AI for Breakfast

Happy Friday! What's good?

No surprise, this email has a slight shift of focus this morning – the new book project demands it. Typically, on Fridays, I send an email detailing the week's news about the business of health technology. For the next few months (at least), I'll be paying much closer attention to news about AI. I'm sorry. Really.

There are already hundreds of AI-focused newsletters, of course, most of them eagerly trumpeting the technology. That trumpeting is making all sorts of horrible noises throughout ed-tech too. So I'm answering the call. I'm ready to resume my work as "ed-tech's Cassandra" and dutifully remind you that the future of human- and machine-learning as envisioned by Silicon Valley's libertarian elite is a pretty shitty one. (Unlike the mythological Cassandra, however, I don't believe that this future need come to pass as we have much more agency than the gods would like.)

Of course, just because I'm paying attention to AI and education doesn't mean I'm ignoring news about fitness and health technology. I mean, I have been following the story about US ultrarunner Camille Herron, who's been involved in a Wikipedia-editing controversy – deleting accolades from her competitors' entries and "fluffing" the achievements listed in her own. It's all pretty wild, because Herron is undoubtedly one of the most accomplished distance runners of all times – she holds 12 world records. But according to one source, this pattern of bullying online extends offline as well, as other runners have accused Herron and her team of unsportsmanlike conduct in races as well.

For all the flaws in the Wikipedia model – all the complaints about over-zealous editors controlling what can and cannot appear in an entry – the site does seem to be one of the few places online that's not overrun with "AI slop." For now, at least. "A thriving underground economy is clogging the internet with AI garbage — and it’s only going to get worse," writes Max Read in New York Magazine. Facebook, Google News, Kindle, blogs, once-respected magazines and newspapers like USA Today and Sports Illustrated – all filled with ChatGPT generated shit. The web is becoming unusable – a great irony since that's supposedly where all this brilliant machine learning is siphoning data to beef up its "intelligence."

Choire Sicha chatted with Read in his NY Mag newsletter, The Dinner Party, and the two discussed Reddit, which too is now full of "AI slop" – a shame, not because Reddit has ever been a great place on the Internet but because it has been a human place (or at least, to borrow the title from the 1960 film, "where the boys are"). Indeed, as Google search has become more and more unreliable in recent years – thanks to SEO manipulation and subsequent changes in the search algorithm that make it almost impossible to "search" – more people have been turning to Reddit for answers. Now, not only is Reddit the new frontier for SEO-optimization, but it too is being overrun with machine-generated storytelling. "This is quite bad for Reddit," Read argues, "because what Reddit has going for it is that it’s one of the last places online where you can consistently encounter humans and human opinions and human stories."

FWIW, this is the kind of stuff I'm really interested in exploring in a book about AI and education: how do we build a humane future of teaching and learning, a human future of knowledge and stories, in a world that has decided to go "all in" on machine-generated word-junk?

Elsewhere in AI in education: Fortune profiles Luis von Ahn, founder of Duolingo. Here's a choice excerpt:

Von Ahn knows things could get messy. He is concerned an AI tutor could go “haywire,” and start “getting into, say, I don’t know, some Nazi stuff,” he said. Von Ahn said he’s “come to peace” with the possibility of backlash set off by one of the company’s interactive AI features. “We're prepared. Not in that we have major contingency plans, but I'm prepared for it to happen,” he said. “And that's okay.” 

Speaking of being at peace with Nazi content, in their discussion on "AI slop," Max Read and Choire Sicha note the affinity that fascists have for AI – the AI-generated images used by Trump supporters, by the far-right party in Germany, and so on. "There’s something about the way that the AI’s default to cliché seems particularly appealing to the kind of proto-fascism or, whatever you want to call it, fascist-adjacent politics that we have today," says Read. Pretty fun to think about how this applies to AI-generated open educational content, right guys?

Anyway...

OpenAI Downgrades: "OpenAI to remove non-profit control and give Sam Altman equity," says Reuters. OpenAI CTO Mira Murati has announced she'll leave the company. I don't know if a robot wrote this essay for Sam Altman – "The Intelligence Age" – but it's... not good. "It is possible that we will have superintelligence in a few thousand days," the essay argues. (So 8 years? Sure, Sam. Remind me to check back in in 2031.) But it's this sort of phrasing – "How did we get to the doorstep of the next leap in prosperity?" – that should remind you that your AI editing sucks.

"Heckuva job, Brownie": "A courts reporter wrote about a few trials. Then an AI decided he was actually the culprit." "Google's AI Mushrooms Could Have 'Devastating Consequences'." Wired encourages you to "Use Zoom’s AI Companion to Take Notes and Summarize Meetings" – reduce the world to bullet points. Should be great for higher ed. Also great to see healthcare automate "care" away: "That Message From Your Doctor? It May Have Been Drafted by A.I." "Could AI end grief?" fuck you fuck you fuck you fuck you fuck you fuck you fuck you fuck you.

Thanks for subscribing to Second Breakfast. I'm writing about ed-tech again, yup. But I'm still thinking about bodies and breakfasts – I'll write more about that in Monday's newsletter, maybe. Regardless, I appreciate your support.