International Students Ignored, AI Ignites
Another weird week. U.S. higher-ed institutions are under growing political and social pressure. International students are facing deportations and are left without strong advocates—despite being crucial to both campus enrollment and the teaching and research missions of faculty. Yet as these challenges play out, the field of AI forges ahead, delivering breakthroughs that promise to transform education and beyond.
OpenAI launched two new reasoning models—o3 and o4‑mini—that extend its “chain-of-thought” approach across coding, mathematics, and visual analysis, marking its most advanced AI release to date (The Verge). The flagship o3 model devotes extra computation to deliberate on complex prompts and, for the first time, can “think with images,” using tools to crop, zoom, and rotate visual inputs during its internal reasoning (The Verge). Its smaller counterpart, o4‑mini, is optimized for speed and cost efficiency, achieving top benchmark scores (e.g., on the AIME exam) while remaining compact enough for broad deployment.
Google is an online advertising monopoly, judge rules
A federal judge in the Eastern District of Virginia found that Google has “illegally built ‘monopoly power’” in its web advertising business by tying together its publisher ad server and ad exchange, harming competition among publishers and advertisers. This marks the second major antitrust victory against Google in under a year—following last December’s ruling on its search engine monopoly—and the third since late 2023. While the court did not deem Google’s prior acquisitions of DoubleClick and Admeld unlawful per se, it indicated that breaking up parts of Google’s ad tech stack, such as divesting Ad Manager, may be necessary to restore competition. Google has announced plans to appeal. Read here.
OpenAI is reportedly working on a social network to rival X
Insider sources report that OpenAI is quietly developing a prototype of its own social media platform—akin to X—built around ChatGPT’s generative-image capabilities. The internal prototype features a feed where users can generate and share AI‑created visuals, though it remains unclear whether OpenAI will launch this as a standalone app or integrate it directly into ChatGPT. CEO Sam Altman has solicited external feedback, and the move is seen as a bid to secure a fresh stream of human‑generated data for training models—while also positioning OpenAI to compete directly with Elon Musk’s X and Meta’s forthcoming AI‑driven social features. Read here.
Google is now leading on multiple fronts on AI metrics
This chart showcases Google’s top rankings across various performance and efficiency benchmarks in artificial intelligence. It underscores Google’s competitive edge in both foundational model quality and compute‑cost efficiency relative to other major AI players.
Other:
As someone who thinks about misinformation and its impact on the society, I found the diagram below helpful.
This diagram uses a ladder to show how each step—from statement to proof—builds on the one below it, yet warns that no single rung guarantees validity: statements may be inaccurate, facts unrepresentative, data inconclusive, and evidence not universally applicable. It urges us to interrogate conclusions by checking context and range, considering rival explanations, looking at the full picture (including failures and omitted successes), and scrutinizing what was actually said or measured.
