Amid all the chaos at OpenAI I hope this paper in Nature gets some attention. It finds that larger language models become *less* reliable for a variety of reasons—for one, where earlier models avoided questions they couldn't answer, newer ones are more likely to make something up

https://twitter.com/lexin_zhou/status/1838961179936293098

There's a lot more in here too — larger models do better with easy prompts but still make enough errors that they can't be relied on, and that the nature of the errors are such that human supervision in cleaning them up doesn't help much. In fact...

On a Labor Day weekend 10 years ago, I stumbled onto a journal article about the Luddites that transformed my perspective. These weren’t idiots who smashed what they didn’t understand, but a labor movement that saw exactly how bosses used tech to exploit them—and fought back. I wrote a piece for VICE called 'You've Got the Luddites All Wrong', based on what I'd learned, and haven't looked back.

As I wrote in 2014:

This week, in a genuinely stunning series of events, a bill that was written to save local journalism from big tech was gutted in a backroom deal—and wound up bailing out Google and *boosting AI* instead. You can't make this stuff up.

Here's how it happened. The CA Journalism Preservation Act, was one of two bills that passed CA's legislative bodies with supermajority, bipartisan support.

Like laws in Canada and AUS that generate millions for newsrooms, they charge big tech for profiting from sharing news on their platforms.

But.

Fed up with the terrible state of privacy on the web, a former Google engineer has built a search engine to uncover every privacy violation on the web.

My latest for @WIRED, on @tim_libert and his new tool, webXray:

wired.com/story/webxray-… Libert has studied the way that web companies like Google and Meta track our online activity for years, primarily through cookies generated when we load a page, first as a grad student at UPenn, a postdoc at Oxford, and a professor at Carnegie.

Then Google came calling.

AI is already killing jobs in video games. My investigation for @Wired reveals that major studios are already using AI for concept art and asset generation, foisting AI trainings on games workers and that AI is playing a role in the mass layoffs that have roiled the field. As one veteran AAA games worker told me, “It’s here. It’s definitely here, right now… I think everyone’s seen it get used, and it’s a matter of how and to what degree. The genie is out of the bottle, Pandora's box is opened.”

wired.com/story/ai-is-al…

This is a great story about exactly how management can use generative AI to degrade and replace jobs—it's a lengthy process, with junior and precarious workers replaced first, and more senior jobs deskilled into ones tasked with overseeing AI output.

bbc.com/future/article…
Workers are downgraded into 'AI checkers' or 'humanizers' and paid a fraction of what they used to get for original writing. Expect to see variations of this in just about every creative text- or image- based industry.

Last week, I read about one of the bleakest uses for gen AI I've seen yet: First Horizon Bank is rolling out a system to detect when a call center worker was on the brink of "losing it"—and play them an AI-made montage of family photos set to their favorite song to calm them down Pretty bleak! Instead of, you know, giving workers a real break, or meaningful time off, this corporation was using AI as a way to paper over their problem—a problem, as it turns out, that was created by overloading on a *previous* automation technology.

Welp. After laying off the editorial staff of VICE, a "venture operator" called Savage Ventures is resurrecting its big verticals—Motherboard, Noisey, Munchies etc—presumably with part-time contractors. Or even, judging by a quick look at Savage's offerings, AI-generated content

https://twitter.com/jayasax/status/1788577508503199892

Savage Ventures hosts a bunch of clickbait web verticals (along with maybe one real one), and even its tagline — "Our collection of brands guide millions of individuals to decisions." — feels like it was written by ChatGPT.

It’s pub day for Blood in the Machine, the book I’ve been working on for over five years. It tells the story of a chapter in history too often papered over—how workers rose up at the dawn of the Industrial Revolution when entrepreneurs used machines to take or degrade their jobs. These workers, these Luddites, did not hate technology at all, though the elites of the age—and every age since—hoped the public might be convinced that they did. That would make it easier to dismiss those inclined to question or protest exploitative technologies.

Okay, how about this. How about we take the "top AI experts," tech giants and CEOs *at their word*: AI is an existential threat to humanity. Given that, let's do what we do with other such threats: Ban it. Ban it from the commercial marketplace. It's the only way to keep us safe. We don't allow virologists to manufacture deadly agents for commercial sale, nor do we let weapons makers to sell nuclear weapons to the public.

AI researchers may apply for permits to work on AI programs for non-commercial uses in regulated environments

https://twitter.com/justinhendrix/status/1663509719120576512

Writers and artists standing up to AI are the modern-day Luddites—and for all our sakes, we’d better hope they win.

I've been researching and writing about the Luddites for oh 3-4 years, and it's a good time to discuss the parallels and the stakes, which get higher every day. First, when I say “Luddite,” I'm talking about the men and women who saw businessmen using automated machinery to drive down wages and subordinate workers into factories—bosses aiming to profit at their direct expense—and organized a spirited rebellion against them.

Tara Hernandez & Damon Lindelof's new show "Mrs. Davis" is getting rave reviews. It's great for another reason, too—it shows that the best approach to consumer AI may simply be to say, "no."

I spoke with the creators about why it's OK to be a Luddite:

latimes.com/business/techn… The show is remarkable for a lot of reasons—I binged it all last week—but chief among them is that it doesn't have a cartoonishly dystopian view of how AI will enslave or destroy us all. It knows how tech addicts and entrances us—and it sends its hero on a mission to destroy it.

This is not a new sentiment by any means but it's fairly staggering what a mess has been made of the big digital media companies. So much talent assembled at BuzzFeed and VICE et al, so much great work. Now they are sinking into the sea, the journalists going down first VICE was never perfect, never even great, especially regarding management, but it *did* feel like there was so much promise, freedom to take swings, do ambitious or weird stories that wouldn't have fit at legacy outlets. It really feels now like that promise has been squandered

This is also a good a wake up call to everyone who enjoys living in a world where written materials are comprehensible, accountants do due diligence, and programs are built with good code—because mgmt is going to *try* to replace it all of with subpar, unreliable ChatGPT output

https://twitter.com/gregisenberg/status/1648677152005451777

For the record, this guy's thread I QT'd above *is not* the good wakeup call, it's utter tripe; derivative business mgmt literature at its worst. Greg here is trying to profit off the moment, by fanning the phenomenon, suggesting AI text output *can* and *should* replace people

This is just such a disaster on every count, it’s so embarrassing for 60 minutes to have bought this hook line and sinker without ever having consulted a credible AI scholar or tech journalist

https://twitter.com/60Minutes/status/1647742247444553732

Google [screens the scene in Her where Scarlett Johansson’s AI transcends for 60 Minutes] “This is the future.”

60 Minutes: “The humanity at superhuman speed was a shock. How is this possible?"

Silicon Valley Bank broke because venture capital-dominated Silicon Valley is broken

latimes.com/business/techn… Wrote this before the news that all deposits, even the ~90% that were not FDIC insured, would be made whole, and am only more convinced it's true now

The big tech layoffs aren't about "economic realities" or simply streamlining operations—Silicon Valley is moving to slash rising wages and bring an increasingly empowered workforce to heel.

My 1st column for @latimes:

latimes.com/business/techn… @latimes Historically, this approach has worked, as Malcolm Harris (@BigMeanInternet) points out in the piece. Big tech has kept labor costs low via a range of techniques, from mass layoffs to, most famously, perhaps, outright collusion. This time, however, it might finally backfire.

This is how it begins

https://twitter.com/jasoncolavito/status/1611710986871767041

AI is not going suddenly “replace” workers, it’s going to give employers a massive tool to wield leverage over them, to justify eroding wages and standards

The lesson from all this good Facebook reporting is very simple, and was clear before the massive leaks: If you build a social network whose profits depend on continually expanding engagement, ultimately, whatever drives that engagement will be allowed to triumph. That's it. The reports today are filled with accounts of dissenting employees and researchers pleading that their work be taken seriously—but whether the failings are because of sprawling, inept bureaucracy or direct favoring of business imperatives by leadership, the failures are the same

Uber and Lyft's pitch was always that they'd reduce congestion, emissions, car ownership and the total costs that cars impose on society. A new paper shows that the complete opposite has happened—taking an Uber costs society ~60% more than your own car.

pubs.acs.org/doi/10.1021/ac… There is only one (1) net environmental benefit of people taking a Lyft and Uber over their own car, and that's with particulate pollution exhaust, which is less because cars are being started less frequently since drivers are moving between riders without turning off their cars

My prediction is that this is going to be a disaster for everyone involved — the workers, Kroger, even customers — a glaring example of shitty automation in action

bloomberg.com/news/features/… First, it only looks fully automated—tucked away in the piece is that each of these facilities will need 400 human workers; drivers, baggers, etc and who will have to work strenuously to keep up with the machines. As we know from Amazon, this is likely going to be a nightmare