Wrong way

I've got one or two documents I'd like to pop into Matt Brittin's weekend reading bag. 

We all knew that Matt was a gospeller for AI, but extracts from his first address to BBC staff have been heard by Jake Kanter of Deadline... 

"He said the BBC could deploy technology to analyze its news and content to establish patterns in output. Brittin said this could mean assessing how often the BBC uses certain words, or analysing the types of contributors appearing across its programming.

“Stories and data together are the way to understand the world..... not to audit people, but as a kind of sat nav around bias or sat nav around these topics … So that’s where I think I’d try to complement our brilliant expert teams.”

So first in his man-bag, The Asserson Report, of 2024, commissioned and shaped by the head of a Tel Aviv HQ'd law firm, and produced with Dr. Haran Shani-Narkiss, who calls himself a "Computational Neuroscientist". 

Second, the Centre for Media Monitoring report of 2025 "We employed Claude 3.5 Sonnet, a state-of-the-art language model that has demonstrated strong performance on the MMLU-Pro benchmark designed to evaluate the multitask capabilities of language models across diverse subjects, achieving 88.3% accuracy."

Both have lots of data, but the textual analysis has parameters set by humans, and both use so called sympathy assessments to come to what many would call their preferred conclusions. This is a dangerous and distracting path, Matt.