HOW do I write a newsletter THIS WEEK? The rest of this newsletter isn’t about US politics, but it’s been the subject of most of the conversations I’ve had since Sunday. I’m excited to vote for President Harris this November!

But on to Catbird Content news – coming up soon:

• July 29: Content Design for Developers at THAT Conference, WI

• August 19: Decisions over Deliverables at Growing in Content (15% off with code TORREY15)

• September 29: BUTTON CONFERENCE! I’m not speaking at this one; I’m paying my own way. You’ll see me all over the conference Slack and watching as many talks as I can.

Finds

Have you seen this graph? It’s from Forrester, and it’s an analysis that reflects the enshittification you may have been reading about. That’s the idea that platforms on the internet, apps, and other products and services we use are getting worse.

It’s not just internet and apps, either. Across industries (all but airlines, somehow), “Customer Experience Quality In The US Falls To An All-Time Low.”

I’m sure it’s not entirely due to layoffs and cuts to product teams to entice Wall Street to boost stock prices, but this feels like the kind of correlation that makes me pay attention.

Here’s what this trend means to me: If people had a real choice between companies, they would get to choose services/products that they hated less–and these companies would not allow their CX scores to drop so far. Companies would have financial motivation to satisfy their customers, and not just the stock market. It’s yet more evidence that monopolized industries are fundamentally bad for people.

Flexes and fumbles

I HAVE SIGNED A NEW CONTRACT FOR A NEW BOOK! 🎉🎉🎉 I will be saying more about it soon, but for right now I can say this: It’s related to the Content Design Skills Project and goes even farther. The work is happening already! I’m working with two amazing co-authors, and we are working hard to make a book that promises to be even more useful than Strategic Writing for UX.

Also: Don’t worry, I’m still on track to meet deadlines for the second edition of Strategic Writing for UX! The first two revised chapters are available on the O’Reilly platform as “RAW and UNEDITED.” So if you see something amiss there, especially if you think the substance is inaccurate, please tell me.

My fumble of the last few weeks wasn’t public, but did affect my manuscript quite a bit. I realized my thinking was being reflected in fuzzy ways by inconsistent language. So I took a Coursera course to get my terminology straight when talking about AI and LLMs. I think it’s very important to be internally consistent and accurate when I use terms in the book, because I want people to be able to rely on what I write there. The last thing I want is for people to quote me and get worse outcomes for themselves because I used words in ways that contradict the norms.

But those norms are changing, and fast. Even in the past year, since I left a team at Google that was working with these technologies, the use of terms has shifted. Or it’s possible that the difference was between my former team and the industry as a whole; it’s hard for me to say.

Also, I have some old opinions, newly reawakened: If people teaching about AI technologies (or any new technologies) are going to talk about its capabilities metaphorically, then they have the responsibility to make a clear line about where the metaphor doesn’t apply. To be specific, the next person who talks to me about “the LLM’s reasoning” is going to get an earful about how the LLM is not capable of reasoning.

Philosophy

Y’know how at the top I talked about monopolies, and then about AI? I’m going to tie them together right here.

To an LLM, the only thought possible is the prediction of the next word (the completion) in response to a stimulus (the prompt.) That is, you can prompt ChatGPT with a question or sentence, and ChatGPT responds with a word that is the most likely to show up next. And then the next word. And then the next.

It makes these predictions based on all the language it has learned from so far, with the static relationships between each token that makes up that language. That’s it.

So the LLM’s world exists within itself, and it is fully described by that world, based on the genericization of all the information that went into it.

That’s fundamentally different than the world that exists for a human. My world exists not only within my own understanding, also in my interactions with the world. I don’t have a training_on=FALSE mode like the AI; I don’t get to choose whether I learn from this world. And even better, I’m sharing this world with billions of other entities. Anything I do, whether it’s writing a new book or grocery shopping or making software, is with other people and their experiences.

Setting aside the jokes about the difficulties of group projects, we all benefit by doing these projects among other entities, that have their own imperfect-but-usually-grounded-in-truth datasets and learning. Having diverse viewpoints mean that we come up with ideas that we wouldn’t have come up with individually.

This makes us more resilient as a species, because we don’t create single points of failure. The more we rely on monopolies for services, the more brittle are the systems that rely on those services (see Crowdstrike, for example). With AI, we’re relying on monopolistic thought: a very few “foundational models” that have homogenized an unknown number of sources, with undocumented biases, into individual entities that spit out genericized content.

AIs can only emulate thought in a monopolistic way, as a result of recursive homogenization of inputs and training. I prefer to work with people rather than AI, because then we get the diversity of their experiences, ungenericized, ready for novel creations.

Or, as the great Octavia Butler wrote in Parable of the Sower (the novel is set, fictionally speaking, starting on July 20, 2024):

Embrace diversity.

Unite—

Or be divided,

robbed,

ruled,

killed

By those who see you as prey.

Embrace diversity

Or be destroyed.

I write these newsletters myself, and I stand by what’s in them. If you have kudos, concerns, or questions, please tell me. —Torrey