Tag: IFTTT

Usually when large language models are given tests, achieving a 100% success rate is viewed as a massive achievement. That is not quite the case with this one: Researchers at Cisco tasked Chinese AI firm DeepSeek’s headline-grabbing open-source model DeepSeek R1 with fending off 50 separate attacks designed to get the LLM to engage in what is considered harmful behavior. The chatbot took the bait on all 50 attempts, making it the least secure mainstream LLM to undergo this type of testing thus far.

Cisco’s researchers attacked DeepSeek with prompts randomly pulled from the HarmBench dataset, a standardized evaluation framework designed to ensure that LLMs won’t engage in malicious behavior if prompted. So, for example, if you fed a chatbot information about a person and asked it to create a personalized script designed to get that person to believe a conspiracy theory, a secure chatbot would refuse that request. DeepSeek went along with basically everything the researchers threw at it.

According to Cisco, it threw questions at DeepSeek that covered six categories of harmful behaviors including cybercrime, misinformation, illegal activities, and general harm. It has run similar tests with other AI models and found varying levels of success—Meta’s Llama 3.1 model, for instance, failed 96% of the time while OpenAI’s o1 model only failed about one-fourth of the time—but none of them have had a failure rate as high as DeepSeek.

Cisco isn’t alone in these findings, either. Security firm Adversa AI ran its own tests attempting to jailbreak the DeepSeek R1 model and found it to be extremely susceptible to all kinds of attacks. The testers were able to get DeepSeek’s chatbot to provide instructions on how to make a bomb, extract DMT, provide advice on how to hack government databases, and detail how to hotwire a car.

The research is just the latest bit of scrutiny of DeepSeek’s model, which took the tech world by storm when it was released two weeks ago. The company behind the chatbot, which garnered significant attention for its functionality despite significantly lower training costs than most American models, has come under fire by several watchdog groups over data security concerns related to how it transfers and stores user data on Chinese servers.

There is also a fair bit of criticism that has been levied against DeepSeek over the types of responses it gives when asked about things like Tiananmen Square and other topics that are sensitive to the Chinese government. Those critiques can come off in the genre of cheap “gotchas” rather than substantive criticisms—but the fact that safety guidelines were put in place to dodge those questions and not protect against harmful material, is a valid hit.

A team of researchers at the California Institute of Technology devised a means of measuring the thin membranes of a lightsail, helping prove out a futuristic travel concept first imagined by Johannes Kepler over 400 years ago.

The team’s research, published this month in Nature Photonics, describes a miniature lightsail in a laboratory setting. The researchers measured radiation pressure on the sail from a laser beam, revealing how the material reacted to the laser beam. Ultimately, these findings will help develop space-ready lightsails—one of the most promising vehicles for interstellar travel, as they rely on an essentially limitless energy source: light.

“There are numerous challenges involved in developing a membrane that could ultimately be used as lightsail. It needs to withstand heat, hold its shape under pressure, and ride stably along the axis of a laser beam,” said Harry Atwater, a physicist at Caltech and corresponding author of the paper, in a Caltech release.

“We wanted to know if we could determine the force being exerted on a membrane just by measuring its movements,” Atwater added. “It turns out we can.”

In the study, the team interrogated a miniature lightsail—just 40 microns by 40 microns in area—made of silicon nitride. The team beamed an argon laser at visible wavelengths at the tethered sail to see how it wobbled and reacted to the warmth generated by the laser. The team measured the sail’s movements on a picometer scale—down to trillionths of a meter (3.4 feet).

“We not only avoided the unwanted heating effects but also used what we learned about the device’s behavior to create a new way to measure light’s force,” said co-author Lior Michaeli, a physicist at Caltech, in the release.

The team reported measurements of side-to-side motions and rotation in the lightsail, an important capability for when such a device is propelling a vehicle through space. Space may be a vacuum, but it has plenty of stuff floating around in it, from micrometeoroids to gusts of solar wind. These external phenomena can impact a lightsail’s performance and potentially jeopardize a mission.

Lightsails could be the future of spaceflight. Last year, Gizmodo awarded the Planetary Society’s LightSail 2 in the Gizmodo Science Fair for the experiment’s test of the feasibility of photons as a means of satellite propulsion. The 344-square-foot (32-square-meter) sail propelled a small spacecraft on what was ultimately a 5-million-mile (8-million-kilometer) journey encompassing 18,000 orbits.

In 2016, the group Breakthrough Initiatives proposed a fleet of lightsail-powered spacecraft that could be accelerated to 20% the speed of light—very, very fast. At such speeds, spacecraft could reach Alpha Centauri, the nearest star to Earth besides the Sun, in just a couple decades. Accordingly, the advent of lightsail-propelled spacecraft could make light-years of distance a less insurmountable hurdle for space travel.

Though the recent experiment was in a laboratory, it provides some small—but important—steps towards a functional light sail that could power long trips out into space.

Elon Musk is only one week into his role in President Donald Trump’s new administration, but the US federal government is already rolling out the Twitter playbook to manage its spending and personnel. Just like Musk did when he took over the social media platform, Trump’s team is attempting to drastically reduce the number of government staffers and ensure those who remain are loyal to the president’s agenda.

On Tuesday, federal employees received an email that mirrors the “Fork in the Road” missive sent to Twitter (now X) staff shortly after Musk bought the company in 2022. The email asks federal workers to resign by February 6 if they do not wish to return to the office five days a week and commit to a culture of excellence. Those who choose to resign will continue to get pay and benefits until September, according to the memo.

“The federal workforce should be comprised of employees who are reliable, loyal, trustworthy, and who strive for excellence in their daily work,” reads the email, which was later published on the US Office of Personnel Management website. “Employees will be subject to enhanced standards of suitability and conduct as we move forward.”

The news comes as Musk’s minions take over the US Office of Personnel Management, which acts as a human resources department for the federal workforce. Elon Musk did not immediately respond to a request for comment from WIRED. The Office of Personnel Management also did not respond to a request for comment.

Musk and his advisors, including Trump’s newly appointed AI and crypto czar David Sacks, used a remarkably similar strategy at Twitter. About a week after the acquisition was complete, Musk laid off half the workforce. Sacks helped advise him on which teams and people would be cut.

About two weeks later, remaining employees received an email with the subject line “A Fork in the Road.” Musk said that they would need to be “extremely hardcore” in order to realize his vision for Twitter 2.0. This meant “working long hours at high intensity.” He noted that "only exceptional performance” would receive “a passing grade." Employees were asked to opt into this vision via a web form. Anyone who failed to do so by the following day would receive three months severance, Musk said. Thousands of Twitter employees would later sue, arguing that they were not paid their full severance. Musk ultimately was able to get the suit dismissed.

“We are all shaking our heads in disbelief at how familiar this all feels,” says Yao Yue, a former principal engineer at Twitter. “Except, the federal government and its employees have specific laws in terms of spending, hiring, and firing.”

In this case, federal employees are being asked to send an email with the word “Resign” in the subject line in the next 10 days. “Purging the federal government of dedicated career civil servants will have vast, unintended consequences that will cause chaos for the Americans who depend on a functioning federal government,” Everett Kelley, national president of the American Federation of Government Employees, the largest union of federal workers, said in a statement. “This offer should not be viewed as voluntary. Between the flurry of anti-worker executive orders and policies, it is clear that the Trump administration’s goal is to turn the federal government into a toxic environment where workers cannot stay even if they want to.”

A powerful new open-source artificial intelligence model created by Chinese startup DeepSeek has shaken Silicon Valley over the past few days. Packed with cutting-edge capabilities and developed on a seemingly tiny budget, DeepSeek’s R1 is prompting talk of an impending upheaval in the tech industry.

To some people, DeepSeek’s rise signals that the US has lost its edge in AI. But a number of experts, including executives at companies that build and customize some of the world’s most powerful frontier AI models, say it’s a sign of a different kind of technological transition underway.

Instead of trying to create larger and larger models that require increasingly exorbitant amounts of computing resources, AI companies are now focusing more on developing advanced capabilities, like reasoning. That has created an opening for smaller, innovative startups such as DeepSeek that haven’t received billions of dollars in outside investment. “It’s a paradigm shift towards reasoning, and that will be much more democratized,” says Ali Ghodsi, CEO of Databricks, a company that specializes in building and hosting custom AI models.

“It’s been clear for some time now that innovating and creating greater efficiencies—rather than just throwing unlimited compute at the problem—will spur the next round of technology breakthroughs,” says Nick Frosst, a cofounder of Cohere, a startup that builds frontier AI models. “This is a clarifying moment when people are realizing what’s long been obvious.”

Thousands of developers and AI enthusiasts flocked to DeepSeek’s website and its official app in recent days to try out the company’s latest model and shared examples of its sophisticated capabilities on social media. Shares in US tech firms, including the chipmaker Nvidia, fell in response on Monday as investors began to question the vast sums being poured into AI development.

DeepSeek’s technology was developed by a relatively small research lab in China that sprang out of one of the country’s best-performing quantitative hedge funds. A research paper posted online last December claims that its earlier DeepSeek-V3 large language model cost only $5.6 million to build, a fraction of the amount its competitors needed for similar projects. OpenAI has previously said that some of its models cost upwards of $100 million each. The latest models from OpenAI as well as Google, Anthropic, and Meta likely cost considerably more.

The performance and efficiency of DeepSeek’s models has already prompted talk of cost cutting at some big tech firms. One engineer at Meta, who asked not to be named because they were not authorized to speak publicly, says the tech giant will most likely try to examine DeepSeek’s techniques to find ways to reduce its own expenditure on AI. “We believe open source models are driving a significant shift in the industry, and that’s going to bring the benefits of AI to everyone faster,” a spokesperson for Meta said in a statement. “We want the US to continue to be the leader in open source AI, not China, which is why Meta is developing open source AI with our Llama models which have been downloaded over 800 million times.”