New Textbook 'AI for Science Students' Argues Bigger Models Need More Data

bravo_abad · x · 2026-09-09

Jorge Bravo Abad announced his new textbook Inteligencia artificial para estudiantes de ciencias (Ediciones Pirámide, 2026), aimed at science students, now available on Amazon.es (€27-31).

One core argument: in scientific problems, data scarcity is the norm. In that regime, model capacity is only useful if your data can support it — increasing complexity often buys you variance before it buys you performance. Classical ML or shallow networks may reach their useful regime quickly, while deeper models need far more data before extra capacity becomes an advantage.

Related event: Bigger Models Can Lose When Data Is Scarce: Learning Curves Matter Most(2 posts)→

Original post →

More from Research

Research channel →