Pantopia Reading Nook 📰🚩 – Telegram
Pantopia Reading Nook 📰🚩
509 subscribers
597 photos
3 videos
66 files
3.53K links
The news channel of the Pantopia Community. We publish articles, short essays, videos and all kinds of media around leftist theory.

Looking for books? Check out @pantopialibrary
Group chat: @pantopiagroup
Download Telegram
"A week and a half ago, Goldman Sachs put out a 31-page-report (noscriptd "Gen AI: Too Much Spend, Too Little Benefit?”) that includes some of the most damning literature on generative AI I've ever seen. And yes, that sound you hear is the slow deflation of the bubble I've been warning you about since March.

The report covers AI's productivity benefits (which Goldman remarks are likely limited), AI's returns (which are likely to be significantly more limited than anticipated), and AI's power demands (which are likely so significant that utility companies will have to spend nearly 40% more in the next three years to keep up with the demand from hyperscalers like Google and Microsoft)."

"The report includes an interview with economist Daron Acemoglu of MIT (page 4), an Institute Professor who published a paper back in May called "The Simple Macroeconomics of AI" that argued that "the upside to US productivity and, consequently, GDP growth from generative AI will likely prove much more limited than many forecasters expect.""

"The training data crisis is one that doesn’t get enough attention, but it’s sufficiently dire that it has the potential to halt (or dramatically slow) any AI development in the near future. As one paper, published in the journal Computer Vision and Pattern Recognition, found, in order to achieve a linear improvement in model performance, you need an exponentially large amount of data.

Or, put another way, each additional step becomes increasingly (and exponentially) more expensive to take. This infers a steep financial cost — not merely in just obtaining the data, but also the compute required to process it — with Anthropic CEO Dario Amodei saying that the AI models currently in development will cost as much as $1bn to train, and within three years we may see models that cost as much as “ten or a hundred billion” dollars, or roughly three times the GDP of Estonia"

"One particular myth Covello dispels is comparing generative AI "to the early days of the internet," noting that "even in its infancy, the internet was a low-cost technology solution that enabled e-commerce to replace costly incumbent solutions," and that "AI technology is exceptionally expensive, and to justify those costs, the technology must be able to solve complex problems, which it isn't designed to do.""

"In essence, on top of generative AI not having any killer apps, not meaningfully increasing productivity or GDP, not generating any revenue, not creating new jobs or massively changing existing industries, it also requires America to totally rebuild its power grid, which Janous regrettably adds the US has kind of forgotten how to do."

"The remaining defense is also one of the most annoying — that OpenAI has something we don't know about. A big, sexy, secret technology that will eternally break the bones of every hater.

Yet, I have a counterpoint: no it doesn't.

Seriously, Mira Murati, CTO of OpenAI, said a few weeks ago that the models it has in its labs are not much more advanced than those that are publicly-available."

https://www.wheresyoured.at/pop-culture/
The analysis, published Friday, is called “Consent in Crisis: The Rapid Decline of the AI Data Commons,” and has found that, in the last year, “there has been a rapid crescendo of data restrictions from web sources” restricting web scraper bots (sometimes called “user agents”) from training on their websites.

Specifically, about 5 percent of the 14,000 websites analyzed had modified their robots.txt file to block AI scrapers. That may not seem like a lot, but 28 percent of the “most actively maintained, critical sources,” meaning websites that are regularly updated and are not dormant, have restricted AI scraping in the last year. An analysis of these sites’ terms of service found that, in addition to robots.txt restrictions, many sites also have added AI scraping restrictions to their terms of service documents in the last year.

This change has happened almost entirely within the last year, the researchers found. In mid 2023, about 1 percent of websites in the researchers’ sample had fully restricted AI scraping. Now, they estimate about 5-7 percent do, and say that the number only captures “full restricted” domains meaning their robots.txt file does not allow for any AI scraping.

The study, led by Shayne Longpre of MIT and done in conjunction with a few dozen researchers at the Data Provenance Initiative, called this change an “emerging crisis” not just for commercial AI companies like OpenAI and Perplexity, but for researchers hoping to train AI for academic purposes. The New York Times said this shows that the data used to train AI is “disappearing fast.”

https://www.404media.co/the-backlash-against-ai-scraping-is-real-and-measurable/
According to a first of its kind study on “corporate carbon damages”, firms would lose 44% of their profits if they had to pay for the damages attributable to their climate pollution. When one of the researchers was asked what the total amount in dollars would be for such damages, co-author Christian Leuz revealed that “At $190 [the U.S. EPA’s current cost per ton of carbon], the utility industry averaged damages more than twice its profits. Materials manufacturing, energy and transportation industries all had average damages that exceeded their profits.” An additional analysis from 2013 that focused on pricing in environmental externalities came to a similar conclusion that does not bode well for “climate capitalism.” Influential climate analyst David Roberts, writing for Grist at the time, concludes: “Of the top 20 region-sectors ranked by environmental impacts, none would be profitable if environmental costs were fully integrated. Ponder that for a moment: None of the world’s top industrial sectors would be profitable if they were paying their full freight. Zero.”

https://sublationmedia.com/making-graphs-to-flatter-the-global-elite/
The Pentagon, Congress, the defense industry, think tanks, lobbyists, and industry-sponsored media outlets are all very real. When combined, they make up what is better termed the “National Security Establishment,” which Americans see in action all the time.

https://responsiblestatecraft.org/military-industrial-complex-2668809022/
Administrative detention in Isreal

Administrative detention is based on secret suspicions, secret evidence and no charges being brought. To conceal its inherent absurdity, hearings are held in-camera and away from the public eye. As such, even the little that is revealed to the defense remains prohibited for publication.

"Like all administrative detention hearings, it was held in-camera, to obscure the fact that detainees' lawyers do their job without access to the facts of the case. Even the few details that are not secret are prohibited for publication. The administrative detention order was approved in full for a period of six months"

" In the past, it was considered, at least officially, a measure reserved for the most extreme of cases. This hypocritical position has always been false, but now there is no longer any need to save face. According to the Israeli army's own data, almost 5,000 arrests were made in the West Bank in the past eight months. These are very conservative numbers, as they don't include the many thousands arrested and released without being indicted.

The data shows that administrative detention, this so-called extreme of extremes, is now the norm. According to Israeli Prison Service numbers, Israel now holds 7016 people who have not yet been convicted in its jails – either awaiting trial or under administrative detention. Of these, 4,299 – more than 60%! – are held without charge or trial. And all that is without saying a single word about the torture, hunger and humiliation to which all Palestinian prisoners held by Israel are subjected these days."
A large-scale nuclear power station in China is the first in the world to be completely impervious to dangerous meltdowns, even during a full loss of external power. The design can’t be adapted to existing nuclear reactors around the world, but could be a blueprint for future ones.

https://www.newscientist.com/article/2440388-chinese-nuclear-reactor-is-completely-meltdown-proof/