Technology

No matter how much you dress up whatever AI service has gaslit you into believing it’s sentient, generative AI is inherently limited, impossibly expensive and economically unviable. Its services cost too much to run, its progenitors have no path to profitability, and no amount of rigged benchmarks and anecdotal examples of theoretical engineering teams that are “10x’d” can make up for the fact that you can’t measure the cost of an LLM-driven task or its return on investment.

Anyone claiming that you have to “measure AI’s ROI differently” is attempting to con either you or themselves. While it’s tough to measure the ROI of a particular worker or project, most workers and projects don’t increase your operating expenses by anywhere from 10% to 100% under the vaguest of promises that you might be “doing the future.” AI is calamitously expensive and, despite years of promises of it getting cheaper for both those running AI services and its customers, costs have only ever increased.

I think that’s by design. AI labs want their costs to be high so that they can continue growing at ridiculous rates, all so that they can keep feeding money to their hyperscaler compute partners who then invest that money right back into them, creating further reasons to keep buying NVIDIA GPUs, so that NVIDIA can then invest that money back into either AI compute providers (who OpenAI and Anthropic pay) or the AI labs themselves.

Concepts like “efficiency” or “cost reduction” run counter to the greater narrative of AI’s voracious sprawl of data center capex and still-theoretical AI revenue. If OpenAI or Anthropic were to seek profitability or sustainability (assuming these things were possible), that would create less demand for AI compute, which would mean less demand for Azure or Google Cloud or Amazon Web Services or CoreWeave or Oracle Cloud Infrastructure, which would in turn mean less demand for NVIDIA GPUs.

The problem with this marvelous plan is that at some point there had to be an honest transaction — real, honest, sustainable demand based on a reliable product that people liked paying for because they understood its value. Right now, AI revenues are either chaotically experimental or so thoroughly-subsidized that labs are giving away hundreds of dollars a user in the hopes that at some point said user might want to pay even more money for measurably less value, the kind of proposition you make when you think your customers are fucking idiots.

Earlier this year Nieman Lab broke the story that major news publishers, including The New York Times, The Guardian, and USA Today Co., had started blocking the Internet Archive for fear that AI companies might scrape the nonprofit’s repositories for training data. As one of the last bastions of archival history, that is, in case you’re not aware, not very good for the public interest.

Four months later and Nieman Lab now notes that the number of news outlets blocking the archive has soared to around 340 organizations:

“Our new analysis shows that more than 340 local news sites across the United States are now limiting the Internet Archive’s ability to access and preserve their stories. Many sites in our sample are owned by five of the seven largest local news publishers in the country: USA Today Co., McClatchy, Advance Local, MediaNews Group, and Tribune Publishing. The latter two are both subsidiaries of the “vulture hedge fund” Alden Global Capital.”

Many of these localities are already effectively news deserts, where most real local journalism was hollowed out and replaced by a smattering of local right wing broadcasters (like Sinclair Broadcasting) or a hedge fund run “local newspaper” that doesn’t do much in the way of actual local reporting. That’s generally also been terrible for informed consensus or shedding a light on local corruption.

Most of the work I've published has been removed from the web. Archival services help us retain history in an era where the physical local paper is not archived by libraries.

How about creating a new protocol for a project ?

The most honest thing you can say about violence is that nobody wants it, but the conditions that produce it are being engineered with extraordinary efficiency by people who have apparently never opened a history book.

Over the decades, there has been no shortage of sites using clever techniques to covertly track visitors’ browsing histories, device fingerprints, and keystrokes and mouse movements in real time. Even Meta and Yandex were recently caught joining in the privacy-invasive free-for-all.

Now sites have a new way to spy on their visitors: measuring subtle interactions with their solid-state drives. The technique, named FROST (fingerprinting remotely using OPFS-based SSD timing), allows sites to monitor other sites a visitor is viewing and what apps are open on their devices.

The technique, laid out in a research paper, exploits a side channel, a form of leak resulting from physical manifestations such as electromagnetic emanations, data caches, or the time required to complete a task. By measuring the manifestations, attackers can decrypt encrypted traffic and infer other confidential data.

The attack that FROST uses is known as a contention side channel, which measures the interaction of various processes all using (or competing for) a given resource. By measuring the timing of certain I/O (input-output) operations of the SSD a visitor is using, the researchers were able to determine the websites open in other tabs—even on other browsers—and the apps that were open on the visitor’s device. FROST requires no interaction from the visitor other than opening the site hosting the attack.

But I'm sure they'll still charge advertisers per impression sted conversion. Enshittification at its finest.

Valnet, the self-proclaimed “leading digital media investment company” behind the likes of Polygon, GameRant, OpenCritic, Collider, and over a dozen more gaming, technology, and lifestyle websites, isn’t known for paying its freelance writers well. But new “Pay Per Session” contracts issued to writers and editors at TheGamer on May 21 threaten to break new ground when it comes to click-mill-style exploitation.

According to details outlined in TheGamer’s Slack, these new terms, mandated by Valnet management, would “reward top-performing articles.” However, Kotaku last learned, based on the details outlined in the newly issued contracts, that the proposed “performance structure” would also result in writers not being paid at all if their articles do not exceed a minimum threshold of views.

Valnet was founded by Hassan Youssef and Sam Youssef, who were previously the owners of Canadian pornographic production company Brazzers and the “silent partners” of Pornhub, in 2012. The company has since earned a negative reputation over the years for the way it treats staff and freelance writers, with one former Collider employee describing it as “a content mill, borderline like almost sweatshop-level” during an interview with TheWrap in 2025 (I briefly freelanced for TheGamer from 2022 to 2023).

Robots have taken over Los Angeles.

It’s not just the AI-generated videos that have caused angst in Hollywood. Our streets are full of driverless Waymo vehicles, covered in more sensors and gadgets than the Batmobile. And our walkways are home to fleets of boxes on wheels, hurrying past pedestrians and navigating outdoor bar-hoppers as the robots deliver smoothies and keto-friendly salads.

And it’s only getting stranger. This month, Serve Robotics, one of the leading companies behind the food-delivery bots, deployed another 500 of them in 40 neighborhoods across the city, up from two neighborhoods in 2023. The other big company, Coco Robotics, founded at UCLA in 2020, has about 300 robots across the city and is looking to expand. Soon a region already known for its lack of walkability will have more obstacles for pedestrians to contend with.

The expansion has sparked consternation in LA and other US cities as residents debate whether our new neighbors are welcome. Neighboring Glendale is considering a moratorium on the bots; Chicago has also limited their expansion. Worse than the sidewalk frustrations, they mean fewer jobs for delivery drivers, even if some are human-controlled.

via

I'm incredibly pleased to announce that the microcode for the Intel 80386 has been decoded.

It was a group effort by a bunch of talented people to extract and correct the physical bits, but the major work of decoding them was done by reenigne - you may know him from such incredible PC demos as 8088 MPH and Area 5150, as well as being the person who decoded the 8088 microcode previously.

Please, check out his writeup.

cross-posted from: https://piefed.world/c/tech/p/1144611/google-disregard-right-now-if-you-want-to-see-where-ai-overviews-fall-short-similar-comm

For comparison, Bing and Kagi responses

Sources

• The Verge;
• Tech Crunch;
• Mac Rumors;
• Business Insider.

Samsung Electronics narrowly averted a walkout by nearly 48,000 workers this week, after executives agreed to a tentative deal over bonus payments. But the labor union’s demand for a bigger share of profits from the company’s semiconductor business has sparked questions — in South Korea and elsewhere — about who benefits from the AI industry, and whether its rewards should be shared more equitably.

Samsung, the world’s biggest memory chip maker, has reported record profits in recent months amid a global shortage of memory chips. The labor union had demanded the company allocate 15% of operating profit to bonuses for all workers, not just those at the memory chip division that supplies Tesla, Nvidia, and other big tech companies.

“As the AI industry drives record operating profits, union members are in a structure where they cannot receive the performance-based rewards they deserve,” Choi Seung-ho, head of Samsung’s union, told Rest of World. “We want to change that.”

Game 7 in the NBA playoffs: a chance to kick back, enjoy the drama of a winner-takes-all shootout between basketball’s big beasts, and … switch over from your regular TV provider to Amazon Prime? The excitement drains from the occasion at the first touch of the remote. Amazon no doubt imagined it had landed a real coup when the Eastern Conference semi-final series between Detroit and Cleveland extended to its maximum length, thereby handing the retail giant’s streaming arm, Prime Video, the right to air a Game 7 in the first season of its partnership with the NBA. In the event, Sunday’s game was a dud: a blowout win for the Cavs, playing on the road, that had all the electricity and charm of a stint in the doctor’s waiting room. Fortunately for viewers, Prime Video did its best to match the moment by producing a broadcast that was every bit as dull and juiceless as events on the court.

The pre-tipoff highlight was an interview with Shai Gilgeous-Alexander, on the occasion of his coronation as this season’s MVP, in which the Oklahoma City star appeared to be speaking from a movie theater for some reason. Blake Griffin, the house beefcake on Prime Video’s studio set, chided ESPN insider Shams Charania for leaking this year’s MVP announcement hours earlier: “It’s Sunday, Shams – go to brunch, you nerd.” If Hillary had won and Shams had kept his trap shut, we’d all be at brunch! The game got under way, and things did not improve. During the half-time show, Dirk Nowitzki rambled Germanly about various topics, while fellow former MVP Steve Nash delivered lines like “That decisiveness in isolation is so important” with all the conviction of a hostage recording a ransom video. Host Taylor Rooks tried valiantly to compensate for the lack of chemistry on set by laughing at even the slightest hint of a joke from any of her panellists. Awkward laughter delivered over dead air on a platform it feels like a punishment to access: that’s the Prime Video NBA playoffs guarantee.

These have been a difficult debut playoffs for Prime as it muscles in on the broadcast territory once ruled by what the media analysts call “linear TV”. The feed dropped out for several minutes during overtime in the play-in game between the Hornets and the Heat; buffering, the nightmare we all thought we outlived in 2006, has plagued the stream in several games; and video has frequently been mistimed with audio, producing delays and mismatches. There’s primetime, which is when the bulk of these playoffs are taking place, and then there’s Prime Video time, which comes in around three seconds later. The audio itself in many games has often, in my experience at least, been strangely soft, requiring a trip all the way to the top of the volume scale to hear what the analysts and announcers are saying.

github.com/select/movies-delux…

As Meta races to recenter itself around artificial intelligence, the tech giant is mandating that more than 7,000 workers must move to new teams, and it’s radically changing some employees’ jobs. The Guardian has also learned that some of these reassigned employees will shift to two new teams: one building AI cloud infrastructure and another that’s building an internal AI agent codenamed Hatch.

Late last week, Meta employees received a notice that engineers had been “selected” for reassignment and would begin reporting to the cloud infrastructure and Hatch teams by the end of this week. Meta made a similar move last month when it reshuffled at least 1,000 engineers on to a new data labeling team called Applied AI, or AAI – at first giving them the option to volunteer, but later telling workers: “Transfers aren’t optional.”

“Our work, infrastructure and our products are fundamentally changing as a result of the continued acceleration of AI,” wrote Peter Hoose, vice-president of production engineering at Meta, in an internal post about the two new teams viewed by the Guardian. “The pace of what we are building is unprecedented, and these are exactly the kind of challenges that define what we do best.”

A Meta employee referenced last month’s reshuffle in a comment on Hoose’s announcement, writing: “Does ‘selected’ imply this is an [Applied AI]-style draft rather than a voluntary move?”

Further proof the Meta doesn't know what the Zuck it's doing. The rebrand has gone swimmingly, given all the time we spend in the legless metaverse.

Intel ME and AMD PSP: The silicon layer nobody certifies

About cloud sovereignty and the often-ignored and unknown on-CPU management engine running below the OS and BIOS.

The article is quite long; it explains how CPUs run firmware that can include remote management over the network, and can be running even when the OS is not. They can be vulnerable to supply chain attacks and firmware replacements. Because it's on hardware, the firmware with open security vulnerabilities is often not updated.

Regarding cloud, the French SecNumCloud is a framework for cloud infrastructure security requirements. It doesn't cover these hardware attack vectors specifically but may mitigate risks through surrounding practices and isolation.

In conclusion, even a cloud provider that meets SecNumCloud must be asked whether and how they manage CPU management engine attack vectors.