Context Window 12
This edition covers the arrival of DeepSeek and what it means strategically and economically, Apple’s iOS 18.3 changes pausing news summaries and switching Apple Intelligence to opt-out, the wider shift among big tech firms to make AI the default rather than the upgrade, the US Copyright Office’s second report on AI and copyrightability, the UK Government’s AI and Copyright consultation alongside a Government tool that uses LLMs to analyse consultation responses, a mini-literature review on generative AI in pedagogy, and details of the AI-heavy IPG Spring Conference.
It’s been another significant week for AI developments: unless you’ve been offline all week you will have seen news coverage of DeepSeek, a Chinese LLM with impressive performance and efficiency gains versus established, largely US models. There’s a whole series of issues here: technological, economic and political. Amid market turmoil, the investor Marc Andreessen framed it in historical terms as a Sputnik moment. (I wonder if there are also parallels with the competition established US auto manufacturers faced from Japanese carmakers with a disruptive business model and consumer proposition post-1973). In other news, Apple is rolling out two significant AI updates with iOS 18.3. It is temporarily removing support for AI summarisation of news headlines, following issues with inaccurate and misleading outputs that I wrote about earlier in the month. So at least its takes on DeepSeek will be more accurate. More significantly for a broad range of users, it is switching Apple Intelligence from opt-in to an opt-out (if you don’t want it on your iPhone, there’s a simple guide to switching it off here). Like the Google Workspace/Gemini integration that I wrote about last week, and recent changes to Microsoft Copilot, this speaks to a change of approach for big technology companies, making AI the standard, not the upgrade (and in many cases turning it off is a great deal harder than on an iOS device). That’s a paradigm that organisational policies and practices need to address. The US Copyright Office published its second report on AI and copyright, focusing on questions of authorship and copyrightability. Key conclusions include that copyright applies to work created by a human author even if that work also includes AI-generated material, so long as there is sufficient human control over the creative process. However, authorship must be assessed on a case-by-case basis, and the USCO concludes that prompts alone do not represent sufficient control (reading their analysis from page 18 onwards as a practitioner rather than a lawyer, it would seem difficult to meet their suggested standard of sufficiency with current AI tools and practices). If you are intending to assert copyright in a particular piece of content, this view, and any parallel developments in the UK and other jurisdictions, are highly significant. Of course, there are also many uses for AI in publishing where there is not such a pressing need to copyright outputs. Like many people in publishing, I’ve been formulating a response to the UK Government’s Consultation on AI and Copyright. With many leading industry and advocacy groups encouraging contributions, it seems likely there will be a big response. So I was interested to read last week that one of the projects developed by the Government’s Incubator for Artificial Intelligence is a tool to use LLMs to analyse the results of large public consultations. There is a lot that I like about the approach: open source code, preserving the original responses alongside the summarised analysis, publicly stated governance and assurance principles, and human-in-the-loop. But that made me wonder: would the responses to the AI and Copyright Consultation be analysed with an AI tool? Surely not. But if so, there would be some follow-up questions… For educational publishers on the list, this post from Dr Philippa Hardman is a useful mini-literature review of five recent studies on generative AI and pedagogy: given the ranges in focus and methodological approach, the papers identify common themes around pedagogical structure and the importance of encouraging critical reflection. Finally, bookings opened this week for the Independent Publishers Guild Spring Conference in London on 25-26 February. There is a strong AI theme: my colleague Rob Burleigh will be demonstrating Burleigh Dodds’s AI-powered AgNetZero platform, and I will be taking part in a hybrid panel with Will Crooks of Publishing Licensing Services and Ed Newton-Rex of Fairly Trained. If you’re attending, I’d love to say hello in person.
This was originally published in my email newsletter. To receive weekly updates on how AI is affecting the publishing industry, sign up here.