This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.
The people paid to train AI are outsourcing their work… to AI
The news: Many people who are paid to train AI models may be themselves outsourcing that work to AI, a new study has found.
How they did it: Researchers hired 44 people on the gig work platform Amazon Mechanical Turk to summarize 16 extracts from medical research papers, then analyzed their responses for telltale signs they’d been produced by AI models. They estimated that somewhere between 33% and 46% of the workers had used AI models like OpenAI’s ChatGPT to complete the task.
Why it matters: Using AI-generated data to train AI could introduce further errors into already error-prone models. If they generate incorrect output that is itself used to train other AI models, the errors can be absorbed by those models and amplified over time, making it more and more difficult to work out their origins. Read the full story.
To read more about what happens when AI is trained on AI, check out my colleague Melissa Heikkilä’s piece on why we’re hurtling toward a glitchy, spammy, scammy, AI-powered internet.
Robotaxis are here. It’s time to decide what to do about them
In some San Francisco neighborhoods, cars with no driver behind the wheel have become a common sight. Many of the city’s ghostly driverless cars are commercial robotaxis, directly competing with taxis and ride-hailing companies.
I spent the past year covering robotaxis for the San Francisco Examiner and have taken nearly a dozen rides in Cruise driverless cars over the past few months. During my reporting, I’ve been struck by the lack of urgency in the public discourse about robotaxis. It’s high time for the public and its elected representatives to play a more active role in shaping the future of this new technology. Like it or not, robotaxis are here.
Now comes the difficult work of deciding what to do about them. Read the full story.
The chip patterning machines that will shape computing’s next act
When we talk about computing these days, we tend to talk about software and the engineers who write it. But without the hardware and the physical sciences that enabled their creation—disciplines like optics, materials science, and mechanical engineering—modern computing would have been impossible.
Semiconductor lithography, the manufacturing process responsible for producing computer chips, stands at the center of a geopolitical competition to control the future of computing power. And the speed at which new lithography systems and components are developed will shape not only the speed of computing progress but also the balance of power and profits within the tech industry. Read the full story.
Both Benjamin and Chris’s stories are from our forthcoming print issue, which is all about accessibility. If you haven’t already, subscribe to make sure you don’t miss out on future stories—subscriptions start from just $80 a year.
The wild race to improve synthetic embryos
Last week, The Guardian blasted out news of a sensational “breakthrough” at the meeting of the International Society for Stem Cell Research in Boston.
The claim was that a researcher named Magdalena Zernicka-Goetz, based at the University of Cambridge and Caltech, had created “synthetic human embryos” using stem cells as the starting point.
What irked me about this story was that this idea isn’t so new. The amazing fact that stem cells will self-organize into structures sharing features with real embryos has been known and studied for several years, as we first reported in 2017.
Since then, several labs have been in a competitive race to make these “embryo models” more complete, more realistic, and ever more like bona fide embryos, complete with placenta tissue.
What irked scientists was thatZernicka-Goetz appeared to claim to have finished the race—but did so in the media, without providing any scientific proof.
The twist in the story is that there actually was a breakthrough, but it came from a different lab… Read the full story.
Antonio is standing in for Jessica Hamzelou, who normally writes The Checkup, her weekly newsletter giving you the inside track on all things biotech. Sign up to receive it in your inbox every Thursday.
I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.
1 The US navy heard the Titan submersible ‘implode’ days ago
But because it wasn’t definitive, officials decided to continue searching. (WSJ $)
+ The tragic mission is now likely to be consumed in messy legal proceedings. (Vox)
+ The operator’s CEO had an unconventional approach to safety measures. (FT $)|
+ Film director James Cameron claims he’d heard safety concerns about the sub. (NYT $)
2 Generative AI models are likely to fall foul of EU rules
Copyright, in particular, is likely to be a real sticking point. (FT $)
+ Some real evidence for why those AI-stealing-our-jobs headlines are overhyped. (Wired $)
+ Stability AI wants its model’s content to look even more realistic. (Bloomberg $)
+ Five big takeaways from Europe’s AI Act. (MIT Technology Review)
3 It’s a year since the US Supreme Court overturned Roe v Wade
We still don’t fully understand what comes next. (WP $)
+ The cognitive dissonance of watching the end of Roe unfold online. (MIT Technology Review)
4 Canada will force Google and Meta to pay for news
It’s possible the platforms could strike a deal, as Facebook did in Australia. (BBC)
6 The FBI demands police silence in exchange for surveillance aid
Forces are made to sign NDAs before helping them to track suspects. (Wired $)
7 E-bike fires are on the rise in New York
13 people are known to have died this year alone, and gig economy workers are at high risk. (The Guardian)
9 How immunotherapy could help correct immune systems
The cancer treatment could treat systems themselves, as well as attacking rogue cells. (Economist $)+ Bacteria is also an effective way to treat some cancers. (New Scientist $)+ The quest to show that biological sex matters in the immune system. (MIT Technology Review)
10 Liquid metal could turn any object into a smart device
Which could be a real help on space missions. (Ars Technica)
Quote of the day
“I don’t have to trust or not to trust.”
—Thierry Breton, the EU commissioner for the internal market, explains to a reporter that his recent trip to “stress test” Twitter’s compliance with Europe’s new digital content law doesn’t rely on his own personal convictions, reports the Wall Street Journal.
The big story
February 2020Google’s most advanced computer isn’t at the company’s headquarters in Mountain View, California, nor anywhere in the febrile sprawl of Silicon Valley. It’s a few hours’ drive south in Santa Barbara, in a flat, soulless office park.
This is the computer that Google is betting on to beat IBM in a race to be among the first to usher in a new era of machines that would make today’s mightiest computer look like an abacus, but through very different approaches. And it’s these differing goals that could influence which—if either—comes out ahead in the quantum computing race. Read the full story.
We can still have nice things
+ A real Italian tried a deep dish pizza, and unsurprisingly, he wasn’t terribly impressed. 🍕
+ How design collective Hipgnosis redefined the album cover: from AC/DC to Pink Floyd.
+ Blogger-zines are back!
+ Remembering the incredible legacy of Alan Turing, who was born on this day 111 years ago.
+ Drop everything: this is what it’s like to eat a $120,000 banana—that also happens to be a work of art.