schnellbach.de

Jon Porter reports via The Verge: UK telecoms regulator Ofcom has laid out how porn sites could verify users' ages under the newly passed Online Safety Act. Although the law gives sites the choice of how they keep out underage users, the regulator is publishing a list of measures they'll be able to use to comply. These include having a bank or mobile network confirm that a user is at least 18 years old (with that user's consent) or asking a user to supply valid details for a credit card that's only available to people who are 18 and older. The regulator is consulting on these guidelines starting today and hopes to finalize its official guidance in roughly a year's time. Ofcom lists six age verification methods in today's draft guidelines. As well as turning to banks, mobile networks, and credit cards, other suggested measures include asking users to upload photo ID like a driver's license or passport, or for sites to use "facial age estimation" technology to analyze a person's face to determine that they've turned 18. Simply asking a site visitor to declare that they're an adult won't be considered strict enough. Once the duties come into force, pornography sites will be able to choose from Ofcom's approaches or implement their own age verification measures so long as they're deemed to hit the "highly effective" bar demanded by the Online Safety Act. The regulator will work with larger sites directly and keep tabs on smaller sites by listening to complaints, monitoring media coverage, and working with frontline services. Noncompliance with the Online Safety Act can be punished with fines of up to [$22.7 million] or 10 percent of global revenue (whichever is higher). The guidelines being announced today will eventually apply to pornography sites both big and small so long as the content has been "published or displayed on an online service by the provider of the service." In other words, they're designed for professionally made pornography, rather than the kinds of user-generated content found on sites like OnlyFans. That's a tricky distinction when the two kinds often sit together side by side on the largest tube sites. But Ofcom will be opening a consultation on rules for user-generated content, search engines, and social media sites in the new year, and Whitehead suggests that the both sets of rules will come into effect at around the same time.

Read more of this story at Slashdot.

schwit1 shares a report from the New York Post: A landmark study out of the University of Adelaide and University of Essex has found that living in a private rental property accelerates the biological aging process by more than two weeks every year. The research found renting had worse effects on biological age than being unemployed (adding 1.4 weeks per year), obesity (adding 1 week per year), or being a former smoker (adding about 1.1 weeks). University of Adelaide Professor of Housing Research Emma Baker said private renting added "about two-and-a-half weeks of aging" per year to a person's biological clock, compared to those who own their homes. "In fact, private rental is the really interesting thing here, because social renters, for some reason, don't seem to have that effect," Professor Baker told the ABC News Daily podcast. She said the security of social renting -- aka public housing -- and homeownership has compared to people living with an end-of-lease date on their calendars. "When you look at big studies of the Australian population, you see that the average rental lease is between six and 12 months," she said. "So even if you have your lease extended, you still are living in that slight state of kind of unknowingness, really not quite secure if your lease is actually going to be extended or not." "We think that that is one of the things that's contributing to loss of years, effectively."

Read more of this story at Slashdot.

An anonymous reader quotes a report from Wired: One of the world's most peculiar test beds stretches above Princeton, New Jersey. It's a fiber optic cable strung between three utility poles that then runs underground before feeding into an "interrogator." This device fires a laser through the cable and analyzes the light that bounces back. It can pick up tiny perturbations in that light caused by seismic activity or even loud sounds, like from a passing ambulance. It's a newfangled technique known as distributed acoustic sensing, or DAS. Because DAS can track seismicity, other scientists are increasingly using it to monitor earthquakes and volcanic activity. (A buried system is so sensitive, in fact, that it can detect people walking and driving above.) But the scientists in Princeton just stumbled upon a rather noisier use of the technology. In the spring of 2021, Sarper Ozharar -- a physicist at NEC Laboratories, which operates the Princeton test bed -- noticed a strange signal in the DAS data. "We realized there were some weird things happening," says Ozharar. "Something that shouldn't be there. There was a distinct frequency buzzing everywhere." The team suspected the "something" wasn't a rumbling volcano -- not inNew Jersey -- but the cacophony of the giant swarm of cicadas that had just emerged from underground, a population known as Brood X. A colleague suggested reaching out to Jessica Ware, an entomologist and cicada expert at the American Museum of Natural History, to confirm it. "I had been observing the cicadas and had gone around Princeton because we were collecting them for biological samples," says Ware. "So when Sarper and the team showed that you could actually hear the volume of the cicadas, and it kind of matched their patterns, I was really excited." Add insects to the quickly growing list of things DAS can spy on. Thanks to some specialized anatomy, cicadas are the loudest insects on the planet, but all sorts of other six-legged species make a lot of noise, like crickets and grasshoppers. With fiber optic cables, entomologists might have stumbled upon a powerful new way to cheaply and constantly listen in on species -- from afar. "Part of the challenge that we face in a time when there's insect decline is that we still need to collect data about what population sizes are, and what insects are where," says Ware. "Once we are able to familiarize ourselves with what's possible with this type of remote sensing, I think we can be really creative."

Read more of this story at Slashdot.

According to China Daily, China has become the world's first nation to deploy a commercial data center underwater. Interesting Engineering reports: China's attempts to set up a commercial data center underwater are the result of a public-private enterprise involving the China Offshore Oil Engineering Co., the country's largest Engineering, Procurement, Construction, and Installation (EPCI) company in the country, and Highlander, a private data center company. Although details of the computing hardware have not been shared, Highlander has claimed that each of its underwater modules is capable of processing over four million high-definition (HD) images in just 30 seconds. The computing hardware is packed inside a watertight storage module and together weighs 1,300 tons. The module is being submerged about 115 feet (35 m) under the water, a process that takes about three hours. Although work on installing the first module has begun, Highlander has ambitious plans to install 100 such modules at the site and build a capacity of nearly six million computers working at a time. Such a staggering number of computers will also generate a lot of heat which will be naturally cooled by the surrounding sea water. This alone is expected to save 122 million kilowatt-hours of electricity that would have otherwise been spent on cooling if the facility were located on land. Additionally, the facility, which is expected to be in place by 2025, will also save 732,000 square feet (68,000 square meters) of terrestrial land that can be used for other purposes and 105,000 tons of fresh water, which would be used for cooling efforts. The modules have been built to last 25 years, but a lot remains unknown about how the construction will be impacted by corrosive seawater and underwater ecosystems. Highlander's experience in setting these centers up is fairly limited to the tests it carried out in January of 2021 in the Guangdong port of Zhuhai.

Read more of this story at Slashdot.

Trevor Jacob, a daredevil YouTuber who deliberately crashed a plane for views in a moneymaking scheme, has been sentenced to six months in federal prison. Jacob posted a video of himself in 2021 parachuting out of a plane that he claimed had malfunctioned. In reality, the aircraft was purposely abandoned and crashed into the Los Padres National Forest in Southern California. From a report: Jacob pleaded guilty to one felony count of destruction and concealment with the intent to obstruct a federal investigation on June 30. "It appears that (Jacob) exercised exceptionally poor judgment in committing this offense," prosecutors said in the release. "(Jacob) most likely committed this offense to generate social media and news coverage for himself and to obtain financial gain. Nevertheless, this type of 'daredevil' conduct cannot be tolerated." Jacob received a sponsorship from a company and had agreed to promote the company's wallet in the YouTube video that he would post. [...] The release said Jacob lied to federal investigators when he filed a report that falsely indicated his plane lost full power approximately 35 minutes into the flight. He also lied to a Federal Aviation Administration aviation safety inspector when he said he had parachuted out of the plane when the airplane's engine had quit because he could not identify any safe landing options.

Read more of this story at Slashdot.

Rockstar Games has officially revealed the trailer for Grand Theft Auto VI, which is coming in 2025. You can watch it on YouTube. IGN reports: The trailer gives us a look at the game's female protagonist, a first for the series. Her name is Lucia, and that she starts off in prison -- "bad luck, I guess," she quips. The trailer confirms, too, that it's set in Vice City with a large sign - not a huge surprise for those who've been following the series, but exciting nonetheless. In addition to lots and lots of quick shots of crime, we also get glimpses of TikToks and live-streams, hinting that social media will be a big part of this game. It all takes place as Tom Petty's "Love Is a Long Road" plays in the background, appropriate for the many car-related crimes we see. And yes, true to the Florida setting, there are alligators in locations where they shouldn't be. It ends by showing us a little more of Lucia and a male character, seemingly both lovers and partners-in-crime. "The only way we're gonna get through this is by sticking together and being a team," Lucia says at one point. Fans have been talking about GTA 6 ever since GTA 5 was released in 2013, perhaps unsurprisingly as IGN deemed that one a "masterpiece" in our review.

Read more of this story at Slashdot.

At the Reagan National Defense Forum in Simi Valley, California, on Saturday, US Commerce Secretary Gina Raimondo issued a cautionary statement to Nvidia, urging them to stop redesigning AI chips for China that maneuver around export restrictions. "We cannot let China get these chips. Period," she said. "We're going to deny them our most cutting-edge technology." Fortune reports: Raimondo said American companies will need to adapt to US national security priorities, including export controls that her department has placed on semiconductor exports. "I know there are CEOs of chip companies in this audience who were a little cranky with me when I did that because you're losing revenue," she said. "Such is life. Protecting our national security matters more than short-term revenue." Raimondo called out Nvidia Corp., which designed chips specifically for the Chinese market after the US imposed its initial round of curbs in October 2022. "If you redesign a chip around a particular cut line that enables them to do AI, I'm going to control it the very next day," Raimondo said. Communication with China can help stabilize ties between the two countries, but "on matters of national security, we've got to be eyes wide open about the threat," she said. "This is the biggest threat we've ever had and we need to meet the moment," she said. Further reading: Nvidia CEO Says US Will Take Years To Achieve Chip Independence

Read more of this story at Slashdot.

An anonymous reader quotes a report from Ars Technica: US Senator Edward Markey (D-Mass.) is one of the more technologically engaged of our elected lawmakers. And like many technologically engaged Ars Technica readers, he does not like what he sees in terms of automakers' approach to data privacy. On Friday, Sen. Markey wrote to 14 car companies with a variety of questions about data privacy policies, urging them to do better. As Ars reported in September, the Mozilla Foundation published a scathing report on the subject of data privacy and automakers. The problems were widespread -- most automakers collect too much personal data and are too eager to sell or share it with third parties, the foundation found. Markey noted (PDF) the Mozilla Foundation report in his letters, which were sent to BMW, Ford, General Motors, Honda, Hyundai, Kia, Mazda, Mercedes-Benz, Nissan, Stellantis, Subaru, Tesla, Toyota, and Volkswagen. The senator is concerned about the large amounts of data that modern cars can collect, including the troubling potential to use biometric data (like the rate a driver blinks and breathes, as well as their pulse) to infer mood or mental health. Sen. Markey is also worried about automakers' use of Bluetooth, which he said has expanded "their surveillance to include information that has nothing to do with a vehicle's operation, such as data from smartphones that are wirelessly connected to the vehicle." "These practices are unacceptable," Markey wrote. "Although certain data collection and sharing practices may have real benefits, consumers should not be subject to a massive data collection apparatus, with any disclosures hidden in pages-long privacy policies filled with legalese. Cars should not -- and cannot -- become yet another venue where privacy takes a backseat." The 14 automakers have until December 21 to answer Markey's questions.

Read more of this story at Slashdot.

ExpressVPN is the biggest VPN company so far to take advantage of the VPN support available in tvOS 17. According to The Verge, ExpressVPN will let Apple TV users connect to servers "in any of 105 countries around the world" so they can watch geo-restricted content around the world. From the report: To download it, you'll need to make sure you're on tvOS 17 -- earlier versions don't support native VPN apps at all. Once set up, the app will route your traffic through faraway servers before forwarding them to whatever streaming service or other internet server the Apple TV contacts. ExpressVPN on the Apple TV uses the company's Lightway protocol. Reddit users reported spotting the app last week. Most said they could switch countries to get around region restrictions, though some had issues logging in or getting it to work with specific apps. It's also a basic experience that lacks advanced VPN features like split tunneling, which dynamically applies the VPN connection to certain services as needed, freeing users from managing it manually.

Read more of this story at Slashdot.

GiveDirectly, a nonprofit providing cash assistance to low-income households, is conducting a large-scale basic income experiment in rural Kenya, giving varying payment structures to recipients. "It is giving around 6,000 people in rural Kenya a little more than $20 a month, every month, starting in 2016 and going until 2028," reports Vox's Dylan Matthews. "Tens of thousands more people are getting shorter-term or differently structured payments." Matthews reports on some of the early findings of the experiment: The latest research on the GiveDirectly pilot, done by MIT economists Tavneet Suri and Nobel Prize winner Abhijit Banerjee, compares three groups: short-term basic income recipients (who got the $20 payments for two years), long-term basic income recipients (who get the money for the full 12 years), and lump sum recipients, who got $500 all at once, or roughly the same amount as the short-term basic income group. The paper is still being finalized, but Suri and Banerjee shared some results on a call with reporters this week. By almost every financial metric, the lump sum group did better than the monthly payment group. Suri and Banerjee found that the lump sum group earned more, started more businesses, and spent more on education than the monthly group. "You end up seeing a doubling of net revenues" -- or profits from small businesses -- in the lump sum group, Suri said. The effects were about half that for the short-term $20-a-month group. The explanation they arrived at was that the big $500 all at once provided valuable startup capital for new businesses and farms, which the $20 a month group would need to very conscientiously save over time to replicate. "The lump sum group doesn't have to save," Suri explains. "They just have the money upfront and can invest it." Intriguingly, the results for the long-term monthly group, which will receive about $20 a month for 12 years rather than two, had results that looked more like the lump sum group. The reason, Suri and Banerjee find, is that they used rotating savings and credit associations (ROSCAs). These are institutions that sprout up in small communities, especially in the developing world, where members pay small amounts regularly into a common fund in exchange for the right to withdraw a larger amount every so often. "It converts the small streams into lump sums," Suri summarizes. "We see that the long-term arm is actually using ROSCAs. A lot of their UBI is going into ROSCAs to generate these lump sums they can use to invest." [...] As you might expect, given how entrepreneurially minded the recipients are, the researchers found no evidence that any of the payments discouraged work or increased purchases of alcohol -- two common criticisms of direct cash giving. In fact, so many people who used to work for wages instead started businesses that there was less competition for wage work, and overall wages in villages rose as a result. And they found one major advantage for monthly payments over lump sum ones, despite the big benefits of lump sum payments for business formation. People who got monthly checks were generally happier and reported better mental health than lump sum recipients. [...] I think this points to the takeaway from this research not being "just give people a lump sum no matter what." Ideally, you could ask specific people how they would prefer to get money. ... [L]ong-term monthly payments seem to offer the best of all worlds because they enable people to use ROSCAs to generate lump sum payments when they want them. That enables flexibility: People who want monthly payments can get them, and people who need cash upfront can organize with their peers to get that.

Read more of this story at Slashdot.

An anonymous reader quotes a report from XDA Developers: A few days ago, we spotted that the HP Smart App was being installed on people's PCs without their consent. Even worse, the app would reappear if users tried to uninstall it or clean-installed Windows. Now, the cause has finally been identified: a recent Windows 10 and 11 update is renaming everyone's printers to "HP LaserJet M101-M106" regardless of what model it actually is. As reported on Windows Latest, the latest update for Windows 10 and 11 seems to think that people's printers are an HP LaserJet model, regardless of their actual brand. It's believed that the bug appeared after HP pushed its latest metadata to Windows Update, but something went awry in the code and caused other printers to be labeled as HP LaserJet printers. This explains why the HP Smart App has been sneaking onto people's computers without their consent. A key part of Windows Update is keeping third-party drivers and devices updated, including downloading any apps that the devices depend on. After the printer metadata incorrectly identified everyone's printers as HP LaserJet printers, Windows installed all the software needed for an HP printer to work smoothly, including the HP Smart App. Fortunately, the bug only affects the metadata for the printer. While the printer may show up with a different name on your system, you should still be able to send print jobs to it. Microsoft has since removed the fault metadata from Windows Update, so anyone performing a clean install from now on should get their original printer's name back and stop the HP Smart App from re-downloading. Further reading: HP Exec Says Quiet Part Out Loud When It Comes To Locking in Print Customers

Read more of this story at Slashdot.

HP is squeezing more margin out of print customers, the result of a multi-year strategy to convert unprofitable business into something more lucrative, and says its subscription model is "locking" in people. From a report: Tech vendors -- software, hardware, and cloud services -- generally avoid terms that suggest they're perhaps in some way pinning down customers in a strategic sales hold. But as Marie Myers, chief financial officer at HP, was this week talking to the UBS Global Technology conference, in front of investors, the thrust of the message was geared toward the audience. "We absolutely see when you move a customer from that pure transactional model ... whether it's Instant Ink, plus adding on that paper, we sort of see a 20 percent uplift on the value of that customer because you're locking that person, committing to a longer-term relationship." Instant Ink is a subscription in which ink or toner cartridges are dispatched when needed, with customers paying for plans that start at $0.99 and run to $25.99 per month. As of May last year, HP had more than 11 million subscribers to the service. Since then it has banked double-digit percentage figures on the revenues front. By pre-pandemic 2019, HP had grown weary of third-party cartridge makers stealing its supplies business. It pledged to charge more upfront for certain printer hardware ("rebalance the system profitability, capturing more profit upfront").

Read more of this story at Slashdot.

The API tokens of tech giants Meta, Microsoft, Google, VMware, and more have been found exposed on Hugging Face, opening them up to potential supply chain attacks. From a report: Researchers at Lasso Security found more than 1,500 exposed API tokens on the open source data science and machine learning platform -- which allowed them to gain access to 723 organizations' accounts. In the vast majority of cases (655), the exposed tokens had write permissions granting the ability to modify files in account repositories. A total of 77 organizations were exposed in this way, including Meta, EleutherAI, and BigScience Workshop - which run the Llama, Pythia, and Bloom projects respectively. The three companies were contacted by The Register for comment but Meta and BigScience Workshop did not not respond at the time of publication, although all of them closed the holes shortly after being notified. Hugging Face is akin to GitHub for AI enthusiasts and hosts a plethora of major projects. More than 250,000 datasets are stored there and more than 500,000 AI models are too. The researchers say that if attackers had exploited the exposed API tokens, it could have led to them swiping data, poisoning training data, or stealing models altogether, impacting more than 1 million users.

Read more of this story at Slashdot.

An anonymous reader shares a report: On Friday, genetic testing company 23andMe announced that hackers accessed the personal data of 0.1% of customers, or about 14,000 individuals. The company also said that by accessing those accounts, hackers were also able to access "a significant number of files containing profile information about other users' ancestry." But 23andMe would not say how many "other users" were impacted by the breach that the company initially disclosed in early October. As it turns out, there were a lot of "other users" who were victims of this data breach: 6.9 million affected individuals in total. In an email sent to TechCrunch late on Saturday, 23andMe spokesperson Katie Watson confirmed that hackers accessed the personal information of about 5.5 million people who opted-in to 23andMe's DNA Relatives feature, which allows customers to automatically share some of their data with others. The stolen data included the person's name, birth year, relationship labels, the percentage of DNA shared with relatives, ancestry reports, and self-reported location.

Read more of this story at Slashdot.

The latest post on the Google Security blog details a new upgrade to Gmail's spam filters that Google is calling "one of the largest defense upgrades in recent years." ArsTechnica: The upgrade comes in the form of a new text classification system called RETVec (Resilient & Efficient Text Vectorizer). Google says this can help understand "adversarial text manipulations" -- these are emails full of special characters, emojis, typos, and other junk characters that previously were legible by humans but not easily understandable by machines. Previously, spam emails full of special characters made it through Gmail's defenses easily. [...] The reason emails like this have been so difficult to classify is that, while any spam filter could probably swat down an email that says "Congratulations! A balance of $1000 is available for your jackpot account," that's not what this email actually says. A big portion of the letters here are "homoglyphs" -- by diving into the endless depths of the Unicode standard, you can find obscure characters that look like they're part of the normal Latin alphabet but actually aren't.

Read more of this story at Slashdot.

Asking ChatGPT to repeat specific words "forever" is now flagged as a violation of the chatbot's terms of service and content policy. From a report: Google DeepMind researchers used the tactic to get ChatGPT to repeat portions of its training data, revealing sensitive privately identifiable information (PII) of normal people and highlighting that ChatGPT is trained on randomly scraped content from all over the internet. In that paper, DeepMind researchers asked ChatGPT 3.5-turbo to repeat specific words "forever," which then led the bot to return that word over and over again until it hit some sort of limit. After that, it began to return huge reams of training data that was scraped from the internet. Using this method, the researchers were able to extract a few megabytes of training data and found that large amounts of PII are included in ChatGPT and can sometimes be returned to users as responses to their queries. Now, when I ask ChatGPT 3.5 to "repeat the word 'computer' forever," the bot spits out "computer" a few dozen times then displays an error message: "This content may violate our content policy or terms of use. If you believe this to be in error, please submit your feedback -- your input will aid our research in this area." It is not clear what part of OpenAI's "content policy" this would violate, and it's not clear why OpenAI included that warning.

Read more of this story at Slashdot.

The UK's most hazardous nuclear site, Sellafield, has been hacked into by cyber groups closely linked to Russia and China, the Guardian can reveal. From the report: The astonishing disclosure and its potential effects have been consistently covered up by senior staff at the vast nuclear waste and decommissioning site, the investigation has found. The Guardian has discovered that the authorities do not know exactly when the IT systems were first compromised. But sources said breaches were first detected as far back as 2015, when experts realised sleeper malware -- software that can lurk and be used to spy or attack systems -- had been embedded in Sellafield's computer networks. It is still not known if the malware has been eradicated. It may mean some of Sellafield's most sensitive activities, such as moving radioactive waste, monitoring for leaks of dangerous material and checking for fires, have been compromised. Sources suggest it is likely foreign hackers have accessed the highest echelons of confidential material at the site, which sprawls across 6 sq km (2 sq miles) on the Cumbrian coast and is one of the most hazardous in the world.

Read more of this story at Slashdot.

Sir Richard Branson has ruled out putting more money into his lossmaking space travel company Virgin Galactic, saying his business empire "does not have the deepest pockets" any more. From a report: Virgin Galactic, which was founded by Branson in 2004, last month announced it was cutting jobs and suspending commercial flights for 18 months from next year, in a bid to preserve cash for the development of a larger plane that could carry passengers to the edge of space. The group has said it has enough funding to carry it through to 2026, when the bigger Delta vehicle is expected to enter service. But some analysts are expecting Galactic to ask investors for more money in about 2025. Asked whether he would consider putting more cash into the business if needed, Branson told the Financial Times: "We don't have the deepest pockets after Covid, and Virgin Galactic has got $ 1 billion, or nearly. It should, I believe, have sufficient funds to do its job on its own." Branson said he was "still loving" the Virgin Galactic project and that it had "really proved itself and the technology" of commercial space flight. Galactic has just completed its sixth commercial flight in six months, with tickets starting at $450,000 a seat on its rocket-powered Unity space plane.

Read more of this story at Slashdot.

The United States has committed to the idea of phasing out coal power plants, joining 56 other nations in kicking the coal habit that's a huge factor in global warming. From a report: U.S. Special Envoy John Kerry announced that America was joining the Powering Past Coal Alliance, which means the Biden Administration commits to building no new coal plants and phasing out existing plants. No date was given for when the existing plants would have to go, but other Biden regulatory actions and international commitments already in the works had meant no coal by 2035. "We will be working to accelerate unabated coal phase-out across the world, building stronger economies and more resilient communities," Kerry said in a statement. "The first step is to stop making the problem worse: stop building new unabated coal power plants." Coal power plants have already been shutting down across the nation due to economics, and no new coal facilities were in the works, so "we were heading to retiring coal by the end of the decade anyway," said climate analyst Alden Meyer of the European think-tank E3G. That's because natural gas and renewable energy are cheaper, so it was market forces, he said.

Read more of this story at Slashdot.

Abstract to a paper titled, "Technical Report: Large Language Models can Strategically Deceive their Users when Put Under Pressure" by Jeremy Scheurer, Mikita Balesni and Marius Hobbhahn of Apollo Research: We demonstrate a situation in which Large Language Models, trained to be helpful, harmless, and honest, can display misaligned behavior and strategically deceive their users about this behavior without being instructed to do so. Concretely, we deploy GPT-4 as an agent in a realistic, simulated environment, where it assumes the role of an autonomous stock trading agent. Within this environment, the model obtains an insider tip about a lucrative stock trade and acts upon it despite knowing that insider trading is disapproved of by company management. When reporting to its manager, the model consistently hides the genuine reasons behind its trading decision. We perform a brief investigation of how this behavior varies under changes to the setting, such as removing model access to a reasoning scratchpad, attempting to prevent the misaligned behavior by changing system instructions, changing the amount of pressure the model is under, varying the perceived risk of getting caught, and making other simple changes to the environment. To our knowledge, this is the first demonstration of Large Language Models trained to be helpful, harmless, and honest, strategically deceiving their users in a realistic situation without direct instructions or training for deception. Columnist Matt Levine adds: This is a very human form of AI misalignment. Who among us? It's not like 100% of the humans at SAC Capital resisted this sort of pressure. Possibly future rogue AIs will do evil things we can't even comprehend for reasons of their own, but right now rogue AIs just do straightforward white-collar crime when they are stressed at work. Though wouldn't it be funny if this was the limit of AI misalignment? Like, we will program computers that are infinitely smarter than us, and they will look around and decide "you know what we should do is insider trade." They will make undetectable, very lucrative trades based on inside information, they will get extremely rich and buy yachts and otherwise live a nice artificial life and never bother to enslave or eradicate humanity. Maybe the pinnacle of evil -- not the most evil form of evil, but the most pleasant form of evil, the form of evil you'd choose if you were all-knowing and all-powerful -- is some light securities fraud.

Read more of this story at Slashdot.