Journalism begins where hype ends

,,

We can only see a short distance ahead, but we can see plenty there that needs to be done.” "

—Alan Turing

OpenAI’s Model Astra Solves Ten Longstanding Mathematical Problems

OpenAI said the tokens needed to find the ten solutions would cost roughly $2,000 at its Sol API rates.

An internal version of OpenAI’s next major model produced solutions spanning high-dimensional geometry, coding theory, group theory, operator algebras, quantum complexity and lattice cryptography. None of the ten results has been independently verified.

Legal AI Lacks ‘Legibility,’ Stanford Researchers Call for Public Benchmarking

A Stanford study calls for independent public benchmarking of legal AI systems, arguing that greater transparency is needed to help legal professionals assess the performance, reliability

Beyond technical evaluations, the researchers argue that building trustworthy legal AI requires independent, community-driven benchmarking supported by the right institutions.

Meet PMX AI: Malaysia’s Prime Minister Anwar Ibrahim Debuts an AI Avatar

Malaysian Prime Minister Anwar Ibrahim's verified Facebook page features a PMX-themed cover image ahead of the planned launch of PMX AI, an AI avatar trained on his speeches, writings and government policies.

Developed by Zetrix AI in collaboration with Parti Keadilan Rakyat (PKR), PMX AI is trained on Prime Minister Anwar Ibrahim’s speeches, writings and government policies to help Malaysians access government services and answer questions in English and Malay.

UK NHS Rolls Out AI Triage Tool for Health App; To Prioritize Patients By Automation

UK NHS has announced roll out of an artificial intelligence triage tool in its NHS app to let automation prioritize patients.

The AI triage tool in NHS app is part of £10 billion funding received by NHS to digitalize healthcare system.

Google DeepMind Says Their AI System Helped Solve Nine ‘Open Erdős Problems’

Google DeepMind said its AI agent autonomously solved nine open problems associated with Hungarian mathematician Paul Erdős. (Image: DeepMind / Wikimedia Commons)

Google DeepMind’s AlphaProof Nexus agent autonomously solved nine open problems associated with Hungarian mathematician Paul Erdős — two of them open for 56 years. The announcement, days after OpenAI claimed to have disproved an Erdős conjecture from 1946, signals a shift in how AI labs are testing mathematical reasoning.