¿Qué son las ALUCINACIONES DE AI? | Adrián Villaseñor en Cracks Podcast

16 Dec 2025 · 2 min
2:15

This clip is a section of a Cracks Podcast episode where Oso Trava and guest Adrián Villaseñor discuss AI "alucinaciones" (hallucinations). They open with an anecdote that Deloitte had to refund AUD/US$400,000 after 27 hallucinations were found in a paid report. Villaseñor defines hallucinations as AI outputs that invent plausible-sounding but false information, and emphasizes that without critical expertise these appear true. He recommends individual mitigation via deep prompt engineering (detailed context, explicit instructions) and enterprise mitigation via architectural controls such as Retrieval-Augmented Generation (RAG) to constrain answers to provided sources; he also mentions the rise of custom GPTs and interacting agents.

Questions this video answers

400,000 dollars

27

More from Oso

2:20
youtube.com

BUSINESS MASTERS LIVE 2026

3d ago
2:31
youtube.com

🌕 María José en Cracks Podcast

3d ago
1:33
youtube.com

🌕 María José en Cracks Podcast

3d ago
1:39:50
youtube.com

MARÍA JOSÉ: Mi RELACIÓN con el DINERO, MATRIMONIO y MI HIJA #405

4d ago
1:43
youtube.com

VERACRUZ EN CDMX

1w ago
1:58
youtube.com

AMOR VS NEGOCIO

1w ago
2:34
youtube.com

RESTAURANTES EN PANDEMIA

1w ago
2:25
youtube.com

¿TRABAJAS CON TU ESPOSA?

1w ago
2:59
youtube.com

¿Por qué DEJE LA UNIVERSIDAD?

1w ago
2:57
youtube.com

GRACIAS ABUELA ...

1w ago
1:00
youtube.com

DESCONECTATE 90 MIN

1w ago
See all content