Context Window 7
This edition covers the UK government’s consultation on copyright and AI, an Amazon job listing that hints at deeper AI use inside the Kindle reading experience, NotebookLM’s enterprise update, Apple Intelligence’s BBC misattribution, a US AI litigation tracker, and what the Betty Crocker Effect can teach us about generative AI.
The big news in the UK was the announcement of the UK government’s consultation on copyright and artificial intelligence. Trade organisations such as the PA and ALCS have put out public statements highlighting the lack of evidence for a change to the copyright regime, and highlighting the proposal that rightsholders opt out, rather than in, to training. If your organisation wants to respond, you have until 25 February. Amazon have been closely involved in AI through their AWS platform, investment in Claude and use of AI in search. But their plans for AI in books have been less clear. This week a job listing offers some insight into their intentions. Amazon are recruiting for a data scientist, responsible for “unlocking capabilities like analysis, enhancement, curation, moderation, translation, transformation and generation in books based on content structure, features, intent, synthesis and publisher details… to enhance the reading and publishing experience.” That’s a pretty expansive charter, and what’s particularly interesting is how many of the points refer to book content, rather than just the retail experience around the book. It sounds like the Kindle UX is going to move forward substantially. What role there is for publishers in that development remains to be seen. Google pushed an update to its NotebookLM tool, aimed at larger users such as enterprise customers, schools and universities and including increased privacy and data security. The key feature of NotebookLM is its ability to generate podcast-style audio summaries of documents and data. As an illustration, I put in the text of this week’s newsletter and in just over a minute, it generated a seven minute podcast-style discussion. Is it great listening? I think not, though admittedly I didn’t put any particular effort into crafting my prompts as I wanted to see what the minimal version sounded like. You can make your own mind up here. But is this the worst version of the tool you’re going to use? And are consumers going to be putting your content into tools like this to create a version that’s convenient to consume? Absolutely. The AI feature on newer iPhones has come under scrutiny after an AI-generated summary of notifications from the BBC falsely suggested the suspect in the United Healthcare CEO murder had shot himself. (FWIW, Apple Intelligence rolled out here last week, and while I haven’t seen any factual inaccuracies, I haven’t really seen any benefit from it either.) The summaries of AI licensing deals were one of the most popular links on last week’s newsletter, so in a similar vein, Wired has published a useful summary—and data visualisation—of ongoing US litigation over AI, mapping who is suing whom. I really liked this piece on what we can learn about AI from consumer psychology, specifically the Betty Crocker Effect (the possibly apocryphal story is that in the 1950s General Mills increased the popularity of powdered cake mix by making the consumer do just a little more). Similarly with generative AI, reduced effort often means a perception of reduced quality, so perhaps the answer is to reframe the effort that goes into it. Thank you to everyone who has subscribed to and shared the newsletter this year. I’d love to know your thoughts on what you’d like to see more or less of in 2025—please feel free to get in touch with any comments. Meanwhile, I’m going to take a seasonal pause next week, with normal service resumed on Friday 3 January. Happy holidays to all of you who celebrate, and best wishes for the New Year.
This was originally published in my email newsletter. To receive weekly updates on how AI is affecting the publishing industry, sign up here.