
END:VCALENDAR