Well, I could say for the next few months (or even years?) that it’s been one of those weeks, but here goes!
Who owns your cloned AI voice?
This powerful piece in MIT Technology Review – especially for me, daughter of a father who died of ALS – talks about people who, due to a neurological disease, lose their muscle strength but can regain the use of their “own” voice thanks to AI. Wow, that’s truly a useful application of AI, wouldn’t you say? Can’t have anything against that, Hermans!
However… not in late-stage surveillance capitalism. Because what does this piece show? If you use badd language—say something like “get your arse down here”—you get banned! It reminded me of that article from 2022 about people whose eye implants suddenly failed because the company behind them went bankrupt.
It perfectly illustrates what I’ve mentioned in several places (including at the very end of last year’s VPRO podcast De Machine): I truly have nothing against AI per se. Someone who trains their own AI on their own voice, text, images—and then uses it in any way they see fit—has my blessing. You’re not excluding anyone, you’re not stealing anything, and you probably won’t end up any dumber; perhaps even the opposite. (And if you’re training it in the middle of the day in your house packed with solar panels—well, that’s perfectly fine; since giving back to the net doesn’t yield much these days, you might as well run a small AI!)
Speaking of voices and AI, there’s also an article in TechCrunch about software that can adjust accents live! Yes, because we certainly don’t want customers exposed to the full spectrum of ways people actually speak! They describe their mission as a “deeply human mission to break barriers and reduce discrimination.” Well, if I were to dress up as a man, I might experience less sexism, but that doesn’t make the world any less sexist.
Using AI doesn’t save you time—it makes you slower
So far, I’ve mostly reported on “regular” news, but there’s plenty happening in the scientific realm of AI as well. One popular claim is that generating text with AI saves time—but is that really so? In an insightful article, three medical researchers outline several reasons why it might not. Synthetic text reads differently than normal text and may actually require more time to process. Plus, the feeling of being responsible for a text that isn’t your own does something to you as a person. Other recent studies confirm this: one Chinese study involving over 6,500 radiologists found that using AI can increase the risk of burnout, because in rare cases a lot of time is needed to build up context by digging into the details. Another study among nearly 200 doctors communicating with patients via AI showed they spent just as much time—if not a bit more—as when doing it manually (though the difference wasn’t statistically significant).
In a completely different field—programming—we see similar results. Recent research from GitClear shows that as AI usage increases, there’s also a rise in “code churn” (lines of code that are frequently modified). The hypothesis for that is that AI-generated code isn’t as robust, and therefore needs fixing more often.
A small study one of my students at VU last year revealed something similar in education: reviewing and refining AI-generated feedback takes more time than doing it yourself. My own argument is that if you truly want to check whether something is correct, you not only have to review the AI’s text but also think about whether something important is missing—and that’s only possible if you’ve done the thinking yourself.
ChatGPT Is Getting More Conservative
Well, last week in NRC I wrote about the covert abandonment of the ambition to be neutral, and now TechCrunch reports that ChatGPT is becoming more conservative—since, of course, everything must be allowed to be said, free speech and all. It reminded me of an article I wrote in 2016 about Paul Graham; back then, he said he’d “go into resistance” if Trump won, even though he was still tight with Peter Thiel (whom we now all know as JD Vance’s mentor). I believe my piece has stood the test of time and still clearly shows why it isn’t such a good idea to allow all ideas into the public debate.
Your Book, My Book?
The Verge reports that starting February 26 (that’s next Wednesday!) you’ll no longer be able to download ebooks from your Amazon ebooks dashboard. This small change (since everyone naturally syncs over Wi-Fi) raises interesting questions about who really owns a book you’ve paid for. It makes clear that in today’s world, when I buy a book (often at a price similar to a physical book), I’m only purchasing the right to read it—not to truly own it, cut it up, pass it on, annotate it, and so on. Also, since Amazon is in “Team Trump,” it might just be that the dozens of feminist books in my library could soon disappear for being too woke. Some really wild things have happened in America lately.
So if I were to advise someone on what they absolutely must not do, it would be to quickly download all your AZW3 files and convert them to epub with something like Epubor Ultimate. I certainly didn’t do that with the no less than 146 books in my library (as they say: buying books and reading books are two entirely different hobbies).
Hype Is Necessary in Silicon Valley
This interview with Meredith Whitaker from Signal in 2023 was widely shared last week, but I’m including it again because of her brilliant take on hype:
“Venture capital looks at valuations and growth, not necessarily at profit or revenue. So you don’t actually have to invest in technology that works, or that even makes a profit, you simply have to have a narrative that is compelling enough to float those valuations.”
Hype and growth are essential—it’s not just about making “plain” profits. It reminded me of that disheartening article (also from 2023) about the Instant Pot, which explains well that we’re not dealing with ordinary capitalism where companies merely want to make money, but with shareholder capitalism (see also the work of the inimitable Ed Zitron).
A Quirky (Online) Nerd Conference
A little treat for the nerds among you! Next week I’m speaking at the online conference HYTRADBOI, and the other talks look really fun too. There are still tickets available!
Good News!
Now for some good news! It might not be earth-shattering, but it’s certainly amusing: the company behind the humane AI pin—a sort of “ChatGPT brooch”—is shutting down because they couldn’t get it off the ground and people didn’t think it was worth the money. Well, who could have predicted that? (Also, remember Google Glass.)
And we could all use a bit of positive government news! Because the government might not be great, but perhaps no government is even worse. Here are nine examples of successful government policies, along with a thoughtful analysis of how Rwanda managed to contain the Marburg virus through swift action (I could comment on America and bird flu, but this is the good news section…).
If you’d like to read more about how governments can actually be pretty decent, then the book The Entrepreneurial State is highly recommended!
This post is a translation of the original newsletter, published in Dutch.
