Recently science author Sam Kean reviewed In a Flight of Starlings, a "slender, uneven collection of essays by Giorgio Parisi about his life in physics, from his student days in Rome to the work that won him a share of the 2021 Nobel Prize in physics."

But the reviewer makes an interesting point: As someone who writes about science history, I have long grumbled about how misleading modern scientific papers are. I understand the need to present scientific findings in a clean, concise way, but the papers also omit all the false starts, blind alleys, broken equipment, and dumb mistakes that beset real scientific research every day. By omitting all the human stuff, the papers fail to explain how science really gets done. Parisi raises a related complaint — that scientific papers omit all sense of intuition. Indeed, the best sections of the book explore the role of intuition in scientific thinking.

He quotes a friend who says that "a good mathematician understands immediately which mathematical statements are true and which are false, whereas a bad mathematician has to try to prove them in order to know." The same applies to science: the early stages of any project are chaotic, and the data can be confusing and even contradictory. Scientists need intuition to cut through the mess and focus on the most promising explanations. Much of this intuition is unconscious and, while still grounded in physical brain processes, remains murky and hard to reconstruct. And for whatever reason, that vagueness makes scientists uncomfortable. "In almost all texts written by scientists," Parisi notes, "these themes are taboo."

So it's refreshing to see a scientist, especially one of Parisi's stature, honestly discuss the fuzzy side of scientific thinking, and not just during the early, groping stages but in the technical phases of a project, too. "The physicist sometimes uses mathematics ungrammatically," he admits, "a license that we grant to poets" as well... In a Flight of Starlings, Parisi writes, "is my attempt to convey to a wide readership something of the beauty, importance, and cultural value of modern science." Does he succeed? At times, yes... Perhaps it's not unlike the hodgepodge of science itself, then...

Cory Doctorow, author, digital rights advocate, and co-editor of the blog Boing Boing, has launched a Kickstarter campaign for his next book, called The Internet Con: How To Seize the Means of Computation. "The book presents an array of policy solutions aimed at dismantling the monopolistic power of Big Tech, making the internet a more open and user-focused space," writes Boing Boing's Mark Frauenfelder. "Key among these solutions is the concept of interoperability, which would allow users to take their apps, data, and content with them when they decide to leave a service, thus reducing the power of tech platforms." From Cory's Medium article announcing the Kickstarter: I won't sell my work with DRM, because DRM is key to the enshittification of the internet. Enshittification is why the old, good internet died and became "five giant websites filled with screenshots of the other four" (h/t Tom Eastman). When a tech company can lock in its users and suppliers, it can drain value from both sides, using DRM and other lock-in gimmicks to keep their business even as they grow ever more miserable on the platform.

Here is how platforms die: first, they are good to their users; then they abuse their users to make things better for their business customers; finally, they abuse those business customers to claw back all the value for themselves. Then, they die.

The Internet Con isn't just an analysis of where enshittification comes from: it's a detailed, shovel-ready policy prescription for halting enshittification, throwing it into reverse and bringing back the old, good internet.

By 1983 there were a whopping 2,000 pieces of software for Apple's pre-Macintosh computer, the Apple II — more than for any other machine in the world. It turns out this left a trail for one historian to understand The Apple II Age: How the Computer Became Personal.

The new book (by New York University academic Laine Nooney) argues that it was the first purchasers of that software who are the true overlooked pioneers during the seven years before the Macintosh. And (as this reviewer explains, with quotes from the book), collectively they form the most compelling story about the history of Apple: It's about all those brave and curious people, the users, who came "Not to hack, but to play... Not to program, but to print..." And you can trace their activities in perfect detail through the decades-old software programs they left behind. It's a fresh and original approach to the history of technology. Yes, the Apple II competed with Commodore's PET 2001 and Tandy's TRS-80... [But] this trove of programs uniquely offers "a glimpse of what users did with their personal computers, or perhaps more tellingly, what users hoped their computers might do."

Looking back in time, Nooney calls the period "one of unusually industrious and experimental software production, as mom-and-pop development houses cast about trying to create software that could satisfy the question, 'What is a computer even good for...?'" The book's jacket promises "a constellation of software creation stories," with each chapter revisiting an especially iconic program that also represents an entire category of software...

[T]he book ultimately focuses more heavily on the lessons that can be learned from what programmers envisioned for these strange new devices — and how the software-buying public did (or didn't) respond... The earliest emergence of personal computing in America was "a wondrous mangle," Nooney writes, saying it turned into an era where "overnight entrepreneurs hastily constructed a consumer computing supply chain where one had never previously existed."
Vice republished an excerpt in May which describes the "roiling debate" that took place over copy protection in 1981.

The New York Times points out that so-called "shadow libraries," like Library Genesis, Z-Library or Bibliotik, "are obscure repositories storing millions of titles, in many cases without permission — and are often used as A.I. training data." A.I. companies have acknowledged in research papers that they rely on shadow libraries. OpenAI's GPT-1 was trained on BookCorpus, which has over 7,000 unpublished titles scraped from the self-publishing platform Smashwords. To train GPT-3, OpenAI said that about 16 percent of the data it used came from two "internet-based books corpora" that it called "Books1" and "Books2." According to a lawsuit by the comedian Sarah Silverman and two other authors against OpenAI, Books2 is most likely a "flagrantly illegal" shadow library.

These sites have been under scrutiny for some time. The Authors Guild, which organized the authors' open letter to tech executives, cited studies in 2016 and 2017 that suggested text piracy depressed legitimate book sales by as much as 14 percent.

Efforts to shut down these sites have floundered. Last year, the F.B.I., with help from the Authors Guild, charged two people accused of running Z-Library with copyright infringement, fraud and money laundering. But afterward, some of these sites were moved to the dark web and torrent sites, making it harder to trace them. And because many of these sites are run outside the United States and anonymously, actually punishing the operators is a tall task.

Tech companies are becoming more tight-lipped about the data used to train their systems.

sciencehabit shares a report from Science: This week, the much anticipated movie Oppenheimer hits theaters, giving famed filmmaker Christopher Nolan's take on the theoretical physicist who during World War II led the Manhattan Project to develop the first atomic bomb. J. Robert Oppenheimer, who died in 1967, is known as a charismatic leader, eloquent public intellectual, and Red Scare victim who in 1954 lost his security clearance in part because of his earlier associations with suspected Communists. To learn about Oppenheimer the scientist, Science spoke with David C. Cassidy, a physicist and historian emeritus at Hofstra University. Cassidy has authored or edited 10 books, including J. Robert Oppenheimer and the American Century. How did Oppenheimer compare to Einstein? Did he actually make any substantiative contributions to THE Bomb? And why did he eventually lose his security clearance?

Artificial intelligence companies are exploring a new avenue to obtain the massive amounts of data needed to develop powerful generative models: creating the information from scratch. From a report: Microsoft, OpenAI and Cohere are among the groups testing the use of so-called synthetic data -- computer-generated information to train their AI systems known as large language models (LLMs) -- as they reach the limits of human-made data that can further improve the cutting-edge technology. The launch of Microsoft-backed OpenAI's ChatGPT last November has led to a flood of products rolled out publicly this year by companies including Google and Anthropic, which can produce plausible text, images or code in response to simple prompts.

The technology, known as generative AI, has driven a surge of investor and consumer interest, with the world's biggest technology companies including Google, Microsoft and Meta racing to dominate the space. Currently, LLMs that power chatbots such as OpenAI's ChatGPT and Google's Bard are trained primarily by scraping the internet. Data used to train these systems includes digitised books, news articles, blogs, search queries, Twitter and Reddit posts, YouTube videos and Flickr images, among other content. Humans are then used to provide feedback and fill gaps in the information in a process known as reinforcement learning by human feedback (RLHF). But as generative AI software becomes more sophisticated, even deep-pocketed AI companies are running out of easily accessible and high-quality data to train on. Meanwhile, they are under fire from regulators, artists and media organisations around the world over the volume and provenance of personal data consumed by the technology.

The Register shares its collection of "signs that Wayland is becoming the favored way to get a GUI on Linux." - The team developing Linux for Apple Silicon Macs said they didn't have the manpower to work on X.org support.

- A year ago, the developers of the Gtk toolkit used by many Linux apps and desktops said that the next version may drop support for X11...

- One of the developers of the Budgie desktop, Campbell Jones, recently published a blog post with a wildly controversial title that made The Reg FOSS desk smile: "Wayland is pretty good, actually." He lays out various benefits that Wayland brings to developers, and concludes: "Primarily, what I've learned is that Wayland is actually really well-designed. The writing is on the wall for X, and Wayland really is the future." Partly as a result of this, it looks likely that the next version of the Budgie desktop, Budgie 11, will only support Wayland, completely dropping support for X11. The team point out that this is not such a radical proposition: there was a proposal to make KDE 6 sessions default to Wayland as long ago as last October...

- The GNOME spin of Fedora has defaulted to Wayland since version 25 in 2017, and the GNOME flavor of Ubuntu since 21.04.

- [T]here's now an experimental effort to get Wayland working on OpenBSD. The effort happened at the recent OpenBSD hackathon in Tallinn, Estonia, and the developer's comments are encouraging. It's already available as part of FreeBSD.

The Internet Archive has taken the rather unusual step of sending a DMCA notice to protect the copyrights of book publishers and authors. The non-profit organization asked GitHub to remove a tool that can strip DRM from books in its library. The protective move is likely motivated by the ongoing legal troubles between the Archive and book publishers. TorrentFreak reports: The Internet Archive sent a takedown request to GitHub, requesting the developer platform to remove a tool that circumvents industry-standard technical protection mechanisms for digital libraries. This "DeGouRou" software effectively allows patrons to save DRM-free copies of the books they borrow. "This DMCA complaint is about a tool made available on github which purports to circumvent technical protections in violation of the copyright act section 1201," the notice reads. "I am reporting a Git which provides a tool specifically used to circumvent industry standard library TPMs which are used by Internet Archive, and other libraries, to permit patrons to borrow an encrypted book, read the encrypted book, and return an encrypted book."

Interestingly, an IA representative states that they are "not authorized by the copyright owners" to submit this takedown notice. Instead, IA is acting on its duty to prevent the unauthorized downloading of copyright-protected books. It's quite unusual to see a party sending takedown notices without permission from the actual rightsholders. However, given the copyright liabilities IA faces, it makes sense that the organization is doing what it can to prevent more legal trouble. Permission or not, GitHub honored the takedown request. It removed all the DeGourou repositories that were flagged and took the code offline. [...] After GitHub removed the code, it soon popped up elsewhere.

Popular Science describes how NASA's Volatiles Investigating Polar Exploration Rover (VIPER) will use a pair of ramps to become the first rover to explore the Moon's south pole when it arrives in late 2024. From the report: "We all know how to work with ramps, and we just need to optimize it for the environment we're going to be in," says NASA's VIPER program manager Daniel Andrews. A VIPER test vehicle recently descended down a pair of metal ramps at NASA's Ames Research Center in California, as seen in the agency's recently published photos, with one beam for each set of the rover's wheels. Because the terrain where VIPER will land -- the edge of the massive Nobile Crater -- is expected to be rough, the engineering team has been testing VIPER's ability to descend the ramps at extreme angles. They have altered the steepness, as measured from the lander VIPER will descend from, and differences in elevation between the ramp for each wheel. "We have two ramps, not just for the left and right wheels, but a ramp set that goes out the back too," Andrews says. "So we actually get our pick of the litter, which one looks most safe and best to navigate as we're at that moment where we have to roll off the lander."

VIPER is a scientific successor to NASA's Lunar Crater Observation and Sensing Satellite, or LCROSS mission, which in 2009 confirmed the presence of water ice on the lunar south pole. "It completely rewrote the books on the moon with respect to water," says Andrews, who also worked on the LCROSS mission. "That really started the moon rush, commercially, and by state actors like NASA and other space agencies." The ice, if abundant, could be mined to create rocket propellant. It could also provide water for other purposes at long-term lunar habitats, which NASA plans to construct in the late 2020s as part of the Artemis moon program.

But LCROSS only confirmed that ice was definitely present in a single crater at the moon's south pole. VIPER, a mobile rover, will probe the distribution of water ice in greater detail. Drilling beneath the lunar surface is one task. Another is to move into steep, permanently shadowed regions -- entering craters that, due to their sharp geometry, and the low angle of the sun at the lunar poles, have not seen sunlight in billions of years. The tests demonstrate the rover can navigate a 15-degree slope with ease -- enough to explore these hidden dark spots, avoiding the need to make a machine designed for trickier descents. "We think there's plenty of scientifically relevant opportunities, without having to make a superheroic rover that can do crazy things," Andrews says.

Developed by NASA Ames and Pittsburgh-based company Astrobotic, VIPER is a square golf-cart-sized vehicle about 5 feet long and wide, and about 8 feet high. Unlike all of NASA's Mars rovers, VIPER has four wheels, not six. "A problem with six wheels is it creates kind of the equivalent of a track, and so you're forced to drive in a certain way," Andrews says. VIPER's four wheels are entirely independent from each other. Not only can they roll in any direction, they can be turned out, using the rover's shoulder-like joints to crawl out of the soft regolith of the kind scientists believe exists in permanently shadowed moon craters. The wheels themselves are very similar to those on the Mars rovers, but with more paddle-like treads, known as grousers, to carry the robot through fluffy regolith. [...] Together with Astrobotic, Andrews and his team have altered the ramps, and they now include specialized etchings down their lengths. The rover can detect this pattern along the rampway, using cameras in its wheel wells. "By just looking down there," the robot knows where it is, he says. "That's a new touch." Andrews is sure VIPER will be ready for deployment in 2024, however many tweaks are necessary. After all, this method is less complicated than a sky crane, he notes: "Ramps are pretty tried and true."

Comedian Sarah Silverman and two authors have filed copyright infringement lawsuits against Meta and OpenAI for allegedly using their content without permission to train artificial intelligence language models. From a report: The proposed class action lawsuits filed by Silverman, Richard Kadrey and Christopher Golden in San Francisco federal court Friday allege Facebook parent company Meta and ChatGPT maker OpenAI used copyrighted material to train chat bots. The lawsuits underscore the legal risks developers of chat bots face when using troves of copyrighted material to create apps that deliver realistic responses to user prompts. Silverman, Kadrey and Golden allege Meta and OpenAI used their books without authorization to develop their so-called large language models, which their makers pitch as powerful tools for automating tasks by replicating human conversation. In their lawsuit against Meta, the plaintiffs allege that leaked information about the company's artificial intelligence business shows their work was used without permission.

Slashdot reader j3x0n shared this report from California newspaper the Sacramento Bee: In 2015, Democratic Elk Grove Assemblyman Jim Cooper voted for Senate Bill 34, which restricted law enforcement from sharing automated license plate reader (ALPR) data with out-of-state authorities. In 2023, now-Sacramento County Sheriff Cooper appears to be doing just that. The Electronic Frontier Foundation (EFF) a digital rights group, has sent Cooper a letter requesting that the Sacramento County Sheriff's Office cease sharing ALPR data with out-of-state agencies that could use it to prosecute someone for seeking an abortion.

According to documents that the Sheriff's Office provided EFF through a public records request, it has shared license plate reader data with law enforcement agencies in states that have passed laws banning abortion, including Alabama, Oklahoma and Texas. Adam Schwartz, EFF senior staff attorney, called automated license plate readers "a growing threat to everyone's privacy ... that are out there by the thousands in California..." Schwartz said that a sheriff in Texas, Idaho or any other state with an abortion ban on the books could use that data to track people's movements around California, knowing where they live, where they work and where they seek reproductive medical care, including abortions.

The Sacramento County Sheriff's Office isn't the only one sharing that data; in May, EFF released a report showing that 71 law enforcement agencies in 22 California counties — including Sacramento County — were sharing such data... [Schwartz] said that he was not aware of any cases where ALPR data was used to prosecute someone for getting an abortion, but added, "We think we shouldn't have to wait until the inevitable happens."
In May the EFF noted that the state of Idaho "has enacted a law that makes helping a pregnant minor get an abortion in another state punishable by two to five years in prison."

Lamborghini's combustion engine models are sold out until the end of production, its chief executive was quoted as saying in the WELT newspaper on Wednesday, as the luxury carmaker transitions towards a pure hybrid lineup. From a report: Order books for its Hurucan and Urus models are full, marking the end of combustion engine vehicle production for the company, Stephan Winkelmann, head of the Volkswagen subsidiary, said. Lamborghini announced last July it would be investing at least 1.8 billion euros ($2 billion) to produce a hybrid lineup by 2024 and more to bring out its fully electric model by the end of the decade.

Two U.S. authors have filed a proposed class action lawsuit against OpenAI, claiming that the company infringed their copyrights by using their works without permission to train its generative AI system, ChatGPT. The plaintiffs, Massachusetts-based writers Paul Tremblay and Mona Awad, claim the data used to train ChatGPT included thousands of books, including those from illegal "shadow libraries." Reuters reports: The complaint estimated that OpenAI's training data incorporated over 300,000 books, including from illegal "shadow libraries" that offer copyrighted books without permission. Awad is known for novels including "13 Ways of Looking at a Fat Girl" and "Bunny." Tremblay's novels include "The Cabin at the End of the World," which was adapted in the M. Night Shyamalan film "Knock at the Cabin" released in February.

Tremblay and Awad said ChatGPT could generate "very accurate" summaries of their books, indicating that they appeared in its database. The lawsuit seeks an unspecified amount of money damages on behalf of a nationwide class of copyright owners whose works OpenAI allegedly misused.

The website Goodreads has become an essential avenue for building readership, but the same features that help generate excitement can also backfire. The New York Times: Cecilia Rabess figured her debut novel, "Everything's Fine," would spark criticism: The story centers on a young Black woman working at Goldman Sachs who falls in love with a conservative white co-worker with bigoted views. But she didn't expect a backlash to strike six months before the book was published. In January, after a Goodreads user who had received an advanced copy posted a plot summary that went viral on Twitter, the review site was flooded with negative comments and one-star reviews, with many calling the book anti-Black and racist. Some of the comments were left by users who said they had never read the book, but objected to its premise.

"It may look like a bunch of one-star reviews on Goodreads, but these are broader campaigns of harassment," Rabess said. "People were very keen not just to attack the work, but to attack me as well." In an era when reaching readers online has become a near-existential problem for publishers, Goodreads has become an essential avenue for building an audience. As a cross between a social media platform and a review site like Yelp, the site has been a boon for publishers hoping to generate excitement for books. But the same features that get users talking about books and authors can also backfire. Reviews can be weaponized, in some cases derailing a book's publication long before its release. "It can be incredibly hurtful, and it's frustrating that people are allowed to review books this way if they haven't read them," said Roxane Gay, an author and editor who also posts reviews on Goodreads. "Worse, they're allowed to review books that haven't even been written. I have books on there being reviewed that I'm not finished with yet."

The average test scores for 13-year-old students in the U.S. have decreased in reading and math since 2020, reaching the lowest levels in decades, with more significant declines in math. NPR reports: The average scores, from tests given last fall, declined 4 points in reading and 9 points in math, compared with tests given in the 2019-2020 school year, and are the lowest in decades. The declines in reading were more pronounced for lower performing students, but dropped across all percentiles. The math scores were even more disappointing. On a scale of 500 points, the declines ranged from 6 to 8 points for middle and high performing students, to 12 to 14 points for low performing students.

The math results also showed widening gaps based on gender and race. Scores decreased by 11 points for female students over 2020 results, compared with a 7-point decrease for male students. Among Black students, math scores declined 13 points, while white students had a 6-point drop. Compared with the 35-point gap between Black and white students in 2020, the disparity widened to 42 points.

While the scores show a drop from the pre-pandemic years, the results also show that there are other factors at work. The decline is even more substantial when compared with scores of a decade ago: The average scores declined 7 points in reading and 14 points in mathematics. The Education Department says plans are underway to address the learning loss. [...] The latest results are from the NAEP Long-Term Trend Assessment, traditionally administered every four years by the National Center for Education Statistics.

Slashdot reader rtfa0987 spotted this on the front page of the San Jose Mercury News. "Silicon Valley is poised once again to cash in on other people's products, making a data grab of unprecedented scale that has already spawned lawsuits and congressional hearings. Chatbots and other forms of generative artificial intelligence that burst onto the technology scene in recent months are fed vast amounts of material scraped from the internet — books, screenplays, research papers, news stories, photos, art, music, code and more — to produce answers, imagery or sound in response to user prompts... But a thorny, contentious and highly consequential issue has arisen: A great deal of the bots' fodder is copyrighted property...

The new AI's intellectual-property problem goes beyond art into movies and television, photography, music, news media and computer coding. Critics worry that major players in tech, by inserting themselves between producers and consumers in commercial marketplaces, will suck out the money and remove financial incentives for producing TV scripts, artworks, books, movies, music, photography, news coverage and innovative software. "It could be catastrophic," said Danielle Coffey, CEO of the News/Media Alliance, which represents nearly 2,000 U.S. news publishers, including this news organization. "It could decimate our industry."

The new technology, as happened with other Silicon Valley innovations, including internet-search, social media and food delivery, is catching on among consumers and businesses so quickly that it may become entrenched — and beloved by users — long before regulators and lawmakers gather the knowledge and political will to impose restraints and mitigate harms. "We may need legislation," said Congresswoman Zoe Lofgren, D-San Jose, who as a member of the House Judiciary Committee heard testimony on copyright and generative AI last month. "Content creators have rights and we need to figure out a way how those rights will be respected...."

Furor over the content grabbing is surging. Photo-sales giant Getty is also suing Stability AI. Striking Hollywood screenwriters last month raised concerns that movie studios will start using chatbot-written scripts fed on writers' earlier work. The record industry has lodged a complaint with federal authorities over copyrighted music being used to train AI.
The article includes some unique perspectives:

• There's a technical solution being proposed by the software engineer-CEO of Dazzle Labs, a startup building a platform for controlling personal data. The Mercury News summarizes it as "content producers could annotate their work with conditions for use that would have to be followed by companies crawling the web for AI fodder."

• Santa Clara University law school professor Eric Goldman "believes the law favors use of copyrighted material for training generative AI. 'All works build upon precedent works. We are all free to take pieces of precedent works. What generative AI does is accelerate that process, but it's the same process. It's all part of an evolution of our society's storehouse of knowledge...."

In San Francisco's District 8, a public library has turned off its Wi-Fi outside of business hours in response to complaints from neighbors and the city supervisor's office about open drug use and disturbances caused by unhoused individuals. The Verge reports: In San Francisco's District 8, a public library has been shutting down Wi-Fi outside business hours for nearly a year. The measure, quietly implemented in mid-2022, was made at the request of neighbors and the office of city supervisor Rafael Mandelman. It's an attempt to keep city dwellers who are currently unhoused away from the area by locking down access to one of the library's most valuable public services. A local activist known as HDizz revealed details behind the move last month, tweeting public records of a July 2022 email exchange between local residents and the city supervisor's office. In the emails, residents complained about open drug use and sidewalks blocked by residents who are unhoused. One relayed a secondhand story about a library worker who had been followed to her car. And by way of response, they demanded the library limit the hours Wi-Fi was available. "Why are the vagrants and drug addicts so attracted to the library?" one person asked rhetorically. "It's the free 24/7 wi-fi."

San Francisco's libraries have been historically progressive when it comes to providing resources to people who are unhoused, even hiring specialists to offer assistance. But on August 1st, reports San Francisco publication Mission Local, city librarian Michael Lambert met with Mandelman's office to discuss the issue. The next day, District 8's Eureka Valley/Harvey Milk Memorial branch began turning its Wi-Fi off after hours -- a policy that San Francisco Public Library (SFPL) spokesperson Jaime Wong told The Verge via email remains in place today.

In the initial months after the decision, the library apparently received no complaints. But in March, a little over seven months following the change, it got a request to reverse the policy. "I'm worried about my friend," the email reads, "whom I am trying to get into long term residential treatment." San Francisco has shelters, but the requester said their friend had trouble communicating with the staff and has a hard time being around people who used drugs, among other issues. Because this friend has no regular cell service, "free wifi is his only lifeline to me [or] for that matter any services for crisis or whatever else." The resident said some of the neighborhood's residents "do not understand what they do to us poor folks nor the homeless by some of the things they do here." Jennifer Friedenbach of San Francisco's Coalition on Homelessness told The Verge in a phone interview that "folks are not out there on the streets by choice. They're destitute and don't have other options. These kinds of efforts, like turning off the Wi-Fi, just exacerbate homelessness and have the opposite effect. Putting that energy into fighting for housing for unhoused neighbors would be a lot more effective."

schwit1 shares a report from VentureBeat: [A]s those following the burgeoning industry and its underlying research know, the data used to train the large language models (LLMs) and other transformer models underpinning products such as ChatGPT, Stable Diffusion and Midjourney comes initially from human sources -- books, articles, photographs and so on -- that were created without the help of artificial intelligence. Now, as more people use AI to produce and publish content, an obvious question arises: What happens as AI-generated content proliferates around the internet, and AI models begin to train on it, instead of on primarily human-generated content?

A group of researchers from the UK and Canada have looked into this very problem and recently published a paper on their work in the open access journal arXiv. What they found is worrisome for current generative AI technology and its future: "We find that use of model-generated content in training causes irreversible defects in the resulting models." Specifically looking at probability distributions for text-to-text and image-to-image AI generative models, the researchers concluded that "learning from data produced by other models causes model collapse -- a degenerative process whereby, over time, models forget the true underlying data distribution ... this process is inevitable, even for cases with almost ideal conditions for long-term learning."

"Over time, mistakes in generated data compound and ultimately force models that learn from generated data to misperceive reality even further," wrote one of the paper's leading authors, Ilia Shumailov, in an email to VentureBeat. "We were surprised to observe how quickly model collapse happens: Models can rapidly forget most of the original data from which they initially learned." In other words: as an AI training model is exposed to more AI-generated data, it performs worse over time, producing more errors in the responses and content it generates, and producing far less non-erroneous variety in its responses. As another of the paper's authors, Ross Anderson, professor of security engineering at Cambridge University and the University of Edinburgh, wrote in a blog post discussing the paper: "Just as we've strewn the oceans with plastic trash and filled the atmosphere with carbon dioxide, so we're about to fill the Internet with blah. This will make it harder to train newer models by scraping the web, giving an advantage to firms which already did that, or which control access to human interfaces at scale. Indeed, we already see AI startups hammering the Internet Archive for training data." schwit1 writes: "Garbage in, garbage out -- and if this paper is correct, generative AI is turning into the self-licking ice cream cone of garbage generation."

A company called Sol Reader is working on a headset designed exclusively for reading books. "The device is simple: It slips over your eyes like a pair of glasses and blocks all distractions while reading," reports TechCrunch. From the article: The $350 device is currently on pre-order, comes in a handful of colors, and contains a pair of side-lit, e-ink displays, much like the Kindle does. The glasses come with a remote (I wish my Kindle had a remote!) and a charger. A full battery gets you around 25 hours of reading. That may not sound like a lot, but if you have an average adult reading speed of around 200 words per minute, you can finish the 577,608-word tome Infinite Jest in about 48 hours. That means you need at least one charging break, but then, if you are trying to read Infinite Jest in a single sitting, you're a bigger book nerd than most.

The product has a diopter adjustment built in, so glasses- and contacts-wearers can use the glasses without wearing additional vision correction (up to a point -- the company doesn't specify the exact adjustment range). The displays are 1.3-inch, e-ink displays with 256x256 per-eye resolution. The glasses have 64MB of storage, which should hold plenty of books for even the longest of escapist holidays.

The company's $5 million funding round was led by Garry Tan (Initialized, Y Combinator) and closed about a year ago. Today, the company is shipping the 'advanced copy' (read: private beta) of the glasses to a small number of early access testers. The company is tight-lipped on when its full production batches will start shipping, and customers are currently advised to join the waiting list if they want to get their mittens on a pair of Sols.

Owen Gingerich, a renowned astronomer and historian of science, has passed away at the age of 93. Gingerich dedicated years to tracking down 600 copies of Nicolaus Copernicus's influential book "De Revolutionibus Orbium Coelestium Libri Sex" and was known for his passion for astronomy, often dressing up as a 16th-century scholar for lectures. He believed in the compatibility of religion and science and explored this theme in his books "God's Universe" and "God's Planet." The New York Times reports: Professor Gingerich, who lived in Cambridge, Mass., and taught at Harvard for many years, was a lively lecturer and writer. During his decades of teaching astronomy and the history of science, he would sometimes dress as a 16th-century Latin-speaking scholar for his classroom presentations, or convey a point of physics with a memorable demonstration; for instance, The Boston Globe related in 2004, he "routinely shot himself out of the room on the power of a fire extinguisher to prove one of Newton's laws." He was nothing if not enthusiastic about the sciences, especially astronomy. One year at Harvard, when his signature course, "The Astronomical Perspective," wasn't filling up as fast as he would have liked, he hired a plane to fly a banner over the campus that read: "Sci A-17. M, W, F. Try it!"

Professor Gingerich's doggedness was on full display in his long pursuit of copies of Copernicus's "De Revolutionibus Orbium Coelestium Libri Sex" ("Six Books on the Revolutions of the Heavenly Spheres"), first published in 1543, the year Copernicus died. That book laid out the thesis that Earth revolved around the sun, rather than the other way around, a profound challenge to scientific knowledge and religious belief in that era. The writer Arthur Koestler had contended in 1959 that the Copernicus book was not read in its time, and Professor Gingerich set out to determine whether that was true. In 1970 he happened on a copy of "De Revolutionibus" that was heavily annotated in the library of the Royal Observatory in Edinburgh, suggesting that at least one person had read it closely. A quest was born. Thirty years and hundreds of thousands of miles later, Professor Gingerich had examined some 600 Renaissance-era copies of "De Revolutionibus" all over the world and had developed a detailed picture not only of how thoroughly the work was read in its time, but also of how word of its theories spread and evolved. He documented all this in "The Book Nobody Read: Chasing the Revolutions of Nicolaus Copernicus" (2004). John Noble Wilford, reviewing it in The New York Times, called "The Book Nobody Read" "a fascinating story of a scholar as sleuth."

Professor Gingerich was raised a Mennonite and was a student at Goshen College, a Mennonite institution in Indiana, studying chemistry but thinking of astronomy, when, he later recalled, a professor there gave him pivotal advice: "If you feel a calling to pursue astronomy, you should go for it. We can't let the atheists take over any field." He took the counsel, and throughout his career he often wrote or spoke about his belief that religion and science need not be at odds. He explored that theme in the books "God's Universe" (2006) and "God's Planet" (2014). He was not a biblical literalist; he had no use for those who ignored science and proclaimed the Bible's creation story historical fact. Yet, as he put it in "God's Universe," he was "personally persuaded that a superintelligent Creator exists beyond and within the cosmos." [...] Professor Gingerich, who was senior astronomer emeritus at the Smithsonian Astrophysical Observatory, wrote countless articles over his career in addition to his books. In one for Science and Technology News in 2005, he talked about the divide between theories of atheistic evolution and theistic evolution. "Frankly it lies beyond science to prove the matter one way or the other," he wrote. "Science will not collapse if some practitioners are convinced that occasionally there has been creative input in the long chain of being." In 2006, Gingerich was mentioned in a Slashdot story about geologists' reacting to the new definition of "Pluton." He was quoted as saying that he was only peripherally aware of the definition, and because it didn't show up on MS Word's spell check, he didn't think it was that important."

"Gingerich lead a committee of the International Astronomical Union charged with recommending whether Pluto should remain a planet," notes the New York Times. "His panel recommended that it should, but the full membership rejected that idea and instead made Pluto a 'dwarf planet.' That decision left Professor Gingerich somehwat dismayed."