Judging gymnastics has now become a job where AI can help. CC-licensed photo by Jeffrey Hyde on Flickr.
You can sign up to receive each day’s Start Up post by email. You’ll need to click a confirmation link, so no spam.
There’s another post coming this week at the Social Warming Substack on Friday at 0845 UK time. Free signup.
A selection of 9 links for you. Tumbling. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. Observations and links welcome.
The AI search war has begun • The Atlantic
Matteo Wong:
»
existing AI products are absolutely filled with media that publishers have received no compensation for. ([Perplexity’s chief business officer Dmitry] Shevelenko told me that Perplexity will not stop citing publishers outside its revenue-sharing deal [announced on Tuesday with Time, Fortune and some other publishers], nor will it show any preference for its paid partners moving forward.) AI companies don’t seem to value human words, human photos, and human videos as works of craft or products of labour; instead they treat the content as strip mines of information.
“People don’t come to Perplexity to consume journalism; they come to Perplexity to consume facts,” Shevelenko told me in an interview before today’s announcement. “Journalists’ content is rich in facts, verified knowledge, and that is the utility function it plays to an AI answer engine.” To Shevelenko, that means Perplexity and journalists are not in direct competition—the former answers questions; the latter breaks news or provides compelling prose and ideas. But even he conceded that AI search will send less traffic to media websites than traditional search engines have, because users have less reason to click on any links—the bot is providing the answer.
The growing number of AI-media deals, then, are a shakedown. Sure, Shevelenko told me that Perplexity thinks revenue-sharing is the right thing to do. But AI is scraping publishers’ content whether they want it to or not: Media companies can be chumps or get paid. Still, the nature of these deals also suggests that publishers may have more power than it seems. Perplexity and OpenAI, for instance, are offering fairly different incentives to media partners—meaning the tech start-ups are themselves competing to win over publishers.
All of these products have made basic mistakes, such as incorrectly citing sources and fabricating information. Having a searchbot ground itself in human-made “verified knowledge” might help alleviate these issues, especially for recent events the AI model wasn’t trained on. Publishers also have at least some ability to limit AI search engines’ ability to read their websites. They can also refuse to sign or renegotiate deals, or even sue AI companies for copyright infringement, as The New York Times has done. AI firms seem to have their own ways around media companies’ barricades, but that is an ongoing arms race without a clear winner.
«
This all seems like bad ideas all over the place.
unique link to this extract
Why we might never know the truth about ultra-processed foods • BBC News
Philippa Roxby:
»
UPFs are defined by how many industrial processes they have been through and the number of ingredients – often unpronounceable – on their packaging. Most are high in fat, sugar or salt; many you’d call fast food.
…a recent meeting of the American Society for Nutrition in Chicago was presented with an observational study of more than 500,000 people in the US. It found that those who ate the most UPFs had a roughly 10% greater chance of dying early, even accounting for their body-mass index and overall quality of diet.
In recent years, lots of other observational studies have shown a similar link – but that’s not the same as proving that how food is processed causes health problems, or pinning down which aspect of those processes might be to blame.
So how could we get to the truth about ultra-processed food?
The kind of study needed to prove definitively that UPFs cause health problems would be extremely complex, suggests Dr Nerys Astbury, a senior researcher in diet and obesity at Oxford University.
It would need to compare a large number of people on two diets – one high in UPFs and one low in UPFs, but matched exactly for calorie and macronutrient content. This would be fiendishly difficult to actually do.Participants would need to be kept under lock and key so their food intake could be tightly managed. The study would also need to enrol people with similar diets as a starting point. It would be extremely challenging logistically.
And to counter the possibility that people who eat fewer UPFs might just have healthier lifestyles such as through taking more exercise or getting more sleep, the participants of the groups would need to have very similar habits.
“It would be expensive research, but you could see changes from the diets relatively quickly,” Dr Astbury says.
Funding for this type of research could also be hard to come by. There might be accusations of conflicts of interest, since researchers motivated to run these kind of trials may have an idea of what they want the conclusions to be before they started.
«
An unexpected twist lights up the secrets of turbulence • Quanta Magazine
David Freedman:
»
It’s time to feed the blob. Seething and voracious, it absorbs eight dinner-plate-size helpings every few seconds.
The blob is a cloud of turbulence in a large water tank in the lab of the University of Chicago physicist William Irvine. Unlike every other instance of turbulence that has ever been observed on Earth, Irvine’s blob isn’t a messy patch in a flowing stream of liquid, gas or plasma, or up against a wall. Rather, the blob is self-contained, a roiling, lumpy sphere that leaves the water around it mostly still. To create it and sustain it, Irvine and his graduate student Takumi Matsuzawa must repeatedly shoot “vortex loops” — essentially the water version of smoke rings — at it, eight loops at a time. “We’re building turbulence ring by ring,” said Matsuzawa.
Irvine and Matsuzawa tightly control the loops that are the blob’s building blocks and study the resulting confined turbulence up close and at length. The blob could yield insights into turbulence that physicists have been chasing for two centuries — in a quest that led Richard Feynman to call turbulence the most important unsolved problem in classical physics. (Quantum turbulence has become an important problem too.) Untangling turbulence might also prove extraordinarily impactful, given that it plays a huge role in stars, aviation, nuclear fusion, weather, changes in the Earth’s core, wind turbines and even human health — arterial flow can become dangerously turbulent.
If the blob does yield big advances in turbulence, it will add to the growing string of surprising and influential breakthroughs that Irvine and his students have produced in the physics of what might loosely be called spinning stuff — systems composed of whirling objects, fluids and even fields.
«
Fascinating piece about a really important, yet very poorly understood field of physics: fusion reactors (yes I know) are very sensitive to fluctuations in the circulating plasma, and that’s all turbulence.
unique link to this extract
Playing Pokémon by ear • Game File
Stephen Totilo:
»
After Ross Minor was blinded in 2006, he still wanted to play Pokémon.
He was a kid. Eight years old. And he found a way.
“I listened to all my friends playing the game, and I would hear the soundtrack and be like, ‘Oh, I remember that. That song plays when you’re in this town.’ So I learned that each town has a different song.
“Then I learned that each Pokémon has a different cry… All the attacks make different sounds.
“The cherry on top was that, when you run into a wall, it plays this boom-boom sound. So, through that alone, I was able to memorize all the games and form this mental map—and beat the games completely by myself without sighted assistance.”
Minor played 2007’s Pokémon Diamond and Pearl that way, then 2011’s Black and White, and then more as each year’s new Pokémon game came out.
For over a decade, his strategy worked. But 2019’s Pokémon Sword and Shield added navigation in three dimensions, and Minor struggled to draft a mental map. He needed assistance from sighted players.
Then came 2022’s Pokémon Scarlet and Violet. Those, Minor told me, “are just completely inaccessible.”
This is what happens, Minor explained, when a video game’s accessibility for players with disabilities is accidental or unintentional.
Players like him can get left behind.
«
Amazing triumph over the odds – and then Nintendo pulls the rug from under him.
unique link to this extract
The gymnastics world braces for an AI future • The Verge
Dvora Meyers:
»
When Simone Biles saluted the judges and stepped onto the mat to vault at the Sportpaleis in Antwerp, Belgium, it seemed like every camera in the packed arena was trained on her. People in the audience pulled their smartphones to record. The photographers zoomed in from their media perches. One TV camera tracked her run on a high-speed dolly, all the way down the runway, as she hurdled into a roundoff onto the springboard. The spider cam, swinging above, caught the upward trajectory of her body as she turned towards the table and blocked up and off, twisting one and a half times before landing on the blue mat and raising her arms above her head. The apex of human athleticism and kinesthetic beauty had been captured.
But there were other cameras that few other people watching in the arena were thinking about as they took in Biles’ prowess on the event: the four placed in each corner of the mat where the vault was situated. These cameras also caught the occasion but not with the purpose of transmitting it to the rest of the world. These were set up by the Japanese technology giant Fujitsu, which, since 2017, has been collaborating with the International Gymnastics Federation (FIG) to create an AI gymnastics judging system.
In its early days, the system used lidar (light detection and ranging) technology to create 3D composites of gymnasts in action. These days, it uses an even more sophisticated system, drawing from four to eight strategically placed hi-def cameras to capture the movement of the athletes, make 3D models, and identify whether the elements they are performing fall into the parameters established by the judging bodies inside the federation.
But the computer system doesn’t make judgments itself. Instead, it is deployed when there is an inquiry from the gymnast or coaches or a dispute within the judging panel itself. The Judging Support System (JSS) can be consulted to calculate the difficulty score of an athlete’s exercise — a second opinion, rather than an initial prognosis. Currently, it is mostly used for edge cases.
«
Neat article: the judges’ task has always seemed incredibly difficult in gymnastics. (Imagine being a fencing judge before the electronic hit systems; even with those you still get human evaluation of some rounds.)
unique link to this extract
The future of science publishing • C&EN
Dalmeet Singh Chawla:
»
In 2015, the Bill and Melinda Gates Foundation, one of the largest charitable research foundations in the world, introduced a new publication policy that promised to pay publication charges for papers its grantees write as long as the content of the final version of the study was freely available to read somewhere online.
The policy stated that the foundation “would pay reasonable fees required by a publisher to effect publication on these terms.” In 2021, the foundation narrowed that support, clarifying that going forward it would pay only for research published in fully open-access journals, which make all their papers freely available to read.
In March, the Gates Foundation surprised many by backtracking on these policies. It announced that starting Jan. 1, 2025, it would no longer cover publishing costs. The decision is causing anxiety among researchers and publishing experts, who wonder how the open-access model can be maintained if funders don’t foot the publishing bill.
The open-access movement started in the 1990s in a bid to make taxpayer- funded research freely available. Before then, subscription journals were the norm. Researchers typically had access to these journals through their institutional libraries, though cash-strapped universities in emerging-market countries often couldn’t afford the subscriptions. Some publishers introduced waivers for certain universities and libraries in such nations.
…the foundation notes that open access in its current form has resulted in “some unsavory publishing practices,” including unchecked pricing from journals and publishers, questionable peer review, and paper mills—people or organizations that produce fake or subpar papers and sell authorship slots on them.
…The Gates Foundation is now suggesting that authors post online preprints of their author-accepted manuscripts—near-final versions of studies accepted by journals for publication before they are typeset or copyedited—and then publish in whichever journals they like.
«
The latter seems the better model, but still assumes that someone will do the peer review (a necessary step). We still haven’t found the ideal model for science publishing.
unique link to this extract
New study confirms mammal-to-mammal avian flu • EurekAlert!
»
A new Cornell University study provides evidence that a spillover of avian influenza from birds to dairy cattle across several US states has now led to mammal-to-mammal transmission – between cows and from cows to cats and a raccoon.
“This is one of the first times that we are seeing evidence of efficient and sustained mammalian-to-mammalian transmission of highly pathogenic avian influenza H5N1,” said Diego Diel, associate professor of virology and director of the Virology Laboratory at the Animal Health Diagnostic Center in the College of Veterinary Medicine.
Diel is co-corresponding author of the study, “Spillover of Highly Pathogenic Avian Influenza H5N1 Virus to Dairy Cattle” published in Nature.
Whole genome sequencing of the virus did not reveal any mutations in the virus that would lead to enhanced transmissibility of H5N1 in humans, although the data clearly shows mammal-to-mammal transmission, which is concerning as the virus may adapt in mammals, Diel said.
So far, 11 human cases have been reported in the US, with the first dating back to April 2022, each with mild symptoms: four were linked to cattle farms and seven have been linked to poultry farms, including an outbreak of four cases reported in the last few weeks in Colorado. These recent patients fell ill with the same strain identified in the study as circulating in dairy cows, leading the researchers to suspect that the virus likely originated from dairy farms in the same county.
While the virus does have the ability to infect and replicate in people, the efficiency of those infections is low. “The concern is that potential mutations could arise that could lead adaptation to mammals, spillover into humans and potential efficient transmission in humans in the future,” Diel said.
«
*bites nails* Just a watching brief, nothing more.
unique link to this extract
Hacker shows how to get free laundry for life • 404 Media
Joseph Cox:
»
Michael Orlitzky was not having a good day with his laundry. First CSC Serviceworks, a laundry management company, replaced all of the machines in his building with new coin-op or app-powered ones. The card reading machines had been an issue for years because the cards would stop working and the recharge machine would steal dollar bills, Orlitzky said. Now he had another enemy with its own quirks to get used to. Plus, CSC had replaced the machines about a week ahead of schedule, meaning that any cash on his or others’ laundry cards was now worthless and unusable.
Then, one of the new machines ate his quarters. The first machine was stuck on the cold setting, and he had to pay another $2 and move all of his belongings to another machine. He called CSC customer service and was on hold for an hour. CSC eventually told him to get a refund through the company’s website, which in turn insisted he install CSC’s app to proceed.
“That was the day I decided laundry would be free,” Orlitzky told 404 Media in an email.
Orlitzky then discovered multiple bypasses to CSC machines that allow him to wash his clothes for free. Since then, he’s been pretty quiet about the whole thing. Orlitzky published a brief write-up of his escapades on his personal website last year, but hasn’t shared it on social media. Some people in his building know his secret, but that’s about it.
That is, until now, with Orlitzky due to speak at the DEF CON hacker conference in August about how he found infinite money cheats for CSC laundry machines. The talk is called “Laundering Money.”
«
Upside: free laundry. Downside: in the launderette.
unique link to this extract
CrowdStrike outage could cost cyber insurers $1.5bn • Data Breach Today
Mathew Schwartz:
»
The global IT outage triggered by a faulty CrowdStrike software update on July 19 could lead to cyber insurers paying out up to $1.5bn in compensation.
That’s the conclusion of cyber risk analytics platform CyberCube, which in a report said the insurer losses range from $400m to $1.5bn. Those figures represent 3% to 10% of the $15bn in global cyber premiums held today.
The final insurance payout total will need time to emerge. “Determining final losses for the industry is likely to be a lengthy process because cyber insurance policy language is not standardized,” Moody’s Reports said in a Monday report. “It will take time for insurers to determine which customers suffered losses from the outage, and whether those losses are covered.”
Most claims will center on losses due to “business interruption, which is a primary contributor to losses from cyber incidents,” it said. “Because these losses were not caused by a cyberattack, claims will be made under ‘systems failure’ coverage, which is becoming standard coverage within cyber insurance policies.” But, not all systems-failure coverage will apply to this incident, it said, since some policies exclude nonmalicious events or have to reach a certain threshold of losses before being triggered.
The outage resembled a supply chain attack, since it took out multiple users of the same technology all at once – including airlines, doctors’ practices, hospitals, banks, stock exchanges and more.
Cyber insurance experts said the timing of the outage will also help mitigate the quantity of claims insurers are likely to see. At the moment CrowdStrike sent its update gone wrong, “more Asia-Pacific systems were online than European and US systems, but Europe and the US have a greater share of cyber insurance coverage than does the Asia-Pacific region,” Moody’s Reports said.
«
I like the idea that there exists a site called “Data Breach Today”. Turns out that there’s more than enough content to fill it, daily.
unique link to this extract
| • Why do social networks drive us a little mad? • Why does angry content seem to dominate what we see? • How much of a role do algorithms play in affecting what we see and do online? • What can we do about it? • Did Facebook have any inkling of what was coming in Myanmar in 2016? Read Social Warming, my latest book, and find answers – and more. |
Errata, corrigenda and ai no corrida: none notified
The Overspill: when there's more that I want to say 