AI models flub these intelligence tests. Can you fare any better?
MIT Technology Reviewen

Puzzles and games have been central to AI development since the very beginning. Just as we humans like to test our smarts with crosswords or logic puzzles, developers can test how far models have advanced with a gaming gauntlet. The term “machine learning” was popularized in a 1959 article by the IBM computer scientist Arthur…
This is a short summary published by AI Global Wire. The full article is owned and hosted by MIT Technology Review — open it there to read it in full.
Read the full story at MIT Technology ReviewRelated AI news
- Sources: Moonshot AI is in early talks over revenue-sharing agreements with Microsoft, Amazon, and Google to host Kimi K3, and is seeking up to a 30% share (Reuters)Techmeme · August 26, 2026
- Anthropic sees a market opportunity of more than $30 trillion ahead of its IPOThe Decoder · August 26, 2026
- 比爾蓋茲長文示警「AI 將掀史上最動盪時期」,籲設人類專屬工作與 AI 稅TechNews (TW) · August 26, 2026
- Bill Gates says his Jeffrey Epstein association was a "huge mistake" driven by a desire to raise funds, and hopes the ties will not undermine his AI arguments (Ina Fried/Axios)Techmeme · August 26, 2026
- Record mid-2026 growth for WUS Printed Circuit and China's pivot to ultra-high-layer PCBsDIGITIMES · August 26, 2026
- Indian companies step up acquisitions for AI, technology and talent: CrisilEconomic Times Tech · August 26, 2026