Wikipedia Takes on Scrapers, o4-mini Fumbles, and MIT Makes Tiny AIs Code Better
Audio is streamed directly from the publisher (podcasts.captivate.fm) as published in their RSS feed. Play Podcasts does not host this file. Rights-holders can request removal through the copyright & takedown page.
Show Notes
On today’s AI Deep Dive, we spotlight Wikipedia’s latest initiative to provide structured data through Kaggle, aiming to support AI development and reduce harmful web scraping. We also dig into OpenAI’s new o3 and o4-mini models, which oddly show more hallucinations than earlier versions—raising questions about reliability. A real-world example from Cursor AI reveals the dangers of these hallucinations, while MIT researchers present a promising technique for improving code generation accuracy in smaller models. Tune in for a full breakdown of these critical AI advancements.