In this issue: What can a publisher’s plugin offer that ChatGPT can’t already do? The latest on AI use and trust in news. A reality check on AI reading model weights. And David Caswell on why we might be underestimating AI in the short term.

What we’re talking about: OpenAI wants you to build plugins for ChatGPT. 4,000 have already been added to the plugin directory. Order a pizza? Make a to-do list? Look up whether granddaddy was in the Nazi party? There is a plugin for that.

At OpenAI’s DevDay in Berlin, the company promised developers that ChatGPT would automatically recommend plugins to users (of which there are 1.2 billion, they reminded everyone) at the right moment. Three ranking factors will come into play: how often suggested plugins are installed, how long people use them, and how often they return.

Should publishers build plugins? That depends. It’s an interesting feature to offer to your existing customers. Connect your chatbot to your subscription. I think this is gonna be pretty standard before long.

But could a plugin help attract new users? I think the chances that ChatGPT will suggest a plugin that just displays articles are rather slim. Just go ahead and prompt the news. Tell Chat to make an interactive HTML page with the Texas election results. Pretty good.

Why should a user install a plugin (and then register or subscribe) instead of just asking Chat? You would need a much better experience or access to exclusive data, or both. This might work out for specialized data. Or local news.

The Economist has a plugin showing Trump’s approval rating. The Central News Agency of Taiwan gives access to news. SPIEGEL has the Nazi files. What else is out there for a general audience? Let me know!

What I’m reading:

And now: What if we’ve got it backwards? David Caswell thinks we may be underestimating what AI can do in the next year, while overestimating how much our lives will change in the long run. He also has a question for the information ecosystem that comes next: How do we keep it trustworthy when machines are producing and consuming more of the information?

Three Questions with David Caswell

David Caswell

David Caswell is co-founder at Signals at Scale and has worked at the BBC, Tribune Publishing and Yahoo! 

What's the most important question right now?

We are clearly nearing a substantial inflection point in how AI is used to mediate societal information, and the next year or two promises to bring AI that can substitute for human beings on almost all complex knowledge tasks. As that takes hold, machines will become the primary producers and consumers of information and therefore vastly more information will flow through society than in our present ecosystem. We urgently need to understand how that new ecosystem will operate and how we can maintain truthfulness and trustworthiness in that environment. I don’t think words like ‘news’ and ‘journalism’ are adequate for this new AI-mediated information ecosystem, and I think that we will probably need to be much more systematic in our thinking about it.

What will we be shaking our heads about a year from now?

I think that there is a lot of insight in the cliché that we tend to overestimate the impact of tech in the short term but underestimate it in the long term, but in this situation I am increasingly thinking that it might be the reverse. In the short term I think we are probably underestimating the AI capabilities that we will have access to within a year – including our ability to fully automate much of editorial work. In the long term, however, I think we may be overestimating how important raw intelligence is in much of human society – including in media and information consumption. We may have surprisingly familiar and recognizably ‘normal’ lives, jobs and even media experiences – based disproportionately on social relationships, habit and identity – all while we live surrounded by near magical AI. Economies, information access and empowerment from information will probably change dramatically, but maybe our lived lives won’t change as much or as fast as we think.

What future are you looking forward to?

I’m already in it! I find that my ability to consume information has already been transformed beyond all recognition by AI – in terms of how much information I can ‘consume’, in terms of how complete or exhaustive my understanding of a subject can be, and also in terms of how much calmer I feel about information. Being overwhelmed by information or experiencing ‘FOMO’ about information is much less common for me now than it used to be. I look forward to a future where this experience extends much further, into all my interests and all the activities of my life, and I can feel much more comfort and security and ease with information and my use of it. I want to feel confident that stuff is being taken care of, and that I am bothered only by the most critical decisions. That kind of experience would probably give us more time for much richer personal interests and hobbies, so it’s probably a very good time to develop personal interests and hobbies! It’s possible that we may spend a lot less time on passive or undirected information consumption from apps and websites, which might be a very good thing indeed!

Hands on: Tough times for publishers. AI platforms can write whole articles, and creators are finding new ways to connect with audiences on social media. Media strategist Lucy Küng is working on a book to help publishers get into the creator space. It’s called Content Shift, and Lucy has started an accompanying newsletter. Like and subscribe, as they say.

One more thing: We’re mourning actor Patrick Stewart, who died last week. I watched him as Captain Picard in Star Trek: The Next Generation. I hope you did too. Just one brief clip: In “The Measure of a Man,” first aired in 1989, Picard defends the android Data’s right to decide what happens to him.

Is Data a person, or is he Starfleet’s property? Nearly four decades later, the questions feel a little less like science fiction.

This is THEFUTURE.