Mike M. – Page 5

Measles outbreak hits 208 cases as federal response goes off the rails

autism, health, Infectious disease, measles, MMR, New Mexico, Science, texas, vaccination / Mike M. / March 8, 2025

Vitamin A is a fat-soluble vitamin that stays in the body. Taking too much over longer periods can cause vomiting, headache, fatigue, joint and bone pain, blurry vision, and skin and hair problems. Further, it can lead to dangerously high pressure inside the skull that pushes on the brain, as well as liver damage, confusion, coma, and other problems, according to the American Academy of Pediatrics.

Nevertheless, in an interview with Fox News this week, Kennedy endorsed an unconventional regimen of a steroid, an antibiotic and cod liver oil, praising two Texas doctors for giving it to patients. One of the doctors Kennedy championed was disciplined by the state medical board in 2003 for “unusual use of risk-filled medications,” according to a report by CNN.

In a yet more worrying sign, Reuters reported Friday afternoon that the CDC is planning to conduct a large study on whether the MMR vaccine is linked to autism. This taxpayer-funded effort would occur despite the fact that decades of research and numerous high-quality studies have already been conducted—and they have consistently disproven or found no connection between the vaccine and autism.

The agency’s move is exactly what Democratic senators feared when Kennedy was confirmed as the country’s top health official. In Senate hearings, Kennedy refused to say that vaccines do not cause autism. Democratic senators quickly warned that his anti-vaccine stance could not only move the country backward in the fight against vaccine-preventable diseases, but also hold back autism research aimed at finding the real cause(s) as well as better treatments.

“When you continue to sow doubt about settled science it makes it impossible for us to move forward,” Senator Maggie Hassan (D-N.H.) said in a Senate hearing. “It’s the relitigating and rehashing … it freezes us in place.”

Measles outbreak hits 208 cases as federal response goes off the rails Read More »

CMU research shows compression alone may unlock AI puzzle-solving abilities

AI, AI benchmarks, AI compression, ARC-AGI, Biz & IT, compression, François Chollet, machine learning, Tech / Mike M. / March 7, 2025

Tis the season for a squeezin’

New research challenges prevailing idea that AI needs massive datasets to solve problems.

A pair of Carnegie Mellon University researchers recently discovered hints that the process of compressing information can solve complex reasoning tasks without pre-training on a large number of examples. Their system tackles some types of abstract pattern-matching tasks using only the puzzles themselves, challenging conventional wisdom about how machine learning systems acquire problem-solving abilities.

“Can lossless information compression by itself produce intelligent behavior?” ask Isaac Liao, a first-year PhD student, and his advisor Professor Albert Gu from CMU’s Machine Learning Department. Their work suggests the answer might be yes. To demonstrate, they created CompressARC and published the results in a comprehensive post on Liao’s website.

The pair tested their approach on the Abstraction and Reasoning Corpus (ARC-AGI), an unbeaten visual benchmark created in 2019 by machine learning researcher François Chollet to test AI systems’ abstract reasoning skills. ARC presents systems with grid-based image puzzles where each provides several examples demonstrating an underlying rule, and the system must infer that rule to apply it to a new example.

For instance, one ARC-AGI puzzle shows a grid with light blue rows and columns dividing the space into boxes. The task requires figuring out which colors belong in which boxes based on their position: black for corners, magenta for the middle, and directional colors (red for up, blue for down, green for right, and yellow for left) for the remaining boxes. Here are three other example ARC-AGI puzzles, taken from Liao’s website:

Three example ARC-AGI benchmarking puzzles. Credit: Isaac Liao / Albert Gu

The puzzles test capabilities that some experts believe may be fundamental to general human-like reasoning (often called “AGI” for artificial general intelligence). Those properties include understanding object persistence, goal-directed behavior, counting, and basic geometry without requiring specialized knowledge. The average human solves 76.2 percent of the ARC-AGI puzzles, while human experts reach 98.5 percent.

OpenAI made waves in December for the claim that its o3 simulated reasoning model earned a record-breaking score on the ARC-AGI benchmark. In testing with computational limits, o3 scored 75.7 percent on the test, while in high-compute testing (basically unlimited thinking time), it reached 87.5 percent, which OpenAI says is comparable to human performance.

CompressARC achieves 34.75 percent accuracy on the ARC-AGI training set (the collection of puzzles used to develop the system) and 20 percent on the evaluation set (a separate group of unseen puzzles used to test how well the approach generalizes to new problems). Each puzzle takes about 20 minutes to process on a consumer-grade RTX 4070 GPU, compared to top-performing methods that use heavy-duty data center-grade machines and what the researchers describe as “astronomical amounts of compute.”

Not your typical AI approach

CompressARC takes a completely different approach than most current AI systems. Instead of relying on pre-training—the process where machine learning models learn from massive datasets before tackling specific tasks—it works with no external training data whatsoever. The system trains itself in real-time using only the specific puzzle it needs to solve.

“No pretraining; models are randomly initialized and trained during inference time. No dataset; one model trains on just the target ARC-AGI puzzle and outputs one answer,” the researchers write, describing their strict constraints.

When the researchers say “No search,” they’re referring to another common technique in AI problem-solving where systems try many different possible solutions and select the best one. Search algorithms work by systematically exploring options—like a chess program evaluating thousands of possible moves—rather than directly learning a solution. CompressARC avoids this trial-and-error approach, relying solely on gradient descent—a mathematical technique that incrementally adjusts the network’s parameters to reduce errors, similar to how you might find the bottom of a valley by always walking downhill.

A block diagram of the CompressARC architecture, created by the researchers. Credit: Isaac Liao / Albert Gu

The system’s core principle uses compression—finding the most efficient way to represent information by identifying patterns and regularities—as the driving force behind intelligence. CompressARC searches for the shortest possible description of a puzzle that can accurately reproduce the examples and the solution when unpacked.

While CompressARC borrows some structural principles from transformers (like using a residual stream with representations that are operated upon), it’s a custom neural network architecture designed specifically for this compression task. It’s not based on an LLM or standard transformer model.

Unlike typical machine learning methods, CompressARC uses its neural network only as a decoder. During encoding (the process of converting information into a compressed format), the system fine-tunes the network’s internal settings and the data fed into it, gradually making small adjustments to minimize errors. This creates the most compressed representation while correctly reproducing known parts of the puzzle. These optimized parameters then become the compressed representation that stores the puzzle and its solution in an efficient format.

An animated GIF showing the multi-step process of CompressARC solving an ARC-AGI puzzle. Credit: Isaac Liao

“The key challenge is to obtain this compact representation without needing the answers as inputs,” the researchers explain. The system essentially uses compression as a form of inference.

This approach could prove valuable in domains where large datasets don’t exist or when systems need to learn new tasks with minimal examples. The work suggests that some forms of intelligence might emerge not from memorizing patterns across vast datasets, but from efficiently representing information in compact forms.

The compression-intelligence connection

The potential connection between compression and intelligence may sound strange at first glance, but it has deep theoretical roots in computer science concepts like Kolmogorov complexity (the shortest program that produces a specified output) and Solomonoff induction—a theoretical gold standard for prediction equivalent to an optimal compression algorithm.

To compress information efficiently, a system must recognize patterns, find regularities, and “understand” the underlying structure of the data—abilities that mirror what many consider intelligent behavior. A system that can predict what comes next in a sequence can compress that sequence efficiently. As a result, some computer scientists over the decades have suggested that compression may be equivalent to general intelligence. Based on these principles, the Hutter Prize has offered awards to researchers who can compress a 1GB file to the smallest size.

We previously wrote about intelligence and compression in September 2023, when a DeepMind paper discovered that large language models can sometimes outperform specialized compression algorithms. In that study, researchers found that DeepMind’s Chinchilla 70B model could compress image patches to 43.4 percent of their original size (beating PNG’s 58.5 percent) and audio samples to just 16.4 percent (outperforming FLAC’s 30.3 percent).

That 2023 research suggested a deep connection between compression and intelligence—the idea that truly understanding patterns in data enables more efficient compression, which aligns with this new CMU research. While DeepMind demonstrated compression capabilities in an already-trained model, Liao and Gu’s work takes a different approach by showing that the compression process can generate intelligent behavior from scratch.

This new research matters because it challenges the prevailing wisdom in AI development, which typically relies on massive pre-training datasets and computationally expensive models. While leading AI companies push toward ever-larger models trained on more extensive datasets, CompressARC suggests intelligence emerging from a fundamentally different principle.

“CompressARC’s intelligence emerges not from pretraining, vast datasets, exhaustive search, or massive compute—but from compression,” the researchers conclude. “We challenge the conventional reliance on extensive pretraining and data, and propose a future where tailored compressive objectives and efficient inference-time computation work together to extract deep intelligence from minimal input.”

Limitations and looking ahead

Even with its successes, Liao and Gu’s system comes with clear limitations that may prompt skepticism. While it successfully solves puzzles involving color assignments, infilling, cropping, and identifying adjacent pixels, it struggles with tasks requiring counting, long-range pattern recognition, rotations, reflections, or simulating agent behavior. These limitations highlight areas where simple compression principles may not be sufficient.

The research has not been peer-reviewed, and the 20 percent accuracy on unseen puzzles, though notable without pre-training, falls significantly below both human performance and top AI systems. Critics might argue that CompressARC could be exploiting specific structural patterns in the ARC puzzles that might not generalize to other domains, challenging whether compression alone can serve as a foundation for broader intelligence rather than just being one component among many required for robust reasoning capabilities.

And yet as AI development continues its rapid advance, if CompressARC holds up to further scrutiny, it offers a glimpse of a possible alternative path that might lead to useful intelligent behavior without the resource demands of today’s dominant approaches. Or at the very least, it might unlock an important component of general intelligence in machines, which is still poorly understood.

Benj Edwards is Ars Technica’s Senior AI Reporter and founder of the site’s dedicated AI beat in 2022. He’s also a tech historian with almost two decades of experience. In his free time, he writes and records music, collects vintage computers, and enjoys nature. He lives in Raleigh, NC.

CMU research shows compression alone may unlock AI puzzle-solving abilities Read More »

No one asked for this: Google is testing round keys in Gboard

android, gboard, Google, Tech / Mike M. / March 7, 2025

Most Android phones ship with Google’s Gboard as the default input option. It’s a reliable, feature-rich on-screen keyboard, so most folks just keep using it instead of installing a third-party option. Depending on how you feel about circles, it might be time to check out some of those alternatives. Google has quietly released an update that changes the shape and position of the keys, and users are not pleased.

In the latest build of Gboard (v15.1.05.726012951-beta-arm64-v8a), Google has changed the key shape from the long-running squares to circle shapes. If you’re using the four-row layout, the keys are like little pills. In five-row mode with the exposed number row, the keys are collapsed further into circles. The reactions seem split between those annoyed by this change and those annoyed that everyone else is so annoyed.

Change can be hard sometimes, so certainly some of the discontent is just a function of having the phone interface changed without warning. If you find it particularly distasteful, you can head into the Gboard settings and open the Themes menu. From there, you can tap on a theme and then turn off the key borders. Thus, you won’t be distracted by the horror of rounded edges. That’s not the only problem with the silent update, though.

The wave of objections isn’t just about aesthetics—this update also moves the keys around a bit. After years of tapping away on keys with a particular layout, people develop muscle memory. Big texters can sometimes type messages on their phone without even looking at it, but moving the keys around even slightly, as Google has done here, can cause you to miss more keys than you did before the update.

No one asked for this: Google is testing round keys in Gboard Read More »

China aims to recruit top US scientists as Trump tries to kill the CHIPS Act

Chips Act, department of government efficiency, doge, Donald Trump, intel, Micron, NSF, NVIDIA, Policy, samsung, semiconductors, TSMC / Mike M. / March 6, 2025

Tech innovation in US likely to stall if Trump ends the CHIPS Act.

On Tuesday, Donald Trump finally made it clear to Congress that he wants to kill the CHIPS and Science Act—a $280 billion bipartisan law Joe Biden signed in 2022 to bring more semiconductor manufacturing into the US and put the country at the forefront of research and innovation.

Trump has long expressed frustration with the high cost of the CHIPS Act, telling Congress on Tuesday that it’s a “horrible, horrible thing” to “give hundreds of billions of dollars” in subsidies to companies that he claimed “take our money” and “don’t spend it,” Reuters reported.

“You should get rid of the CHIPS Act, and whatever is left over, Mr. Speaker, you should use it to reduce debt,” Trump said.

Instead, Trump potentially plans to shift the US from incentivizing chips manufacturing to punishing firms dependent on imports, threatening a 25 percent tariff on all semiconductor imports that could kick in as soon as April 2, CNBC reported.

The CHIPS Act was supposed to be Biden’s legacy, and because he made it a priority, much of the $52.7 billion in subsidies that Trump is criticizing has already been finalized. In 2022, Biden approved $39 billion in subsidies for semiconductor firms, and in his last weeks in office, he finalized more than $33 billion in awards, Reuters noted.

Among the awardees are leading semiconductor firms, including the Taiwan Semiconductor Manufacturing Co. (TSMC), Micron, Intel, Nvidia, and Samsung Electronics. Although Trump claims the CHIPS Act is one-sided and only serves to benefit firms, according to the Semiconductor Industry Association, the law sparked $450 billion in private investments increasing semiconductor production across 28 states by mid-2024.

With the CHIPS Act officially in Trump’s crosshairs, innovation appears likely to stall the longer that lawmakers remain unsettled on whether the law stays or goes. Some officials worried that Trump might interfere with Biden’s binding agreements with leading firms already holding up their end of the bargain, Reuters reported. For example, Micron plans to invest $100 billion in New York, and TSMC just committed to spending the same over the next four years to expand construction of US chips fabs, which is already well underway.

So far, Commerce Secretary Howard Lutnick has only indicated that he will review the finalized awards, noting that the US wouldn’t be giving TSMC any new awards, Reuters reported.

But the CHIPS Act does much more than provide subsidies to lure leading semiconductor companies into the US. For the first time in decades, the law created a new arm of the National Science Foundation (NSF)—the Directorate of Technology, Innovation, and Partnerships (TIP)—which functions unlike any other part of NSF and now appears existentially threatened.

Designed to take the country’s boldest ideas from basic research to real-world applications as fast as possible to make the US as competitive as possible, TIP helps advance all NSF research and was supposed to ensure US leadership in breakthrough technologies, including AI, 6G communications, biotech, quantum computing, and advanced manufacturing.

Biden allocated $20 billion to launch TIP through the CHIPS Act to accelerate technology development not just at top firms but also in small research settings across the US. But as soon as the Department of Government Efficiency (DOGE) started making cuts at NSF this year, TIP got hit the hardest. Seemingly TIP was targeted not because DOGE deemed it the least consequential but simply because it was the youngest directorate at NSF with the most workers in transition when Trump took office and DOGE abruptly announced it was terminating all “probationary” federal workers.

It took years to get TIP ready to flip the switch to accelerate tech innovation in the US. Without it, Trump risks setting the US back at a time when competitors like China are racing ahead and wooing US scientists who suddenly may not know if or when their funding is coming, NSF workers and industry groups told Ars.

Without TIP, NSF slows down

Last month, DOGE absolutely scrambled the NSF by forcing arbitrary cuts of so-called probationary employees—mostly young scientists, some of whom were in transition due to promotions. All those cuts were deemed illegal and finally reversed Monday by court order after weeks of internal chaos reportedly stalling or threatening to delay some of the highest-priority research in the US.

“The Office of Personnel Management does not have any authority whatsoever under any statute in the history of the universe to hire and fire employees at another agency,” US District Judge William Alsup said, calling probationary employees the “life blood” of government agencies.

Ars granted NSF workers anonymity to discuss how cuts were impacting research. At TIP, a federal worker told Ars that one of the probationary cuts in particular threatened to do the most damage.

Because TIP is so new, only one worker was trained to code automated tracking forms that helped decision-makers balance budgets and approve funding for projects across NSF in real time. Ars’ source likened it to holding the only key to the vault of NSF funding. And because TIP is so different from other NSF branches—hiring experts never pulled into NSF before and requiring customized resources to coordinate projects across all NSF fields of research—the insider suggested another government worker couldn’t easily be substituted. It could take possibly two years to hire and train a replacement on TIP’s unique tracking system, the source said, while TIP’s (and possibly all of NSF’s) efficiency is likely strained.

TIP has never been fully functional, the TIP insider confirmed, and could be choked off right as it starts helping to move the needle on US innovation. “Imagine where we are in two years and where China is in two years in quantum computing, semiconductors, or AI,” the TIP insider warned, pointing to China’s surprisingly advanced AI model, DeepSeek, as an indicator of how quickly tech leadership in global markets can change.

On Monday, NSF emailed all workers to confirm that all probationary workers would be reinstated “right away.” But the damage may already be done as it’s unclear how many workers plan to return. When TIP lost the coder—who was seemingly fired for a technicality while transitioning to a different payscale—NSF workers rushed to recommend the coder on LinkedIn, hoping to help the coder quickly secure another opportunity in industry or academia.

Ars could not reach the coder to confirm whether a return to TIP is in the cards. But Ars’ source at TIP and another NSF worker granted anonymity said that probationary workers may be hesitant to return because they are likely to be hit in any official reductions in force (RIFs) in the future.

“RIFs done the legal way are likely coming down the pipe, so these staff are not coming back to a place of security,” the NSF worker said. “The trust is broken. Even for those that choose to return, they’d be wise to be seeking other opportunities.”

And even losing the TIP coder for a couple of weeks likely slows NSF down at a time when the US seemingly can’t afford to lose a single day.

“We’re going to get murdered” if China sets the standard on 6G or AI, the TIP worker fears.

Rivals and allies wooing top US scientists

On Monday, six research and scientific associations, which described themselves as “leading organizations representing more than 305,000 people in computing, information technology, and technical innovation across US industry, academia, and government,” wrote to Congress demanding protections for the US research enterprise.

The groups warned that funding freezes and worker cuts at NSF—and other agencies, including the Department of Energy, the National Institute of Standards & Technology, the National Aeronautics and Space Administration, the National Institutes of Health—”have caused disruption and uncertainty” and threaten “long-lasting negative consequences for our competitiveness, national security, and economic prosperity.”

Deeming America’s technology leadership at risk, the groups pointed out that “in computing alone, a federal investment in research of just over $10 billion annually across 24 agencies and offices underpins a technology sector that contributes more than $2 trillion to the US GDP each year.” Cutting US investment “would be a costly mistake, far outweighing any short-term savings,” the groups warned.

In a separate statement, the Computing Research Association (CRA) called NSF cuts, in particular, a “deeply troubling, self-inflicted setback to US leadership in computing research” that appeared “penny-wise and pound-foolish.”

“NSF is one of the most efficient federal agencies, operating with less than 9 percent overhead costs,” CRA said. “These arbitrary terminations are not justified by performance metrics or efficiency concerns; rather, they represent a drastic and unnecessary weakening of the US research enterprise.”

Many NSF workers are afraid to speak up, the TIP worker told Ars, and industry seems similarly tight-lipped as confusion remains. Only one of the organizations urging Congress to intervene agreed to talk to Ars about the NSF cuts and the significance of TIP. Kathryn Kelley, the executive director of the Coalition for Academic Scientific Computation, confirmed that while members are more aligned with NSF’s Directorate for Computer and Information Science and Engineering and the Office of Advanced Cyberinfrastructure, her group agrees that all NSF cuts are “deeply” concerning.

“We agree that the uncertainty and erosion of trust within the NSF workforce could have long-lasting effects on the agency’s ability to attract and retain top talent, particularly in such specialized areas,” Kelley told Ars. “This situation underscores the need for continued investment in a stable, well-supported workforce to maintain the US’s leadership in science and innovation.”

Other industry sources unwilling to go on the record told Ars that arbitrary cuts largely affecting the youngest scientists at NSF threatened to disrupt a generation of researchers who envisioned long careers advancing US tech. There’s now a danger that those researchers may be lured to other countries heavily investing in science and currently advertising to attract displaced US researchers, including not just rivals like China but also allies like Denmark.

Those sources questioned the wisdom of using the Elon Musk-like approach of breaking the NSF to rebuild it when it’s already one of the leanest organizations in government.

Ars confirmed that some PhD programs have been cancelled, as many academic researchers are already widely concerned about delayed or cancelled grants and generally freaked out about where to get dependable funding outside the NSF. And in industry, some CHIPS Act projects have already been delayed, as companies like Intel try to manage timelines without knowing what’s happening with CHIPS funding, AP News reported.

“Obviously chip manufacturing companies will slow spending on programs they previously thought they were getting CHIPS Act funding for if not cancel those projects outright,” the Semiconductor Advisors, an industry group, forecasted in a statement last month.

The TIP insider told Ars that the CHIPS Act subsidies for large companies that Trump despises mostly fuel manufacturing in the US, while funding for smaller research facilities is what actually advances technology. Reducing efficiency at TIP would likely disrupt those researchers the most, the TIP worker suggested, proclaiming that’s why TIP must be saved at all costs.

Ashley is a senior policy reporter for Ars Technica, dedicated to tracking social impacts of emerging policies and new technologies. She is a Chicago-based journalist with 20 years of experience.

China aims to recruit top US scientists as Trump tries to kill the CHIPS Act Read More »

Elon Musk loses initial attempt to block OpenAI’s for-profit conversion

elon musk, openai, Policy / Mike M. / March 6, 2025

A federal judge rejected Elon Musk’s request to block OpenAI’s planned conversion from a nonprofit to for-profit entity but expedited the case so that Musk’s core claims can be addressed in a trial before the end of this year.

Musk had filed a motion for preliminary injunction in US District Court for the Northern District of California, claiming that OpenAI’s for-profit conversation “violates the terms of Musk’s donations” to the company. But Musk failed to meet the burden of proof needed for an injunction, Judge Yvonne Gonzalez Rogers ruled yesterday.

“Plaintiffs Elon Musk, [former OpenAI board member] Shivon Zilis, and X.AI Corp. (‘xAI’) collectively move for a preliminary injunction barring defendants from engaging in various business activities, which plaintiffs claim violate federal antitrust and state law,” Rogers wrote. “The relief requested is extraordinary and rarely granted as it seeks the ultimate relief of the case on an expedited basis, with a cursory record, and without the benefit of a trial.”

Rogers said that “the Court is prepared to offer an expedited schedule on the core claims driving this litigation [to] address the issues which are allegedly more urgent in terms of public, not private, considerations.” There would be important public interest considerations if the for-profit shift is found to be illegal at a trial, she wrote.

Musk said OpenAI took advantage of him

Noting that OpenAI donors may have taken tax deductions from a nonprofit that is now turning into a for-profit enterprise, Rogers said the court “agrees that significant and irreparable harm is incurred when the public’s money is used to fund a non-profit’s conversion into a for-profit.” But as for the motion to block the for-profit conversion before a trial, “The request for an injunction barring any steps towards OpenAI’s conversion to a for-profit entity is DENIED.”

Elon Musk loses initial attempt to block OpenAI’s for-profit conversion Read More »

Trump’s 25% tariffs take effect; Canadian PM calls it “a very dumb thing”

Policy, trump tariffs / Mike M. / March 5, 2025

President Trump’s 25 percent tariffs on Canada and Mexico took effect today, and the White House increased a tariff on China from 10 to 20 percent in an executive order. Canada, Mexico, and China announced retaliatory moves, and stock markets sank globally.

Industry groups and companies have warned the Trump tariffs will raise prices for cars, groceries, consumer technology, and other products.

Canada was hit with a 10 percent tariff on energy exports to the US, while other Canadian exports are subject to the 25 percent tariff. Prime Minister Justin Trudeau said his country would impose 25 percent tariffs on over $100 billion worth of US goods.

China responded with new tariffs of 10 to 15 percent on US agricultural products starting March 10 and other retaliatory moves such as blacklisting 15 US companies. Mexican President Claudia Sheinbaum said she will announce retaliatory tariffs and other measures on Sunday.

“There’s no motive, reason or justification that supports this decision that will affect our people and nations,” Sheinbaum reportedly said at a news conference. “We will always seek a negotiated solution, in a framework of respect.” Sheinbaum said Mexico wants “dialogue, with reasoning and rationality.”

Trudeau: Trump tariffs “a very dumb thing to do”

Trudeau said that Trump imposing tariffs is “a very dumb thing to do” and that Canada “will not back down from a fight.” Addressing US residents, Trudeau said, “We don’t want this, but your government has chosen to do this to you.”

Trump’s executive orders have blamed Canada for “the flow of illicit drugs across our northern border,” Mexico for a “sustained influx of illegal aliens and illicit opioids and other drugs,” and China for a “sustained influx of synthetic opioids, including fentanyl.”

Trump’s 25% tariffs take effect; Canadian PM calls it “a very dumb thing” Read More »

Researchers surprised to find less-educated areas adopting AI writing tools faster

AI, AI research, AI writing, AI writing detectors, Biz & IT, chatgpt, chatgtp, large language models, machine learning, Stanford, Tech / Mike M. / March 4, 2025

From the mouths of machines

Stanford researchers analyzed 305 million texts, revealing AI-writing trends.

Since the launch of ChatGPT in late 2022, experts have debated how widely AI language models would impact the world. A few years later, the picture is getting clear. According to new Stanford University-led research examining over 300 million text samples across multiple sectors, AI language models now assist in writing up to a quarter of professional communications across sectors. It’s having a large impact, especially in less-educated parts of the United States.

“Our study shows the emergence of a new reality in which firms, consumers and even international organizations substantially rely on generative AI for communications,” wrote the researchers.

The researchers tracked large language model (LLM) adoption across industries from January 2022 to September 2024 using a dataset that included 687,241 consumer complaints submitted to the US Consumer Financial Protection Bureau (CFPB), 537,413 corporate press releases, 304.3 million job postings, and 15,919 United Nations press releases.

By using a statistical detection system that tracked word usage patterns, the researchers found that roughly 18 percent of financial consumer complaints (including 30 percent of all complaints from Arkansas), 24 percent of corporate press releases, up to 15 percent of job postings, and 14 percent of UN press releases showed signs of AI assistance during that period of time.

The study also found that while urban areas showed higher adoption overall (18.2 percent versus 10.9 percent in rural areas), regions with lower educational attainment used AI writing tools more frequently (19.9 percent compared to 17.4 percent in higher-education areas). The researchers note that this contradicts typical technology adoption patterns where more educated populations adopt new tools fastest.

“In the consumer complaint domain, the geographic and demographic patterns in LLM adoption present an intriguing departure from historical technology diffusion trends where technology adoption has generally been concentrated in urban areas, among higher-income groups, and populations with higher levels of educational attainment.”

Researchers from Stanford, the University of Washington, and Emory University led the study, titled, “The Widespread Adoption of Large Language Model-Assisted Writing Across Society,” first listed on the arXiv preprint server in mid-February. Weixin Liang and Yaohui Zhang from Stanford served as lead authors, with collaborators Mihai Codreanu, Jiayu Wang, Hancheng Cao, and James Zou.

Detecting AI use in aggregate

We’ve previously covered that AI writing detection services aren’t reliable, and this study does not contradict that finding. On a document-by-document basis, AI detectors cannot be trusted. But when analyzing millions of documents in aggregate, telltale patterns emerge that suggest the influence of AI language models on text.

The researchers developed an approach based on a statistical framework in a previously released work that analyzed shifts in word frequencies and linguistic patterns before and after ChatGPT’s release. By comparing large sets of pre- and post-ChatGPT texts, they estimated the proportion of AI-assisted content at a population level. The presumption is that LLMs tend to favor certain word choices, sentence structures, and linguistic patterns that differ subtly from typical human writing.

To validate their approach, the researchers created test sets with known percentages of AI content (from zero percent to 25 percent) and found their method predicted these percentages with error rates below 3.3 percent. This statistical validation gave them confidence in their population-level estimates.

While the researchers specifically note their estimates likely represent a minimum level of AI usage, it’s important to understand that actual AI involvement might be significantly greater. Due to the difficulty in detecting heavily edited or increasingly sophisticated AI-generated content, the researchers say their reported adoption rates could substantially underestimate true levels of generative AI use.

Analysis suggests AI use as “equalizing tools”

While the overall adoption rates are revealing, perhaps more insightful are the patterns of who is using AI writing tools and how these patterns may challenge conventional assumptions about technology adoption.

In examining the CFPB complaints (a US public resource that collects complaints about consumer financial products and services), the researchers’ geographic analysis revealed substantial variation across US states.

Arkansas showed the highest adoption rate at 29.2 percent (based on 7,376 complaints), followed by Missouri at 26.9 percent (16,807 complaints) and North Dakota at 24.8 percent (1,025 complaints). In contrast, states like West Virginia (2.6 percent), Idaho (3.8 percent), and Vermont (4.8 percent) showed minimal AI writing adoption. Major population centers demonstrated moderate adoption, with California at 17.4 percent (157,056 complaints) and New York at 16.6 percent (104,862 complaints).

The urban-rural divide followed expected technology adoption patterns initially, but with an interesting twist. Using Rural Urban Commuting Area (RUCA) codes, the researchers found that urban and rural areas initially adopted AI writing tools at similar rates during early 2023. However, adoption trajectories diverged by mid-2023, with urban areas reaching 18.2 percent adoption compared to 10.9 percent in rural areas.

Contrary to typical technology diffusion patterns, areas with lower educational attainment showed higher AI writing tool usage. Comparing regions above and below state median levels of bachelor’s degree attainment, areas with fewer college graduates stabilized at 19.9 percent adoption rates compared to 17.4 percent in more educated regions. This pattern held even within urban areas, where less-educated communities showed 21.4 percent adoption versus 17.8 percent in more educated urban areas.

The researchers suggest that AI writing tools may serve as a leg-up for people who may not have as much educational experience. “While the urban-rural digital divide seems to persist,” the researchers write, “our finding that areas with lower educational attainment showed modestly higher LLM adoption rates in consumer complaints suggests these tools may serve as equalizing tools in consumer advocacy.”

Corporate and diplomatic trends in AI writing

According to the researchers, all sectors they analyzed (consumer complaints, corporate communications, job postings) showed similar adoption patterns: sharp increases beginning three to four months after ChatGPT’s November 2022 launch, followed by stabilization in late 2023.

Organization age emerged as the strongest predictor of AI writing usage in the job posting analysis. Companies founded after 2015 showed adoption rates up to three times higher than firms established before 1980, reaching 10–15 percent AI-modified text in certain roles compared to below 5 percent for older organizations. Small companies with fewer employees also incorporated AI more readily than larger organizations.

When examining corporate press releases by sector, science and technology companies integrated AI most extensively, with an adoption rate of 16.8 percent by late 2023. Business and financial news (14–15.6 percent) and people and culture topics (13.6–14.3 percent) showed slightly lower but still significant adoption.

In the international arena, Latin American and Caribbean UN country teams showed the highest adoption among international organizations at approximately 20 percent, while African states, Asia-Pacific states, and Eastern European states demonstrated more moderate increases to 11–14 percent by 2024.

Implications and limitations

In the study, the researchers acknowledge limitations in their analysis due to a focus on English-language content. Also, as we mentioned earlier, they found they could not reliably detect human-edited AI-generated text or text generated by newer models instructed to imitate human writing styles. As a result, the researchers suggest their findings represent a lower bound of actual AI writing tool adoption.

The researchers noted that the plateauing of AI writing adoption in 2024 might reflect either market saturation or increasingly sophisticated LLMs producing text that evades detection methods. They conclude we now live in a world where distinguishing between human and AI writing becomes progressively more difficult, with implications for communications across society.

“The growing reliance on AI-generated content may introduce challenges in communication,” the researchers write. “In sensitive categories, over-reliance on AI could result in messages that fail to address concerns or overall release less credible information externally. Over-reliance on AI could also introduce public mistrust in the authenticity of messages sent by firms.”

Researchers surprised to find less-educated areas adopting AI writing tools faster Read More »

Gemini Live will learn to peer through your camera lens in a few weeks

gemini live, Google, Google Gemini, Tech / Mike M. / March 4, 2025

At Mobile World Congress, Google confirmed that a long-awaited Gemini AI feature it first teased nearly a year ago is ready for launch. The company’s conversational Gemini Live will soon be able to view live video and screen sharing, a feature Google previously demoed as Project Astra. When Gemini’s video capabilities arrive, you’ll be able to simply show the robot something instead of telling it.

Right now, Google’s multimodal AI can process text, images, and various kinds of documents. However, its ability to accept video as an input is spotty at best—sometimes it can summarize a YouTube video, and sometimes it can’t, for unknown reasons. Later in March, the Gemini app on Android will get a major update to its video functionality. You’ll be able to open your camera to provide Gemini Live a video stream or share your screen as a live video, thus allowing you to pepper Gemini with questions about what it sees.

Gemini Live with video.

It can be hard to keep track of which Google AI project is which—the 2024 Google I/O was largely a celebration of all things Gemini AI. The Astra demo made waves as it demonstrated a more natural way to interact with the AI. In the original video, which you can see below, Google showed how Gemini Live could answer questions in real time as the user swept a phone around a room. It had things to say about code on a computer screen, how speakers work, and a network diagram on a whiteboard. It even remembered where the user left their glasses from an earlier part of the video.

Gemini Live will learn to peer through your camera lens in a few weeks Read More »

Half-Life 3 is just the hot exclusive Valve needs to propel SteamOS past Windows

gaming, Half-Life 3, steam machine, valve / Mike M. / March 2, 2025

The ultimate system seller

Opinion: Just as Half-Life 2 helped launch Steam, a sequel could help establish non-Windows PC gaming.

We found this logo hidden deep in an abandoned steel forge, Credit: Aurich Lawson | Steam

A little over 20 years ago, Valve was getting ready to release a new Half-Life game. At the same time, the company was trying to push Steam as a new option for players to download and update games over the Internet.

Requiring Steam in order to play Half-Life 2 led to plenty of grumbling from players in 2004. But the high-profile Steam exclusive helped build an instant user base for Valve’s fresh distribution system, setting it on a path to eventually become the unquestioned leader in the space. The link between the new game and the new platform helped promote a bold alternative to the retail game sales and distribution systems that had dominated PC gaming for decades.

Today, all indications suggest that Valve is getting ready to release a new Half-Life game. At the same time, the company is getting ready to push SteamOS as a new option for third-party hardware makers and individual users to “download and test themselves.”

Requiring SteamOS to play Half-Life 3 would definitely lead to a lot of grumbling from players. But the high-profile exclusive could help build an instant user base for Valve’s fresh operating system, perhaps setting it on the path to become the unquestioned leader in the space. A link between the new game and the new platform could help promote a bold alternative to the Windows-based systems that have dominated PC gaming for decades.

Not another Steam Machine

Getting players to change the established platform they use to buy and play games (either in terms of hardware or software) usually requires some sort of instantly apparent benefit for the player. Those benefits can range from the tangible (e.g., an improved controller, better graphics performance) to the ancillary (e.g., social features, achievements) to the downright weird (e.g., a second screen on a portable). Often, though, a core reason why players switch platforms is for access to exclusive “system seller” games that aren’t available any other way.

Half-Life 2‘s role in popularizing early Steam shows just how much a highly anticipated exclusive can convince otherwise reluctant players to invest time and effort in a new platform. To see what can happen without such an exclusive, we only need to look to Valve’s 2015 launch of the Steam Machine hardware line, powered by the first version of the Linux-based SteamOS.

Valve offered players very little in the way of affirmative reasons to switch to a SteamOS-powered Steam Machine in 2015. Credit: Alienware

At the time, Valve was selling SteamOS mainly as an alternative to a new Windows 8 environment that Valve co-founder Gabe Newell saw as a “catastrophe” in the making for the PC gaming world. Newell described SteamOS as a “hedging strategy” against Microsoft’s potential ability to force all Windows 8 app distribution through the Windows Store, a la Apple’s total control of iPhone app distribution.

When Microsoft failed to impose that kind of hegemonic control over Windows apps and games, Valve was left with little else to convince players that it was worth buying a Windows-free Steam Machine (or going through the onerous process of installing the original SteamOS on their gaming rigs). Sure, using SteamOS meant saving a few bucks on a Windows license. But it also meant being stuck with an extremely limited library of Linux ports (especially when it came to releases from major publishers) and poor technical performance compared to Windows even when those ports were available.

Given those obvious downsides—and the lack of any obvious upsides—it’s no wonder that users overwhelmingly ignored SteamOS and Steam Machines at the time. But as we argued way back in 2013, a major exclusive on the scale of Half-Life 3 could have convinced a lot of gamers to overlook at least some of those downsides and give the new platform a chance.

A little push

Fast forward to today, and the modern version of SteamOS is in a much better place than the Steam Machine-era version ever was. That’s thanks in large part to Valve’s consistent work on the Proton compatibility layer, which lets the Linux-based SteamOS run almost any game that’s designed for Windows (with only a few major exceptions). That wide compatibility has been a huge boon for the Steam Deck, which offered many players easy handheld access to vast swathes of PC gaming for the first time. The Steam Deck also showed off SteamOS’s major user interface and user experience benefits over clunkier Windows-based gaming portables.

The Steam Deck served as an excellent proof of concept for the viability of SteamOS hardware with the gaming masses. Credit: Kyle Orland

Still, the benefits of switching from Windows to SteamOS might seem a bit amorphous to many players today. If Valve is really interested in pushing its OS as an alternative to Windows gaming, a big exclusive game is just the thing to convince a critical mass of players to make the leap. And when it comes to massive PC gaming exclusives, it doesn’t get much bigger than the long, long-awaited Half-Life 3.

We know it might sound ludicrous to suggest that Valve’s biggest game in years should ignore the Windows platform that’s been used by practically every PC gamer for decades. Keep in mind, though, that there would be nothing stopping existing Windows gamers from downloading and installing a free copy of the Linux-based SteamOS (likely on a separate drive or partition) to get access to Half-Life 3.

Yes, installing a new operating system (especially one based on Linux) is not exactly a plug-and-play process. But Valve has a long history of streamlining game downloads, updates, and driver installations through Steam itself. If anyone can make the process of setting up a new OS relatively seamless, it’s Valve.

And let’s not forget that millions of gamers already have easy access to SteamOS through Steam Deck hardware. Those aging Steam Decks might not be powerful enough to run a game like Half-Life 3 at maximum graphics settings, but Valve games have a history of scaling down well on low-end systems.

Valve’s leaked “Powered by SteamOS” initiative also seems poised to let third-party hardware makers jump in with more powerful (and more Half-Life 3-capable) desktops, laptops, and handhelds with SteamOS pre-installed. And that’s before we even consider the potential impact of a more powerful “Steam Deck 2,” which Valve’s Pierre-Loup Griffais said in 2023 could potentially come in “the next couple of years.”

Time for a bold move

Tying a major game like Half-Life 3 to a completely new and largely untested operating system would surely lead to some deafening pushback from gamers happy with the Windows-based status quo. An exclusive release could also be risky if SteamOS ends up showing some technical problems as it tries to grow past its Steam Deck roots (Linux doesn’t exactly have the best track record when it comes to things like game driver compatibility across different hardware).

The Lenovo Legion Go S will be the first non-Valve hardware to be officially “Powered by SteamOS.” A Windows-sporting version will be more expensive Credit: Lenovo

Despite all that, we’re pretty confident that the vast majority of players interested in Half-Life 3 would jump through a few OS-related hoops to get access to the game. And many of those players would likely stick with Valve’s gaming-optimized OS going forward rather than spending money on another Windows license.

Even a timed exclusivity window for Half-Life 3 on SteamOS could push a lot of early adopters to see what all the fuss is about without excluding those who refuse to switch away from Windows. Failing even that, maybe a non-exclusive Half-Life 3 could be included as a pre-installed freebie with future versions of SteamOS, as an incentive for the curious to try out a new operating system.

With the coming wide release of SteamOS, Valve has a rare opportunity to upend the PC gaming OS dominance that Microsoft more or less stumbled into decades ago. A game like Half-Life 3 could be just the carrot needed to get PC gaming as a whole over its longstanding Windows dependence.

Kyle Orland has been the Senior Gaming Editor at Ars Technica since 2012, writing primarily about the business, tech, and culture behind video games. He has journalism and computer science degrees from University of Maryland. He once wrote a whole book about Minesweeper.

Half-Life 3 is just the hot exclusive Valve needs to propel SteamOS past Windows Read More »

Federal firings could wreak havoc on Great Lakes fishery

conservation, doge, great lakes, Invasive species, Science, syndicatino, syndication, USFWS / Mike M. / March 2, 2025

Her performance reviews for the last year had been glowing, so the letter made no sense. “It’s not a real explanation,” she said.

The USFWS layoffs will not affect the sea lamprey control program in Canada, McClinchey said. “The Canadian government has assured us that the money from Canada will continue to be there and we’re on track to deliver a full program in Canadian waters,” he said. “That’s great, but this program works because it’s border blind.”

In other words: Cuts to lamprey control in US waters are a threat to fish and fishermen everywhere on the Great Lakes.

Just a week ago, the Great Lakes Fishery Commission faced a more dire staffing situation, as the USFWS informed directors they’d also be unable to hire seasonal workers to spread lampricide come April. Within a few days, that hiring freeze was reversed, said McClinchey.

This reversal gives him a bit of hope. “That at least tells us no one is rooting for the lamprey,” he said.

McClinchey is currently in DC for appropriation season, presenting the commission’s work to members of Congress and defending the agency’s budget. It’s an annual trip, but this year he’s also advocating for the reinstatement of laid-off lamprey control employees.

He is optimistic. “It seems clear to me that it’s important we preserve this program, and so far everyone we’ve encountered thinks that way and are working to that end,” he said.

Cutting back the program isn’t really on the table for the commission. Even minor cuts to scope would be devastating for the fishery, he said.

Even the former USFWS employee from Marquette is remaining hopeful. “I still think that they’re going to scramble to make it happen,” she said. “Because it’s not really an option to just stop treating for a whole season.”

This story originally appeared on Inside Climate News.

Federal firings could wreak havoc on Great Lakes fishery Read More »

Details on AMD’s $549 and $599 Radeon RX 9070 GPUs, which aim at Nvidia and 4K

AMD, amd fsr, Blackwell, gaming, GeForce, GPU, NVIDIA, Nvidia Blackwell, Radeon, radeon 9070, radeon rx 9070, rdna4, Tech / Mike M. / March 2, 2025

AMD is releasing the first detailed specifications of its next-generation Radeon RX 9070 series GPUs and the RDNA4 graphics architecture today, almost two months after teasing them at CES.

The short version is that these are both upper-midrange graphics cards targeting resolutions of 1440p and 4K and meant to compete mainly with Nvidia’s incoming and outgoing 4070- and 5070-series GeForce GPUs, including the RTX 4070, RTX 5070, RTX 4070 Ti and Ti Super, and the RTX 5070 Ti.

AMD says the RX 9070 will start at $549, the same price as Nvidia’s RTX 5070. The slightly faster 9070 XT starts at $599, $150 less than the RTX 5070 Ti. The cards go on sale March 6, a day after Nvidia’s RTX 5070.

Neither Nvidia nor Intel has managed to keep its GPUs in stores at their announced starting prices so far, though, so how well AMD’s pricing stacks up to Nvidia in the real world may take a few weeks or months to settle out. For its part, AMD says it’s confident that it has enough supply to meet demand, but that’s as specific as the company’s reassurances got.

Specs and speeds: Radeon RX 9070 and 9070 XT

	RX 9070 XT	RX 9070	RX 7900 XTX	RX 7900 XT	RX 7900 GRE	RX 7800 XT
Compute units (Stream processors)	64 RDNA4 (4,096)	56 RDNA4 (3,584)	96 RDNA3 (6,144)	84 RDNA3 (5,376)	80 RDNA3 (5,120)	60 RDNA3 (3,840)
Boost Clock	2,970 MHz	2,520 MHz	2,498 MHz	2,400 MHz	2,245 MHz	2,430 MHz
Memory Bus Width	256-bit	256-bit	384-bit	320-bit	256-bit	256-bit
Memory Bandwidth	650 GB/s	650 GB/s	960 GB/s	800 GB/s	576 GB/s	624 GB/s
Memory size	16GB GDDR6	16GB GDDR6	24GB GDDR6	20GB GDDR6	16GB GDDR6	16GB GDDR6
Total board power (TBP)	304 W	220 W	355 W	315 W	260 W	263 W

As is implied by their similar price tags, the 9070 and 9070 XT have more in common than not. Both are based on the same GPU die—the 9070 has 56 of the chip’s compute units enabled, while the 9070 XT has 64. Both cards come with 16GB of RAM (4GB more than the 5070, the same amount as the 5070 Ti) on a 256-bit memory bus, and both use two 8-pin power connectors by default, though the 9070 XT can use significantly more power than the 9070 (304 W, compared to 220 W).

AMD says that its partners are free to make Radeon cards with the 12VHPWR or 12V-2×6 power connectors on them, though given the apparently ongoing issues with the connector, we’d expect most Radeon GPUs to stick with the known quantity that is the 8-pin connector.

AMD says that the 9070 series is made using a 4 nm TSMC manufacturing process and that the chips are monolithic rather than being split up into chiplets as some RX 7000-series cards were. AMD’s commitment to its memory controller chiplets was always hit or miss with the 7000-series—the high-end cards tended to use them, while the lower-end GPUs were usually monolithic—so it’s not clear one way or the other whether this means AMD is giving up on chiplet-based GPUs altogether or if it’s just not using them this time around.

Details on AMD’s $549 and $599 Radeon RX 9070 GPUs, which aim at Nvidia and 4K Read More »

AMD’s FSR 4 upscaling is exclusive to 90-series Radeon GPUs, won’t work on other cards

AMD, AMD FidelityFX Super Resolution, FidelityFX Super Resolution, fsr 4, gaming, Radeon, radeon 9070, rdna4, Tech / Mike M. / March 2, 2025

AMD’s new Radeon RX 90-series cards and the RDNA4 architecture make their official debut on March 5, and a new version of AMD’s FidelityFX Super Resolution (FSR) upscaling technology is coming along with them.

FSR and Nvidia’s Deep Learning Super Sampling (DLSS) upscalers have the same goal: to take a lower-resolution image rendered by your graphics card, bump up the resolution, and fill in the gaps between the natively rendered pixels to make an image that looks close to natively rendered without making the GPU do all that rendering work. These upscalers can make errors, and they won’t always look quite as good as a native-resolution image. But they’re both nice alternatives to living with a blurry, non-native-resolution picture on an LCD or OLED display.

FSR and DLSS are especially useful for older or cheaper 1080p or 1440p-capable GPUs that are connected to a 4K monitor, where you’d otherwise have to decide between a sharp 4K image and a playable frame rate; it’s also useful for hitting higher frame rates at lower resolutions, which can be handy for high-refresh-rate gaming monitors.

But unlike past versions of FSR, FSR 4 is upscaling images using hardware-backed machine-learning algorithms, hardware newly added to RDNA4 and the RX 90-series graphics cards. This mirrors Nvidia’s strategy with DLSS, which has always leveraged the tensor cores found in RTX GPUs to run machine-learning models to achieve superior image quality for upscaled and AI-generated frames. If you don’t have an RDNA4 GPU, you can’t use FSR 4.

AMD’s FSR 4 upscaling is exclusive to 90-series Radeon GPUs, won’t work on other cards Read More »

Author name: Mike M.