Issue #49. The Prodigal Newsletter, Holiday Edition

On today’s quest:

— Brittanica embraces AI
— AI use cases are fragmented and varied
— AI use is widespread in business
— ChatGPT launches Canvas for writers
— OpenAI and Anthropic improve their models’ writing abilities
— Apple’s AI news alerts sent out misinformation
— People seem to prefer AI poetry

Brittanica embraces AI

Brittanica is best known as the parent company of Encyclopedia Britannica, but it also owns Merriam-Webster and gets its biggest chunk of revenue from educational software. The company is all in on AI and now uses the technology “in creating, fact-checking and translating content for its products, including the online Britannica encyclopedia.” It has also built AI chatbots based on its data and is integrating AI further into its products for schools and libraries. Of special interest to editors and writers, Brittanica plans to expand Merriam-Webster’s thesaurus with the help of AI.

The company says it currently sees profits of 45% and is making plans to go public. — Gizmodo and New York Times

What people are really doing with AI

Anthropic released a report describing the top activities it sees people doing with its chatbot, Claude:

• Web and mobile app development (e.g., coding) (>10%) 

• Content creation and communication (>9%)

• Academic research and writing (>7%)

• Language translation (~4.5%)

Survey finds AI use is widespread in business

The Federal Reserve Bank of St. Louis reported that an August survey showed “nearly a quarter of the US workforce was already using [AI] weekly” and that the number was closer to 50% in the software and financial industries. Reinforcing the findings above from Anthropic, the Financial Times reported that “most of these users were turning to tools such as ChatGPT to help with writing and research, often as an alternative to Google, as well as using it as a translation tool or coding assistant.”

Thinking about job loss, energy use, and the tsunami of AI slop

I recently released a bonus episode of the Grammar Girl podcast about the downsides of AI that originally went to just Grammarpalooza subscribers back in October. I have a lot of concerns that I bounced off Christopher Penn, who thinks about this stuff full time.

ChatGPT launches Canvas for writers

ChatGPT’s new Canvas mode gives you an interface designed for writing projects. It has side-by-side panels showing the piece you are working on next to the chat panel. As before, you can ask ChatGPT to make edits to the document, but it’s easier to see the document now that it’s not interspersed with prompts. More importantly, you can also make edits to the text, and if you ask ChatGPT for feedback on the document, it can add highlighted comments to the text, much like a human collaborator would leave comments. It doesn’t have full change-tracking features like EditGPT, but you can click “apply” on the comments to have it implement its suggestions, and you can click another button to highlight the changes it has made. The first 8 minutes of the video are worth watching. — OpenAI

OpenAI and Anthropic improve their models’ writing abilities

OpenAI also recently announced an upgrade to GPT-4o, specifically mentioning that the model’s “creative writing ability has leveled up” after Anthropic made a similar announcement about Claude 3.5 Sonnet’s writing. Scale compared the writing from the two chatbots in a fascinating posts with lots of short examples of creative writing prompts. Because the post was written by AI guys and not writers, it also has interesting tidbits about how different models get their different “personalities” through character training. Read to the end for a trippy response from Claude. — Scale

Apple’s AI news alerts sent out misinformation

Apple’s new AI system, Apple Intelligence, recently introduced misinformation into news alerts. The BBC complained that the system sent a news alert attributed to the BBC that suggested Luigi Mangione, the alleged shooter of the United Health CEO, had shot himself, which is not true. The alert read “Luigi Mangione shoots himself.” — BBC

Why do people seem to prefer AI poetry?

A new study found that people couldn’t tell the difference between poetry written by humans and poetry written by generative AI. Further, people actually preferred the AI poetry. The study participants were just regular people, and I found the researchers’ final hypothesis interesting: “The simplicity of AI-generated poems may be easier for non-experts to understand, leading them to prefer AI-generated poetry and misinterpret the complexity of human poems as incoherence generated by AI.” — Scientific Reports

Typos, good?

I feel like I’m seeing the idea more often that writing errors are actually good because they prove your human.* Here’s one example from Rua M. Williams at Bluesky:

Quick Hits

Duolingo expands AI tutor after seeing 54% increase in subscriptions — Wall Street Journal

I’m a Doctor. ChatGPT’s Bedside Manner Is Better Than Mine — New York Times

The IRS will use AI to do more audits — Barrons

ByteDance is laying off hundreds of Malaysian TikTok moderators and replacing them with AI — TechCrunch

Another podcaster ponders the future of the industry after Notebook LM — Erik Jones

News Corp outlets sue Perplexity over scraped stories — TechCrunch

Spammers are using AI to populate zombie news websites and using the names of real people for the fake bylines — OPB

Science teachers in Britain used ChatGPT to cut lesson-planning time by ~30% without any loss in quality — TES Magazine

What is AI Sidequest?

Using AI isn’t my main job, and it probably isn’t yours either. I’m Mignon Fogarty, and Grammar Girl is my main gig, but I haven’t seen a technology like AI since the development of the internet, and I want to learn about it. I bet you do too.

So here we are! Sidequesting together.

If you like the newsletter, please share it with a friend.

* See what I did there?