Library News Shelves

Why language models hallucinate
OpenAI
September 5, 2025
OpenAI explains that language models hallucinate because they’re trained to produce confident, fluent answers even when unsure, rewarding style over accuracy. The company is developing methods to teach models to recognize uncertainty and ground their outputs in verified sources to reduce false information.

Sora 2 is here
OpenAI
September 30, 2025
Sora 2 is OpenAI’s next-gen video + audio generation model, offering more physically accurate, controllable, and realistic outputs (including synchronized dialogue and sound) as part of the new Sora app. It also introduces “cameos,” letting users upload short clips of themselves (or others, with permission) to be seamlessly inserted into generated scenes, while including built-in moderation and safety controls.

This CEO laid off nearly 80% of his staff because they refused to adopt AI fast enough. 2 years later, he says he’d do it again
Fortune
August 17, 2025
IgniteTech’s CEO defended his decision to lay off 80% of employees who resisted using AI, arguing that the cuts were necessary to modernize operations and eliminate internal sabotage against automation. He claimed the drastic move ultimately boosted productivity and affirmed his belief that companies must fully embrace AI or be left behind.
