Month: September 2024

Enlarge (credit: Getty Images)

Networking hardware-maker Zyxel is warning of nearly a dozen vulnerabilities in a wide array of its products. If left unpatched, some of them could enable the complete takeover of the devices, which can be targeted as an initial point of entry into large networks.

The most serious vulnerability, tracked as CVE-2024-7261, can be exploited to “allow an unauthenticated attacker to execute OS commands by sending a crafted cookie to a vulnerable device,” Zyxel warned. The flaw, with a severity rating of 9.8 out of 10, stems from the “improper neutralization of special elements in the parameter ‘host’ in the CGI program” of vulnerable access points and security routers. Nearly 30 Zyxel devices are affected. As is the case with the remaining vulnerabilities in this post, Zyxel is urging customers to patch them as soon as possible.

But wait... there’s more

The hardware manufacturer warned of seven additional vulnerabilities affecting firewall series including the ATP, USG-FLEX, and USG FLEX 50(W)/USG20(W)-VPN. The vulnerabilities carry severity ratings ranging from 4.9 to 8.1. The vulnerabilities are:

Enlarge

Wars of necessity spawn weapons innovation as each side tries to counter the other's tactics and punch through defenses. For instance—as the Russian invasion of Ukraine has made drone warfare real, both sides have developed ways to bring down drones more easily. One recent Ukrainian innovation has been building counter-drone ramming drones that literally knock Russian drones from the sky.

In the case of the trench warfare that currently dominates the Russian invasion of eastern Ukraine, the Ukrainians have another new tactic: dragon's fire. Delivered by drone.

Videos have begun to circulate on Telegram and X this week from Ukrainian units showing their new weapon. (You can see three of them below.) The videos each show a drone moving deliberately along a trench line as it releases a continuous stream of incendiary material, which often starts fires on the ground below (and ignites nearby ammunition).

Enlarge / ASIC evaluators found AI summaries were often "wordy and pointless—just repeating what was in the submission." (credit: Getty Images)

As large language models have continued to rise in prominence, many users and companies have focused on their useful ability to quickly summarize lengthy documents for easier human consumption. When Australia's Securities and Investments Commission (ASIC) looked into this potential use case, though, it found that the summaries it was able to get from the Llama2-70B model were judged as significantly worse than those provided by humans.

ASIC's proof-of-concept study (PDF)—which was run in January and February, written up in March, and published in response to a Senate inquiry in May—has a number of limitations that make it hard to generalize about the summarizing capabilities of state-of-the-art LLMs in the present day. Still, the government study shows many of the potential pitfalls large organizations should consider before simply inserting LLM outputs into existing workflows.

Keeping score

For its study, ASIC teamed up with Amazon Web Services to evaluate LLMs' ability to summarize "a sample of public submissions made to an external Parliamentary Joint Committee inquiry, looking into audit and consultancy firms." For ASIC's purposes, a good summary of one of these submissions would highlight any mention of ASIC, any recommendations for avoiding conflicts of interest, and any calls for more regulation, all with references to page numbers and "brief context" for explanation.

Enlarge (credit: BING-JHEN HONG | iStock Editorial / Getty Images Plus)

The Department of Justice is reportedly deepening its probe into Nvidia. Officials have moved on from merely questioning competitors to subpoenaing Nvidia and other tech companies for evidence that could substantiate allegations that Nvidia is abusing its "dominant position in AI computing," Bloomberg reported.

When news of the DOJ's probe into the trillion-dollar company was first reported in June, Fast Company reported that scrutiny was intensifying merely because Nvidia was estimated to control "as much as 90 percent of the market for chips" capable of powering AI models. Experts told Fast Company that the DOJ probe might even be good for Nvidia's business, noting that the market barely moved when the probe was first announced.

But the market's confidence seemed to be shaken a little more on Tuesday, when Nvidia lost a "record-setting $279 billion" in market value following Bloomberg's report. Nvidia's losses became "the biggest single-day market-cap decline on record," TheStreet reported.

Enlarge / A ratgdo, version 2.53i. (credit: Ratcloud LLC)

I live in suburbia, which means I've got a garage (or a carhole, if you’re not so fancy). It's a detached garage, so part of my nightly routine when I check to make sure the house is all locked up is to peek out the back window. I like to know the garage door is closed and our cars are tucked in safely.

But actually looking out a window with my stupid analog eyeballs is lame, so I figured I could make things easier by adding some smarts to my garage. The first thing I did was use this fellow’s instructions (the original site is sadly offline, but the Wayback Machine is forever) to cobble together a Raspberry Pi-based solution that would fire off an email every time the garage door opened or closed. I couldn’t remotely open or close the door from inside the house myself (well, I mean, I could with the actual garage door opener remote control), but I could just glance at my inbox to see if the garage door was open or shut in the evenings.

This worked great for a couple of years, until Texas summers murdered the poor Pi. (This was possibly my fault, too, because of the PoE hat that I’d slapped onto the Pi, which resulted in extra heat.) So, I was back to peeking out my window to check on the garage in the evenings. Like a sucker.

Conventional wisdom would suggest that all dogs love to play fetch, while most cats would simply refuse to do so. But a new paper published in the journal PLoS ONE suggests that cats not only play fetch with their owners, they do so far more frequently than previously believed. And while most dogs play fetch at least sometimes, about 12 percent do not. More-trainable dog breeds are more likely to exhibit fetching behavior, while in both species, animals that are more active and playful—and usually male—are more likely to enjoy playing fetch, suggesting that it is a form of play.

"We were surprised to find that there were very few studies of fetching behavior in dogs," said co-author Mikel Delgado, a veterinary medicine researcher at Purdue University. "And personally, as a life-long cat person, I have to admit that I thought all dogs fetched. So it was interesting to get a better sense of how common this behavior is in cats and dogs. We hope that the study draws more attention to fetching behavior in cats, who are often portrayed as independent and aloof. In fact, they can be very social, and this is a nice example of one way they are interactive with humans."

As previously reported, many different animal species exhibit play behavior, and it's most common in mammals and birds. Contrary to what one might expect from cats, fetching behavior has been observed across multiple cat breeds all over the world, usually emerging in kittenhood. One owner who participated in a 2022 study noted that their cat was so obsessed with fetch that it would sometimes drop its favorite toy on their face in the middle of the night.

Enlarge / Ilya Sutskever, OpenAI Chief Scientist, speaks at Tel Aviv University on June 5, 2023. (credit: JACK GUEZ via Getty Images)

On Wednesday, Reuters reported that Safe Superintelligence (SSI), a new AI startup cofounded by OpenAI's former chief scientist Ilya Sutskever, has raised $1 billion in funding. The three-month-old company plans to focus on developing what it calls "safe" AI systems that surpass human capabilities.

The fundraising effort shows that even amid growing skepticism around massive investments in AI tech that so far have failed to be profitable, some backers are still willing to place large bets on high-profile talent in foundational AI research. Venture capital firms like Andreessen Horowitz, Sequoia Capital, DST Global, and SV Angel participated in the SSI funding round.

SSI aims to use the new funds for computing power and attracting talent. With only 10 employees at the moment, the company intends to build a larger team of researchers across locations in Palo Alto and Tel Aviv, Reuters reported.