Context Window 45
This edition covers UK Technology Secretary Liz Kendall’s confirmation that no commitments were made to AI companies on copyright during Trump’s state visit, Amazon’s agentic upgrades to Seller Assistant, Google’s new Agents-to-Payments protocol, takeaways from an IPG Autumn Conference panel on optimising publisher websites for AI, new OpenAI research on how people are actually using ChatGPT, librarians being asked to find hallucinated books, and an essay arguing most organisations are in the business of translating work from one format to another.
A slightly shorter email than usual today, as it’s been a very busy week of travel. I had conversations about AI in some interesting places: a publishing conference, an Oxford college and the House of Lords. The latter two particularly felt a long way from Silicon Valley, but the discussions underlined how much the technology is permeating every aspect of our lives. International readers may have missed President Trump’s state visit to the UK this week, but domestically it’s been hard to ignore. There was a slew of major AI investments being announced, but for publishers the most consequential aspect of the week was confirmation from new Technology Secretary Liz Kendall that no commitments had been made to AI companies on UK copyright and AI training. A couple of major updates on AI and ecommerce. Amazon announced an upgrade to its Seller Assistant, which now includes AI agents for tasks like inventory management. Publishers or retailers using FBA can use the AI to review stock levels, minimise charges, identify product opportunities and plan marketing campaigns. This goes beyond basic automation into generating proactive insights. This is interesting in its own right, but also speaks to Amazon’s bullish strategy on AI: this tooling is for FBA today, but it shows the direction of travel for Vendor Central and other systems. Secondly, on the customer side of the transaction, Google launched Agents Payment Protocol (AP2), a standard way for AI agents to be authorised to make purchases on behalf of consumers. More than a third of UK consumers (more than half in the US) are already using generative AI for research and product discovery. Whether purchasers will be happy with the idea of an AI personal shopper is a big question. I was at the Independent Publishers Guild Autumn Conference on Tuesday, and had the pleasure of chairing a panel with Isaac Jones of the BMJ and Lauren Ingram of Next Big Thing, discussing the impact of AI on websites, and how to optimise website content for AI models. Publishers need to strike a balance between controlling how the content on their website is used by AI models, preserving traffic, and being visible and hyperlinked in AI outputs—or, as I thought about it afterwards, having your click and feeding it. This is a fundamental question given that point above on the importance of AI for product discovery. This is a complex and nuanced topic, but as a start on a checklist for publishers, my takeaways from the session were: there will be a range of perspectives even within a publisher—editorial, marketing and technical stakeholders look at this through their own lens, so coordination is essential; quality of content (including specificity, uniqueness and credibility) is fundamental, which should play into publishers’ strengths; and experimentation is key, including putting yourself in your audience’s shoes and using the platforms they do. Do get in touch if you’d like to discuss what this means for your business. Some highlights from OpenAI’s new research paper on how people are actually using ChatGPT: around 10% of the world’s adult population has used ChatGPT; 18 billion prompts are being sent each week; and non-work use cases (advice, shopping) account for 70% of usage, up from about half a year ago. There’s a sobering report here on librarians’ experience of AI: patrons showing reduced critical thinking, and requesting hallucinated books. Thanks to Thad McIlroy for highlighting an essay on AI in the workplace that argues most organisations are essentially in the business of converting work from one format to another. As it points out, there’s a natural role for AI in this, and the implications it identifies for organisations echo the HBR piece I shared last week on consulting firms. Broadly it rings true. But two nuances strike me. First, AI is very good at summarising—many of the examples in the essay are essentially reductive—but often less capable when the task is to expand short form into long form. Second, what AI produces often loses individuality. In Oxford this week, I spoke with a Mandarin translator about how AI translation can be powerful but flattens linguistic nuance. For some tasks in business and publishing, the output may be good enough. But publishing is a business of meaning-making as well as information—and that’s where the human voice remains indispensable.
This was originally published in my email newsletter. To receive weekly updates on how AI is affecting the publishing industry, sign up here.