Issue #43

On today’s quest:

— More climate-friendly AI?
— Apple’s new AI features
— A new high school with mostly AI teachers
— AI checkers are still bad
— Publishers including anti-AI-training statements
— AI worse than humans at summarizing
— An AI resume builder
— AI Rickrolled a company’s client

A more climate- and privacy-friendly AI

I did an interview with Christopher Penn about AI last week that will run in the Grammar Girl podcast in about a month, but I didn’t want to wait to tell you one of the most interesting things I learned.

The effect of AI on the climate is one of my big reservations about the technology, but one way to reduce your energy use is to install an AI system that runs on your own computer, which also has the added benefit of being more private than submitting your info to online AI chatbots. My house runs on solar, so in theory, I can use AI that’s powered entirely by solar energy if I do it this way. Chris recommended the new Llama 3.1 model.

That’s all I know right now, but I’ll be looking into it, and I thought some of you might want to look into it as well instead of waiting for me to figure it out and report back. (This how-to video looks clear and helpful.)

Apple’s new AI features

Apple just revealed the AI features it will be integrating into the next round of products. Siri is getting an AI upgrade, but I’m more excited about the Apple Watch that will flag sleep apnea, and, if approved by the FDA, the new AirPods that will function as clinical-grade hearing aids. Read more on the event page.

A new high school with mostly AI teachers

David Game College in London is offering a program for students 15 to 17 that will be taught almost exclusively by AI with minor “learning coach” support from humans. It appears to be similar to what we would call an alternative high school in the United States, targeting homeschoolers and students who “have faced challenges in maintaining a presence within a conventional school setting.” The school says learning coaches will be in the room to monitor behavior and teach classes AI can’t do well such as art and sex ed. — David Game College, Sky News

AI checkers are still bad

This post from Mike Masnick about his kid with a precocious vocabulary being dinged by an AI checker was making the rounds on Bluesky recently.

Publishers including anti-AI-training statements

Bill Waters highlighted an article about AI on the Mercer County Library website in which the author said, “I am noticing a trend where printed books now include statements right after the copyright information that prohibit AI from training on their text. This is increasingly common among smaller publishers.”

I have no idea if such a statement would deter AI companies from using the material, but it seems like there’s no harm in including it, and it’s an interesting development.

AI worse than humans at summarizing

Catharine Cellier-Smart pointed me to an article about an Australian government study that found that AI is actually “worse than humans in every way at summarizing documents and might actually create additional work for people.”

Human summarizers scored 81% on the government rubric while AI summarizers came in at 47%. The AI summaries “often missed emphasis, nuance and context.” 

However, the report admits that today’s models are better than the model used in this study and speculated that AI may still be able to competently create summaries in the future. It also seems important to consider how the summary will be used and the vast difference in time it takes to get an AI summary compared to a human summary.

An AI resume builder

Nancy Friedman pointed me to an interesting Instagram Reel of a user who goes by @dareablewearables talking about her experience with an AI resume builder called resume.co. She said the tool got her about 85% of the way there in writing a resume and made the process fun and interesting instead of drudgery.

The most interesting part was the ability to change the wording of the resume to give it a different spin or tone — to make you sound more bold or innovative. The downside, as with most generative AI, is that you have to watch out for hallucinations and strange writing, such as exaggerating a job role or putting the same information in two places.

AI Rickrolled a company’s client

The first surprising part of this story is that a customer service AI tool sent a customer who asked for a training video to a video of Rick Astley’s “Never Gonna Give You Up,” apparently having learned about Rickrolling in its training data.

The second, more surprising part is that the person who made the story public was the CEO of the company that makes the tool. Angrily tweeting about how your own product failed doesn’t seem like a good PR strategy to me, but what do I know? ¯\_(ツ)_/¯ — Futurism 

Quick Hits

Here’s how 7 news audience directors are thinking about Google’s AI overviews — Nieman Lab

Audible rolling out beta program for narrators to create voice replicas using AI — Seeking Alpha

Getty Images drops ‘cleanest’ visual dataset for training foundation models. Getty says the 3,750 images come from “Getty’s wholly-owned creative library,” making them free of copyright problems. — Venture Beat

Canva says its AI features are worth the 300 percent price increase (I downgraded my account.) — The Verge

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 this transformative 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.

^{Written by a human.}