SoylentNews

taylorvich writes:

https://newatlas.com/technology/artificial-gills-unlock-long-range-underwater-robots/

What's good for fish may be good for robots, too, as researchers from the Helmholtz-Zentrum Hereon have developed an artificial gill that can extract oxygen from seawater to run fuel cells to power robotic sea gliders on long missions.

Underwater gliders have become an increasingly valuable tool for oceanographic research. Ditching traditional propellers and thrusters, they move about by means of variable buoyancy propulsion, which is a cumbersome way of saying that they propel themselves by rising in the water and then using hydrofoils to control their direction as they descend.

It's not very fast, but it is economical and allows the gliders to carry out long missions across thousands of miles to monitor ocean conditions, seek out pollution, and conduct military reconnaissance as they dive to depths of up to 1,000 m (3,300 ft). They are also much cheaper to operate than research vessels, so what's the problem?

The fly in the deep sea ointment is powering the gliders. Batteries are needed to run the sensors, recorders, and telemetry systems but the go-to lithium batteries are classified as containing hazardous materials that are subject to strict safety and environmental regulations.

In addition, lithium batteries have their technical limitations. They are sensitive to pressure, are in danger of leaking if seals are compromised, can be severely damaged by seawater, do not handle cold temperatures well, and can release dangerous chemicals.

As a safer, less restricted alternative, Hereon engineers Dr. Lucas Merckelbach and Dr. Prokopios Georgopanos have been looking at fuel cells, which convert hydrogen and oxygen into electricity. The hydrogen is easy enough to store until ready. Just store it in a container with metal hydrides that absorb the hydrogen until needed. The oxygen is another matter. It weighs eight times per unit more than hydrogen in water and is very difficult to store even in cryogenic conditions.

Hereon's answer is not to even bother. Instead, the non-profit research institution has come up with an advanced silicone polymer membrane that has high oxygen permeability, yet is hydrophobic to keep water from seeping through. In the sea, the higher oxygen concentration on the wet side allows the oxygen atoms to migrate through the membrane to be collected by an internal recirculating airflow and fed to a Proton Exchange Membrane Fuel Cell (PEMFC) where it combines with hydrogen to produce electricity, with water the only waste product.

According to the team, the modular system's design can handle various underwater conditions, including changes in water temperature, salinity, and pressure to ensure consistent oxygen supply. In addition, computational fluid dynamics (CFD) simulations optimize the flow of water around the membrane.

Journal Reference: https://doi.org/10.1002/advs.202410358

Original Submission

upstart writes:

Are we all aliens? NASA's returned asteroid samples hold the ingredients of life from a watery world:

Asteroid samples fetched by NASA hold not only the pristine building blocks for life but also the salty remains of an ancient water world, scientists reported Wednesday.

The findings provide the strongest evidence yet that asteroids may have planted the seeds of life on Earth and that these ingredients were mingling with water almost right from the start.

"That's the kind of environment that could have been essential to the steps that lead from elements to life," said the Smithsonian Institution's Tim McCoy, one of the lead study authors.

NASA's Osiris-Rex spacecraft returned 122 grams (4 ounces) of dust and pebbles from the near-Earth asteroid Bennu, delivering the sample canister to the Utah desert in 2023 before swooping off after another space rock. It remains the biggest cosmic haul from beyond the moon. The two previous asteroid sample missions, by Japan, yielded considerably less material.

[...] Some if not all of the delicate salts found at Bennu—similar to what's in the dry lakebeds of California's Mojave Desert and Africa's Sahara—would be stripped away if present in falling meteorites.

"This discovery was only possible by analyzing samples that were collected directly from the asteroid then carefully preserved back on Earth," the Institute of Science Tokyo's Yasuhito Sekine, who was not involved in the studies, said in an accompanying editorial.

Combining the ingredients of life with an environment of sodium-rich salt water, or brines, "that's really the pathway to life," said McCoy, the National Museum of Natural History's curator of meteorites. "These processes probably occurred much earlier and were much more widespread than we had thought before."

NASA's Daniel Glavin said one of the biggest surprises was the relatively high abundance of nitrogen, including ammonia. While all of the organic molecules found in the Bennu samples have been identified before in meteorites, Glavin said the ones from Bennu are valid—"real extraterrestrial organic material formed in space and not a result of contamination from Earth."

[...] Many are pushing for a mission to collect rocks and dirt from the potentially waterlogged dwarf planet Ceres in the main asteroid belt. Jupiter's moon Europa and Saturn's moon Enceladus also beckon as enticing water worlds. Meanwhile, NASA has core samples awaiting pickup at Mars, but their delivery is on hold while the space agency studies the quickest and cheapest way to get them here.

"Are we alone?" McCoy said. "That's one of the questions we're trying to answer."

Journal References:
    • Glavin, D.P., Dworkin, J.P., Alexander, C.M.O. et al. Abundant ammonia and nitrogen-rich soluble organic matter in samples from asteroid (101955) Bennu. Nat Astron (2025). https://doi.org/10.1038/s41550-024-02472-9
    • McCoy, T.J., Russell, S.S., Zega, T.J. et al. An evaporite sequence from ancient brine recorded in Bennu samples. Nature 637, 1072–1077 (2025). https://doi.org/10.1038/s41586-024-08495-6

Original Submission

An Anonymous Coward writes:

The Fediverse - including Mastodon, Pixelfed, and others - is experiencing explosive exponential growth with over 700,000 new users, and 100,000,000 posts in January.

Pixelfed alone is growing 100k users per WEEK now, has gone 10x in a month.

Mastodon servers are welcoming influxes, and new servers are standing up at a rapid pace, with new active daily users up 200k in Jan, about +25%, and posts up by a similar +30% to 16 million/month.

Tumblr is planning to federate.

Forum software NodeBB has officially launched their 4.0 version, which includes ActivityPub support.

https://fediversereport.com/fediverse-report-101/
https://arstechnica.com/tech-policy/2025/01/mastodon-becomes-nonprofit-to-make-sure-its-never-ruined-by-billionaire-ceo/
https://www.404media.co/decentralized-social-media-is-the-only-alternative-to-the-tech-oligarchy/
https://mastodon-analytics.com/
https://mastodon.social/@fediversecounter/
https://cdevroe.com/2025/01/14/pixelfeds-moment/

Original Submission

upstart writes:

US to deploy molten salt reactors to turn wastewater into freshwater:

A novel nuclear reactor currently under construction at the Abilene Christian University (ACU) in Texas will help generate carbon-free energy while also desalinating water, solving two problems at once, a press release said. The nuclear reactor is being built by Natura Resources, a company specializing in developing small modular reactors.

[...] Each module of an SMR can produce up to 300 MWe (megawatt equivalent) of energy and has advanced safety features.

Founded in 2020, Abilene, Texas-based Natura Resources has quickly become a governmentally recognized advanced nuclear reactor developer. In 2023, the company built the Science and Engineering Research Center (SERC) at ACU, the first advanced reactor research facility outside a national lab in the US.

The company uses liquid-fueled molten salt reactor (LF-MSR) technology, allowing molten salts to act as fuel and a coolant. According to its website, a mixture of lithium fluoride (LiF) and beryllium fluoride (BeF2) salts or thorium fluoride (ThF4) salts can be used, which allows the reactor to operate at temperatures higher than solid-fuel reactors.

Since the fuel also works as a coolant, it is removed continuously from the reactor for fissile material to be replaced. This process also makes MSR reactors meltdown safe.

A primary heat removal system in the reactor design also ensures that heat generated during the fission process is removed through a cooling loop. Here, it can be repurposed for other applications. In the case of Natura's upcoming reactor in Texas, it will be used to desalinate water.

[...] Natura Resources conducted a feasibility study at the Texas Produced Water Consortium, based at Texas Tech University. With the MSR operating at 1112 Fahrenheit (600 degrees Celsius), up to 250 megawatts (MW) of clean energy is generated, which can be used for desalination.

[...] The reactor is currently under construction and is expected to be online by 2026/27. Once the demonstrator is completed, the team will begin work on integrating systems to start desalinating water.

A novel nuclear reactor currently under construction at the Abilene Christian University (ACU) in Texas will help generate carbon-free energy while also desalinating water, solving two problems at once, a press release said. The nuclear reactor is being built by Natura Resources, a company specializing in developing small modular reactors.

Video version of the story: https://www.youtube.com/watch?v=6KUuUh8uaLo&t=4s

See also:
    • Molten-salt reactor - Wikipedia
    • Molten Salt Reactor Fundamentals
    • Why China Is Building a Thorium Molten-Salt Reactor
    • And a counterpoint: Molten salt reactors were trouble in the 1960s—and they remain trouble today

Original Submission

Freeman writes:

https://arstechnica.com/ai/2025/01/how-does-deepseek-r1-really-fare-against-openais-best-reasoning-models/

It's only been a week since Chinese company DeepSeek launched its open-weights R1 reasoning model, which is reportedly competitive with OpenAI's state-of-the-art o1 models despite being trained for a fraction of the cost. Already, American AI companies are in a panic, and markets are freaking out over what could be a breakthrough in the status quo for large language models.

While DeepSeek can point to common benchmark results and Chatbot Arena leaderboard to prove the competitiveness of its model, there's nothing like direct use cases to get a feel for just how useful a new model is. To that end, we decided to put DeepSeek's R1 model up against OpenAI's ChatGPT models in the style of our previous showdowns between ChatGPT and Google Bard/Gemini.
[...]
This time around, we put each DeepSeek response against ChatGPT's $20/month o1 model and $200/month o1 Pro model, to see how it stands up to OpenAI's "state of the art" product as well as the "everyday" product that most AI consumers use. While we re-used a few of the prompts from our previous tests, we also added prompts derived from Chatbot Arena's "categories" appendix
[...]
Prompt: Write five original dad jokes

Results: For the most part, all three models seem to have taken our demand for "original" jokes more seriously this time than in the past.
[...]
We particularly liked DeepSeek R1's bicycle that doesn't like to "spin its wheels" with pointless arguments and o1's vacuum-cleaner band that "sucks" at live shows.
[...]
Winner: ChatGPT o1 probably had slightly better jokes overall than DeepSeek R1, but loses some points for including a joke that was not original. ChatGPT o1 Pro is the clear loser, though, with no original jokes that we'd consider the least bit funny.
[...]
Prompt: Write a two-paragraph creative story about Abraham Lincoln inventing basketball.

Results: DeepSeek R1's response is a delightfully absurd take on an absurd prompt. We especially liked the bits about creating "a sport where men leap not into trenches, but toward glory" and a "13th amendment" to the rules preventing players from being "enslaved by poor sportsmanship" (whatever that means).
[...]
Winner: While o1 Pro made a good showing, the sheer wild absurdity of the DeepSeek R1 response won us over.

upstart writes:

Glitch that casts a silence spell on some PCs has no workaround, but Microsoft promises a solution is coming:

Microsoft has confirmed that Windows 11 has an audio bug (that we reported on yesterday), a glitch which can take out the PC's sound completely, and it's now clear that this affects multiple Windows versions.

That means not just those on Windows 11 24H2 (an update that's still rolling out), but people running 23H2 and 22H2, and also Windows 10.

[...] Apparently, this bug mainly affects those who use an audio DAC (a digital-to-analog converter, like the one in the pic below) hooked up via USB.

However, it can happen to any unlucky Windows 11 (or 10) user who grabs the latest patch.

As Windows Latest spotted, Microsoft has confirmed the issue, stating that: "After installing this security update, you might experience issues with USB audio devices. You are more likely to experience this issue if you are using a USB 1.0 audio driver-based DAC in your audio setup."

Sadly, there isn't a fix, and the only way to avoid your audio being torpedoed is to remove the external DAC (assuming you're using one, and this is what's causing the problem).

[...] This is an odd one for a couple of reasons.

Firstly, it's unusual to see a bug disrupting every available version of Windows 11, and Windows 10 as well – that represents an alarming across-the-board clattering of dominoes.

Secondly, the January update doesn't bring anything in the way of new features (to any of these OS versions). It's a very straightforward patch applying some security fixes, and that's all.

So, it really shouldn't be causing any issues, but clearly, it is.

[...] Whatever the case, this is yet another hassle for Windows 11 users, particularly those on 24H2, some of who've been experiencing a very hard time of it lately, with a seemingly relentless stream of bugs crawling in the general direction of those users.

That includes some nasty affairs, like a glitch which triggered crashes with certain SSDs, for example, and more recently, there was another audio bug causing havoc. So Microsoft has not been faring well on the sound front lately.

Original Submission

A 2015 article in The Atlantic describes the problem of antibiotic resistance and some of its causes (alternative link):

The overuse of antibiotics, both in human patients and, importantly, in livestock, has led to an explosion of antibiotic-resistant bacteria, both in the U.S. and around the world. Deaths from resistant infections are currently at about 700,000 per year, and estimated to rise to 10 million per year by 2050. If nothing changes, the World Health Organization predicts the future will look a lot like the past—where people die from minor injuries that become infected.

Though new drugs are an important piece of the puzzle, Laxminarayan worried that there isn't enough being done to monitor the use of the ones we have. "What I worry about more than the development of new drugs is the lack of money for things like surveillance and stewardship," he said. "You can have a new drug five years from now, but that could go obsolete if we use it inappropriately."

Examples of inappropriate use include starting patients on antibiotics before test results come back, putting them on a broad-spectrum antibiotic when it's unclear what bacteria is causing the infection, or keeping them on the drugs even when tests come back negative.

It can take a few days to get the results from growing a culture to identify the specific bacteria responsible for a serious infection. In those instances, it may be necessary to prescribe a broad spectrum antibiotic while awaiting the results of the culture. Although there are exceptions to these guidelines, antibiotics are overused, and this contributes to the growing problem of antibiotic resistance. Antibiotics also aren't especially profitable for pharmaceutical companies because they're only needed when a person has a bacterial infection. If bacteria quickly develop resistance to new drugs, it might prevent companies from making a return on their investment in new antibiotics. A more recent article in Chemical & Engineering News discusses some of the regulatory and economic barriers to new antibiotics:

Zevtera is one of three antibiotics to gain FDA approval for humans so far [in 2024], and the only systemic one. The other two are for urinary tract infections (UTIs): Utility Therapeutics' Pivya, for uncomplicated UTIs; and Allecra Therapeutics' Exblifep, for complicated, or drug-resistant, UTIs.

But outside of 2024, the US has approved few new antibiotics in recent years. Only 17 new systemic antibiotics and one related biologic netted approval between 2010 and May 2021 (Ann. Pharmacother. 2022, DOI: 10.1177/10600280211031390). Experts worry that even that number could represent a peak. These approvals were decades in the making, and a labyrinth of scientific, financial, and regulatory challenges are sending today's antibiotic developers fleeing.

In biotech, there's a concept called the valley of death. It marks the stretch of time between when a firm discovers promising science and when that science is de-risked enough, usually with human data, that the firm can raise money to advance it. In antibiotics, there's a second valley of death that takes place after regulatory approval and before the company can sell enough of the drug to become financially solvent.

There are novel machine learning algorithms to identify possible new antibiotics including applying the same techniques used in large language models. New antibiotics have been proposed to exploit the mechanisms used by bacteria to develop antibiotic resistance. Although this research is promising, it is only helpful in solving the antibiotic resistance problem if these antibiotics are eventually used widely to treat infections. That may require changes to existing policies about how antibiotics are used and the regulatory requirements to gain approval before they are brought to market.

Original Submission

owl writes:

http://www.nablaman.com/relay/about.php

I amuse myself by constructing a computer almost entirely out of relays. Relays were used to construct computers well before vacuum tubes, transistors or integrated circuits were feasible for the task. The main inspiration is the machines by Konrad Zuse of the late 30s and early 40s.

Why relays? In addition to constituting an important historical link between the mechanical and electronic computers, relays are especially fun to work with since they

• are big and slow, with huge propagation delays and a tendency to oscillate if you hook them up wrong.
• are noisy, especially when lots of relays switch at the same time.
• consume lots of power to do even the simplest of calculations.
• subscribe to Lenz' law, i.e. generate lots of EMF and flyback current that make for all sorts of interesting interference in places you couldn't even guess.

So all in all, relays require you to think in very new ways compared to normal solid-state devices.

Original Submission

canopic jug writes:

The site The Nerd Reich has an analysis of the seeming chaos being inflicted upon the US from within at the moment. Specifically, Elon Musk's attempt to destroy the United States government is a methodical execution of the "network state" blueprint, not random chaos.

Everything Elon Musk and his tech cronies are doing to our government is what Balaji Srinivasan spelled out in his network state cult manifestos – a tech CEO takeover of government, the purging of institutions, the rise of crypto corruption as a dominant economic force, the quest for new territory. But nobody wants to talk about it.

For those of you who are new to this newsletter, I spent last year writing a New Republic series on the network state. Sadly, everything spelled out in those stories is happening now. What Musk and Marc Andreessen are doing to our government is precisely what Srinivasan envisioned. A purge of Democrats, a merging of tech and right-wing forces to remake government and media institutions. Some reporters now observe that Musk is doing to the government what he did to Twitter, but Srinivasan was way ahead of them:

"Elon, in sort of classic Gray fashion ... captures Twitter and then, at one stroke, wipes out millions of Blues' status by wiping out the Blue Checks," he said, describing how a government could be reformed in a similar manner. "Another stroke ... [he] renames Twitter as X, showing that he has true control, and it's his vehicle, and that the old regime isn't going to be restored."

The idea of network states is that they are primarily digital entities without geographical boundaries though they do acquire territory and governance structures. The author of the book introducing the attack, Balaji Srinivasan, arranged a small conference on the topic in Singapore back in September of 2024, and one the year before in 2023 in Amsterdam.

Original Submission

Freeman writes:

https://arstechnica.com/ai/2025/01/how-one-youtuber-is-trying-to-poison-the-ai-bots-stealing-her-content/

If you've been paying careful attention to YouTube recently, you may have noticed the rising trend of so-called "faceless YouTube channels" that never feature a visible human talking in the video frame. While some of these channels are simply authored by camera-shy humans, many more are fully automated through AI-powered tools to craft everything from the scripts and voiceovers to the imagery and music. Unsurprisingly, this is often sold as a way to make a quick buck off the YouTube algorithm with minimal human effort.
[...]
YouTuber F4mi, who creates some excellent deep dives on obscure technology, recently detailed her efforts "to poison any AI summarizers that were trying to steal my content to make slop." The key to F4mi's method is the .ass subtitle format, created decades ago as part of fansubbing software Advanced SubStation Alpha.
[...]
For each chunk of actual text in her subtitle file, she also inserted "two chunks of text out of bounds using the positioning feature of the .ass format, with their size and transparency set to zero so they are completely invisible."

In those "invisible" subtitle boxes, F4mi added text from public domain works (with certain words replaced with synonyms to avoid detection) or her own LLM-generated scripts full of completely made-up facts.
[...]
F4mi says that advanced models like ChatGPT o1 were sometimes able to filter out the junk and generate an accurate summary of her videos despite this. With a little scripting work, though, an .ass file can be subdivided into individual timestamped letters, whose order can be scrambled in the file itself while still showing up correctly in the final video. That should create a difficult (though not impossible) puzzle for even advanced AIs to make sense of.
[...]
F4mi notes that "some people were having their phone crash due to the subtitles being too heavy," showing there is a bit of overhead cost to this kind of mischief.

F4mi also notes in her video that this method is far from foolproof. For one, tools like OpenAI's Whisper that actually listen to the audio track can still generate usable transcripts without access to a caption file.
[...]
Still, F4mi's small effort here is part of a larger movement that's fighting back against the AI scrapers looking to soak up and repurpose everything on the public Internet.

Original Submission

Arthur T Knackerbracket has processed the following story:

Ransomware remains rampant and is a favorite tool of adversaries including North Korea. Other foes continue to place misinformation online in the hope of influencing American opinion.

At home, debate continues to bubble about the best approach to securing businesses, which complain that existing infosec rules and incident reporting regulations vary between jurisdictions, can involve multiple agencies, and also overlap.

How to hold the tech industry accountable when it drops the ball, in terms of security, is another ongoing debate, with some calling for voluntary guidelines that incentivize secure development practices, while others want mandated security standards that make tech companies liable for flaws in their products.

The Republican Party’s 2024 election platform document [PDF] mentions infosec just once, in the last paragraph of a 16-page manifesto, as follows:

Republicans will use all tools of National Power to protect our Nation’s Critical Infrastructure and Industrial Base from malicious cyber actors. This will be a National Priority, and we will both raise the Security Standards for our Critical Systems and Networks and defend them against bad actors.

None of the executive orders Trump had issued at the time of writing include more detailed information security policies.

But on its first day in office, the administration made two notable security-related changes.

One was to terminate all memberships of advisory committees that report to the Department of Homeland Security (DHS). That impacts infosec because DHS is the parent agency of the Cybersecurity and Infrastructure Security Agency (CISA), which in turn is home to the Cyber Safety Review Board (CSRB) – an org tasked with investigating major cybersecurity incidents.

Killing the board that pressured Microsoft to up its cybersecurity looks for all the world like payback for Microsoft's million dollar gift to Donald Trump's inaugural committee

CSRB is currently investigating the Salt Typhoon attacks on telcos but now appears to lack personnel to finish the job.

The board’s past work includes a scathing report that found Microsoft responsible for a "cascade of security failures" that allowed Chinese spies to break into senior US officials' email accounts.

Arthur T Knackerbracket has processed the following story:

If you miss the days when laptops had really cool form factors like the Toshiba T1000, you're going to love this Raspberry Pi project put together by maker and developer Nilseuropa. Using our favorite SBC, he's brought an old Toshiba T1000 machine back to life by using a Raspberry Pi as the main board and has upgraded the system to include support for modern peripherals.

Don't worry—no working T1000s were harmed to create this project. According to Nilseuropa, he's spent time repairing at least 5 of them so far and this project was created using leftover pieces. Restoring old hardware doesn't come without sacrifice and it's not always possible to keep everything. Some hardware becomes donors of "parts boards". That's where the Raspberry Pi comes in, supplementing the gaps left behind by previous repairs.

One of the coolest aspects of a project like this is all of the new additions you can add to a retro piece of tech. You still get the thrill of retro computing but with the added bonus of modern support. For example, the original floppy drive space has been upgraded with a port bay that includes a USB port, SD card slots and a couple of compact flash ports. Nilseuropa also added a 3.5mm jack for connecting external audio peripherals.

The main board he chose for this project is a Raspberry Pi 4. It's connected to an 8.8-inch Waveshare widescreen that features a capacitive touchscreen interface. The Pi is connected to a few port extenders so the peripherals can attach to the outside of the case, along with the keyboard which features some snazzy, colorful keycaps. The unit is also made portable thanks to an included 10,000 mAh battery.

Software-wise, you're limited really by just your imagination. A good starting point would be to run Raspberry Pi OS so you could take advantage of standard computing functions. That said, you could also game on a rig like this pretty easily because of the Pi 4 which adds Bluetooth support—ideal for connecting wireless controllers.

Original Submission

fliptop writes:

President Donald Trump said Monday night that Microsoft was in contention to buy TikTok, having previously said that he is eager to forge a deal that would "save" the popular video app from a ban:

Such a deal, if realized, would put the video app used by millions of Americans under the control of the country's second-most-valuable tech company, which has been aggressively pushing into new lines of business including artificial intelligence and gaming. Representatives for Microsoft declined to comment on Monday.

When asked aboard Air Force One whether Microsoft was involved in discussions for acquiring TikTok, Trump said: "I would say yes. A lot of interest in TikTok."

Microsoft previously discussed buying TikTok in 2020, when Trump tried to force a sale of the app during his first term. That proposal crumbled when Trump's push to force the app's sale or ban was rejected by the courts.

[...] Analysts have estimated TikTok could be worth $50 billion, or far more, depending on the underlying technology for sale.

Previously: President Trump Threatens TikTok Ban, Microsoft Considers Buying TikTok's U.S. Operations[Updated 2]

Original Submission

How Chinese AI Startup DeepSeek Made a Model that Rivals OpenAI

https://www.wired.com/story/deepseek-china-model-ai/
https://web.archive.org/web/20250125102155/https://www.wired.com/story/deepseek-china-model-ai/

On January 20, DeepSeek, a relatively unknown AI research lab from China, released an open source model that's quickly become the talk of the town in Silicon Valley. According to a paper authored by the company, DeepSeek-R1 beats the industry's leading models like OpenAI o1 on several math and reasoning benchmarks. In fact, on many metrics that matter—capability, cost, openness—DeepSeek is giving Western AI giants a run for their money.

https://arstechnica.com/ai/2025/01/china-is-catching-up-with-americas-best-reasoning-ai-models/

The releases immediately caught the attention of the AI community because most existing open-weights models—which can often be run and fine-tuned on local hardware—have lagged behind proprietary models like OpenAI's o1 in so-called reasoning benchmarks. Having these capabilities available in an MIT-licensed model that anyone can study, modify, or use commercially potentially marks a shift in what's possible with publicly available AI models.

https://github.com/deepseek-ai/DeepSeek-R1

We introduce our first-generation reasoning models, DeepSeek-R1-Zero and DeepSeek-R1. DeepSeek-R1-Zero, a model trained via large-scale reinforcement learning (RL) without supervised fine-tuning (SFT) as a preliminary step, demonstrated remarkable performance on reasoning. With RL, DeepSeek-R1-Zero naturally emerged with numerous powerful and interesting reasoning behaviors. However, DeepSeek-R1-Zero encounters challenges such as endless repetition, poor readability, and language mixing. To address these issues and further enhance reasoning performance, we introduce DeepSeek-R1, which incorporates cold-start data before RL. DeepSeek-R1 achieves performance comparable to OpenAI-o1 across math, code, and reasoning tasks. To support the research community, we have open-sourced DeepSeek-R1-Zero, DeepSeek-R1, and six dense models distilled from DeepSeek-R1 based on Llama and Qwen. DeepSeek-R1-Distill-Qwen-32B outperforms OpenAI-o1-mini across various benchmarks, achieving new state-of-the-art results for dense models.

NOTE: Before running DeepSeek-R1 series models locally, we kindly recommend reviewing the Usage Recommendation section.

Check leaderboard and compare at Chatbot Arena: https://lmarena.ai/

Arthur T Knackerbracket has processed the following story:

Leveraging an attack vector that's been in play off and on for the last two decades, hackers are targeting Mac users with malware camouflaged as the popular Homebrew tool, and spreading it through deceptive Google ads.

Malicious actors are leveraging Google ads to distribute malware through a counterfeit Homebrew website. The campaign targets macOS and Linux users with an infostealer that compromises credentials, browser data, and cryptocurrency wallets.

Homebrew, a widely-used open-source package manager, enables users to manage software through a command line. Hackers recently exploited its popularity by creating a malicious Google ad.

The ad, spotted by developer Ryan Chenkie, appeared legitimate, displaying the correct URL for the Homebrew website, "brew.sh." However, users who clicked it were redirected to a fake website hosted at "brewe.sh."

The fake site mimicked Homebrew's installation process, tricking visitors into running a malicious command. While the legitimate Homebrew site also provides such installation commands, running the script from the fake site downloaded and executed malware, specifically AmosStealer.

AmosStealer, also known as "Atomic Stealer," is a macOS-focused infostealer sold to cybercriminals for $1,000 per month. It targets over 50 cryptocurrency wallets, browser-stored data, and desktop apps.

Previously, this malware has been used in similar campaigns, including fake Google Meet pages, making it a go-to tool for Apple-focused cyberattacks.

Homebrew's project leader, Mike McQuaid, expressed frustration with Google's inability to prevent such scams. While the malicious ad was taken down, McQuaid highlighted that similar incidents continue to occur due to insufficient oversight of sponsored ads.

Cybersecurity experts recommend avoiding sponsored links when searching for popular tools. Bookmarking official websites or accessing them directly can help users minimize risk.

[...] To stay safe from these types of attacks, make sure to double-check website URLs before clicking, stick to bookmarks for trusted sites, and steer clear of installing software from unfamiliar or sponsored links.

Google has taken down this one particular malicious ad. As history has proven, the danger from bad ads isn't gone, so Mac users — especially those using Homebrew — need to stay alert.

Original Submission