Baidu's founder Robin Li declared his company's large language model has finally caught up with OpenAI's advanced GPT-4, claiming the lead in his country's race to develop AI that can rival the US. From a report: The billionaire took the stage in Beijing Tuesday to run Ernie 4.0 through a Q&A designed to showcase its ability to provide answers and solve complicated puzzles on the fly. Ernie has matched OpenAI's seminal product in terms of sophistication and general capabilities, Li told a packed house at a converted steel mill that now serves as an auditorium. The marquee Ernie chatbot now surpassed 45 million users -- a milestone that still lags ChatGPT's estimated 180 million, though the US bot launched months earlier. China's search leader, often referred to as a homegrown equivalent to Google, is pinning its hopes on AI to help it surpass rivals from Alibaba Group Holding Ltd. to Tencent that control the rest of the internet.

Baidu is leading a wave of aggressive investment across China after ChatGPT demonstrated the disruptive potential of generative AI -- which can craft video and content from simple commands. It's regarded as a leader in a race with local big tech firms and scores of startups to create a next-generation platform for the world's biggest internet market. They're trying to compete with American names from Microsoft to Google to create services like ChatGPT and Dall-E, but US sanctions on Chinese access to the most advanced chips to train and run AI models, coupled with Beijing's stringent censorship, could cloud their prospects. Washington is tightening curbs on shipments of AI chips to the country, stoking that uncertainty. "Ernie is not inferior in any respect to GPT-4," Li told the audience.

While the tech in virtual reality headsets has "undoubtedly gotten better," iFixit says "repair is getting left off of designers' priority lists." In a recent teardown video, the DIY repair site disassembled Meta's Quest 3 headset to find that it's not super repairable," giving it a repairability score of 4 out of 10 due the absence of manuals, OEM spare parts, and "any sign of repairability considerations whatsoever." The Register reports: As the iFixit team tore into the headset, the first major failure from a repairability perspective was the "extremely complicated procedure of replacing the lithium polymer battery pack." "Replacing the battery in the Quest 3 is as difficult as it was in the Quest 2, and far more difficult than the Quest Pro." That said, the batteries in the controllers are AAs rather than the lithium-ion cells of the Quest Pro, so it's a win there.

Faced with a multitude of screws and the lack of a service manual, iFixit stripped the headset back to its bare components, revealing the new time of flight sensor -- essential for hand and controller tracking as well as mapping out the space around the user -- and, beyond the fan, the mainboard. The Quest 3 is powered by a Snapdragon 8, the XR2 Gen2. According to iFixit: "Leaked benchmarks suggest that this newer SoC improves on the XR2+ found in the Quest Pro both in terms of performance and power efficiency."

However, it is the battery that disappoints. Although it is a standard unit so theoretically replaceable, iFixit noted: "It's taken me three Fixmats, a single tray of plastic, and very careful organizing of about 50 screws to get this far." Yikes. Not really a user-serviceable part at all. [...] Overall, the team gave the device a provisional 4 out of 10 in its teardown, principally due to the absence of manuals, OEM spare parts, and "any sign of repairability considerations whatsoever." But hey, at least you can swap out the AAs in the controllers when they die.

From the South China Morning Post: Scientists in China say their latest quantum computer has solved an ultra-complicated mathematical problem within a millionth of a second — more than 20 billion years quicker than the world's fastest supercomputer could achieve the same task. The JiuZhang 3 prototype also smashed the record set by its predecessor in the series, with a one million-fold increase in calculation speed, according to a paper published on Tuesday by the peer-reviewed journal Physical Review Letters...

The series uses photons — tiny particles that travel at the speed of light — as the physical medium for calculations, with each one carrying a qubit, the basic unit of quantum information... The fastest classical supercomputer Frontier — developed in the US and named the world's most powerful in mid-2022 — would take over 20 billion years to complete the same task, the researchers said.
The article claims they've increased the number of photons from 76 to 113 in the first two versions, improving to 255 in the latest iteration.

Thanks to long-time Slashdot reader hackingbear for sharing the news.

samleecole writes: Tickets for rapper Travis Scott's upcoming tour sold out fast. Check StubHub right now, however, and you can find thousands of tickets to "sold out" shows in many cities for between $10 and $20, far below the face value for his cheapest tickets at $61.50 before fees when they first went on sale. In ticket reseller lingo, Scott's tour is a "bloodbath," the result of overzealous brokers and noobs "overbuying" tickets based on a miscalculation of the likely value of his tickets on the secondary market. Many brokers now stand to lose a lot of money on Scott's shows. At least part of this buying frenzy was fueled by a bet placed by PFS Buyers Club, a credit card maxing site I wrote about earlier this week that has recently pivoted from buying rare coins to buying concert tickets. PFS told its members to buy as many tickets to Scott's shows as possible, according to emails viewed by 404 Media.

PFS itself stands to lose more than $1 million on Travis Scott alone when all is said and done, it told members. The entire situation, which has become a complicated mess, sheds light on a little-known segment of the ticket broker industry, where resellers partner with credit card "buyers clubs" to obtain tickets. The fiasco also highlights the risks associated with ticket reselling and shows how Ticketmaster profits from the secondary market, helping it sell out artists even before their ability to sell out venues is guaranteed, and passing that risk on to resellers.

First the Washington Post summarizes what scientists believed in the 1970s. Chlorofluorocarbons, or (CFCs, "could float up into the stratosphere and break down a protective layer of ozone, allowing more ultraviolet light to enter the atmosphere and harm humans, crops, and entire ecosystems. In fact, this had already happened: There was a hole in the ozone layer over the South Pole." Experts view the subsequent treaty to cut down on the use of CFCs — the 1987 Montreal Protocol — as a landmark environmental achievement. Scientists estimate that the pact has prevented millions of cases of skin cancer. Today, the ozone hole is recovering well. But a provocative scientific paper published Friday in the journal AGU Advances suggests that the link between the ozone layer and human health is more complicated than it seems. Under certain circumstances, the researchers write, small decreases in the ozone layer could now save lives...

The researchers initially were examining something else: what would happen to the chemistry of the atmosphere if humans injected sulfates into the stratosphere, a controversial strategy to cool the planet. But in the process, they found that the chemicals would alter the atmosphere's ozone content — with consequences for human health. Sulfate chemicals are known to deplete ozone high in the atmosphere, but, the paper shows, they could also decrease ground-level air pollution. Ozone, or O3, occurs in two forms in the atmosphere: what scientists call "good ozone" in the stratosphere, the layer of the atmosphere that sits 6 to 31 miles above the surface, and "bad ozone" in the troposphere, the atmospheric layer that reaches to the ground... an air pollutant in the troposphere that comes from power plants, cars, and industrial sites. It can be deadly, exacerbating respiratory diseases. According to one study, over 400,000 people died from long-term exposure to ozone in 2019 alone.

The new paper shows that "good ozone" and "bad ozone" can interact in unexpected ways. When good ozone is depleted, more UV light reaches the troposphere, which increases the rate of skin cancer. But UV light also catalyzes chemical reactions in the troposphere, including one in which hydroxide, or OH — which some scientists call the "Pac-Man of the atmosphere" — swallows up pollutants. The more UV light, the more OH eats up dangerous pollutants. This decrease in ground-level air pollution, according to the study, could actually outweigh the rise in skin cancer. A small decrease in stratospheric ozone, according to their study, could save between 33,000 and 86,000 lives every year.

Only a few papers have made this connection, including one in 2018 that similarly found that a small decrease in the ozone layer could save lives from air pollution... One way to read the study is as another warning of how dangerous ground-level air pollution is and how far the world still needs to go to clean it up. (Outdoor air pollution writ large is associated with an estimated 4.2 million premature deaths every year.)

"Here is another radical replacement for the traditional combustion engine," writes long-time Slashdot reader Inzkeeper. "Check out the Astron Aerospace H2 Starfire Omega 1... an ICE engine with a turbine configuration."

The company "is claiming that it is a viable alternative to EVs," reports TopSpeed: Astron have showcased a 3D rendering of their engine which helps to understand this extremely complicated new powerplant in all of its glory. They also showed a functioning prototype which gives us a glimpse into how the engine could potentially function... The company claims that it weighs an absolutely mind-boggling 35 pounds yet produces horsepower in the region of 160 and about 170 pound-feet of torque. These are insane figures. The Omega 1 boasts an alleged 60 percent efficiency, which is absurd if true given that piston engines rarely ever top 40 percent efficiency. On top of this, Omega 1 can run on any kind of combustible fuel, meaning that hydrogen could easily be used to reduce emissions so close to zero that it's negligible.
HotCars adds that "According to Astron Aerospace, the engine idles at 1,000 rpm and redlines at an incredible 25,000 rpm — much higher than all the other rotaries we've seen. This is thanks to the circular movement, rather than the epitrochoidal movement used for Reuleaux triangle rotaries." The awesome thing about this engine is that it is stackable, meaning two of them will make 320 hp and 340 lb-ft, three will produce 480 hp and 510 lb-ft, etc... Astron Aerospace also stated that due to the design, the engine is easily scalable for other applications — for instance, marine engines. According to them and one of their renders, a larger version can easily reach 4,500 hp...

[I]t is not only more efficient than the equivalent piston engine — 80% compared to a mere 34% — but the engine is also much smaller and lighter. This translates to better fuel economy and a lighter overall vehicle. The engine is air-cooled as well, which means there aren't any additional radiators or other cooling systems needed to keep the engine working. Air-cooled may sound a bit old-fashioned, but in this case, it simplifies the whole package. The maintenance on such an engine would also be minimal, with Astron Aerospace claiming 60,000 miles further usage over a typical piston engine before maintenance is required.

The disadvantage of this engine is that it hasn't yet been thoroughly tested in real-world conditions. Astron Aerospace has patented the engine and has a working prototype but has found no investors to begin mass testing and production. The engine needs to be worked hard to flesh out any potential weak points and new materials need to be used to cope with the internal stresses and wear.

An anonymous reader quotes a report from Gizmodo: Google's Privacy Sandbox, a controversial set of tools and settings meant to replace third-party cookies, is now on almost every single Chrome browser, according to a company blog post published Thursday. Google says Privacy Sandbox is now available to around 97% of Chrome users, and that number will reach 100% in the next few months. The news comes on the heels of the browser's 15th anniversary, which Google is celebrating by redesigning Chrome to make it look and feel more closely aligned with the design paradigm of Android and the rest of the Google suite. The final step in this process comes in 2024, when Google will disable third-party cookies in Chrome for good, marking the end of their decades-long reign of privacy-violating terror.

Back in 2019, Google said the cookie era was coming to a close. In place of third-party cookies, Privacy Sandbox will implement a long list of new tools for the ad industry. Google, after all, makes all of its money by spying on you and turning the insights into ads, so it's not about to put itself out of business. In fairness, this new system is really more private, though it's private on Google's terms. The biggest change is "Ad Topics," a.k.a. the Topics API if you're a huge nerd who's been following this stuff for years. With Topics, Chrome will keep track of all the websites you're looking at and sort you into a variety of categories. This tracking happens in your browser and the data stays on your device. Neither Google nor anyone else gets to see your browsing history or learn anything about you as an individual throughout this process. Websites and advertising companies will know there's a person interested in a certain Topic, but they won't be able to tell who you are specifically.

There's also an extremely complicated technique websites can use to tag you with subjects they want you to see ads about, called "Site Suggested Ads." Google is also rolling out a tool called "Ad Measurement," which helps companies keep track of how well their ads are working through metrics such as the time of day you saw an ad and whether you clicked on it. Google gives users some control over how these tools are implemented. With the rollout of Privacy Sandbox comes new settings listed as "Ad privacy controls," which you can adjust in Chrome's preferences. Further reading: Chrome is About To Look a Bit Different

"France and Germany lead the camps in disagreeing on the future of nuclear in Europe," write two climate policy journalists. On the Energy Post blog they explore why — citing energy experts and politicians.

Germany "ultimately completed its nuclear exit in April 2023," while France "has the highest share of nuclear in the energy mix of any country in the world." [A] major concern is that more nuclear means less renewables, at a time when wind and solar need all the scale they can get... In a joint attempt to provide greater technical clarity on the nuclear power debate, French think tank IDDRI and German Agora Energiewende set out in 2018 to understand how nuclear energy will influence the transformation of energy systems in both countries. They found that if a high share of coal or nuclear based conventional power capacity stays online in both countries, this will likely to delay the time when market prices allow renewable power operators to cover their production costs and run the operations at a profit. They also found that exporting surplus electricity with conventional plants bites into renewable power investments abroad. At the same time, the growing share of renewables would eventually render most conventional plants unprofitable. "In order to avoid stranded assets, it is essential to gradually reduce conventional capacities," the bi-national report concluded...

Xavier Moreno, president of French think tank Ecological Realities and Energy Mix Study Circle (Cereme) and former vice president of French utility company Suez, said the all-renewables approach was complicated by a lack of viable electricity storage technologies. "Technically speaking, it would be necessary to store up to 20 percent to be able to smoothen renewable power supply." Those who believe that this will be possible through a combination of different storage options are chasing "a dream," Moreno argued.
The issue comes up when trading power in Europe's integrated energy market: should gate closure times be based on a decentralised, flexible renewables-based system, or a centralised grid based on nuclear baseloads? Rainer Hinrichs-Rahlwes, European policy expert for the German Renewable Energy Federation lobby group, says "Nuclear power plants and their inflexible output can cause grid congestion, the opposite of what is needed to accommodate large shares of wind and solar in a modern and flexible grid system."

The article notes that France plans to eliminate coal use by 2038, and already has one of the lowest emissions per head of any rich country. But "In mid-2023, 800 French scientists warned against the risks of the country's new nuclear programme, pointing to unresolved questions of radioactive waste management, which remain largely unresolved in most of the EU, including in France. The scientists also warned against risks of accidental contamination or meltdown."

Thanks to Slashdot reader AleRunner for submitting the article.

U.S. Commerce Secretary Gina Raimondo opened talks with Chinese government officials on Monday, saying it is "profoundly important" for the world's two largest economies to have a stable economic relationship. From a report: Raimondo is looking to boost business ties as U.S. firms have reported increasing challenges with operating in China, while China has sharply criticized U.S. efforts to block its access to advanced semiconductors. Raimondo said the entire world expects the United States and China will have a stable economic relationship; the two countries share more than $700 billion in annual trade. "It's a complicated relationship. "It's a challenging relationship. We will of course disagree on certain issues," Raimondo said. "I think we can make progress if we are direct, open and practical."

The federal government is escalating efforts to make cryptocurrency investors comply with tax law, nearly 15 years after people started trading bitcoin. From a report: The Treasury Department proposed new rules Friday with twin goals: making it harder for crypto investors to dodge income taxes when they sell digital assets, and simplifying complicated tax messes for people who are trying to follow the law. When they are fully implemented, the rules will require crypto exchanges such as Coinbase to deal with the Internal Revenue Service in a manner similar to brokers who handle investors' stock and mutual-fund portfolios.

The crypto exchanges will send annual reports on Form 1099s to the IRS and to taxpayers that show the gross proceeds from transactions. That starts in 2026 for tax year 2025. Later, they will start reporting how much customers paid for the assets, known as their cost basis. Capital gains are the difference between sale price and cost basis, and investors face federal taxes of up to 23.8%.

Monday sees the release of "The Billion Dollar Heist," a documentary about the theft of $81 million from the Bangladesh Bank, considered the biggest cyber-heist of all time. The film's executive producer wrote the book Dark Market: How Hackers Became the New Mafia (and is also a rector at the Institute for Human Sciences).

But he's also written an article for the Financial Times outlining the complicated background of Russian-speaking hacker gangs responsible for malware and ransomware, starting with "one of the most remarkable if little-known events in post-cold war history: the first and, to my knowledge, the last publicly organised conference of avowed criminals" in May, 2002.

The First Worldwide Carders Conference was the brainchild of the administrators of a landmark website, carderplanet.com. Known as "the family", this was a mixed group of young men, both Ukrainians and Russians, who had spent the previous 10 years growing up in a lively atmosphere of gangster capitalism. During the 1990s, conventional law and order in the former Soviet Union had broken down. The collapse of the communist system had left a vacuum in which new forms of economic activity were emerging...

Founded a year before the conference, CarderPlanet revolutionised web-based criminal activity, especially the lucrative trade in stolen or cloned credit card data, by solving the conundrum that until then had faced every bad guy on the web: how can I do business with this person, as I know he's a criminal, so he must be untrustworthy by definition? To obviate the problem, the CarderPlanet administrators created an escrow system for criminals. They would act as guarantor of any criminal sale of credit and debit card data — a disinterested party mediating between the vendor and the purchaser... The escrow system led to an explosion of credit card crime around the world in which many criminal fortunes were made....

Roman Stepanenko Vega, a Russian-speaking Ukrainian national who was one of the founders and administrators of CarderPlanet, explained to me how "two days before the conference's opening, we received a visit from an FSB [Federal Security Service] officer in Moscow. He explained that Moscow had no objections to us cloning credit cards or defrauding banks in Europe and the United States but anywhere within the CIS was off limits." In addition, the FSB officer let CarderPlanet know that if the Russian state ever required assistance from criminal gangs, it would be expected to co-operate...

Members of criminal gangs were later recruited into notorious state-backed hacking teams such as Advanced Persistent Threat 28.
A 2021 ransomware attack on Colonial Pipeline brought warnings of a U.S. counterattack, the article notes, after which "Russian police started arresting and imprisoning cyber criminal groups." Ransomware attacks now seem particularly focused on Europe, and "According to cyber-security experts, the Russian government is giving these criminal groups information on potential targets." But once more the hackers have been careful not to cross what the Americans consider red lines, as advised, presumably, by Russia's security services. Russia is probably confident that disrupting European businesses will be unlikely to provoke a cyber attack. But the U.S. — whether its government, municipalities or police — remains strictly off-limits.
Thanks to long-time Slashdot reader Geoffrey.landis for sharing the article.

A recently released research reveals a fundamental challenge of developing artificial intelligence: ChatGPT has become worse at performing certain basic math operations. From a report: The researchers at Stanford University and the University of California, Berkeley said the deterioration is an example of a phenomenon known to AI developers as drift, where attempts to improve one part of the enormously complex AI models make other parts of the models perform worse.

[...] Thus far, they have tested two versions of ChatGPT: version 3.5, available free online to anyone, and version 4.0, available via a premium subscription. The results aren't entirely promising. They gave the chatbot a basic task: identify whether a particular number is a prime number. This is the sort of math problem that is complicated for people but simple for computers.

Is 17,077 prime? Is 17,947 prime? Unless you are a savant you can't work this out in your head, but it is easy for computers to evaluate. A computer can just brute force the problem -- try dividing by two, three, five, etc., and see if anything works. To track performance, the researchers fed ChatGPT 1,000 different numbers. In March, the premium GPT-4, correctly identified whether 84% of the numbers were prime or not. (Pretty mediocre performance for a computer, frankly.) By June its success rate had dropped to 51%. Across eight different tasks, GPT-4 became worse at six of them. GPT-3.5 improved on six measures, but remained worse than its advanced sibling at most of the tasks.

During its 54th flight on Mars, NASA's helicopter Ingenuity captured an image of the space agency's Perseverance rover. Space.com reports: Perseverance is nearly out of frame at the top of the photo, which Ingenuity took when it was about 16 feet (5 meters) above the red dirt. Unlike previous sorties, the Aug. 3 flight wasn't a scouting run to aid Perseverance's science activities. It lasted just 24 seconds, reached a maximum altitude of 16 feet and covered no ground laterally, according to Ingenuity's flight log. The mission team designed this short and simple hop in an attempt to help understand what happened during Ingenuity's previous flight, which was cut short unexpectedly.

The mission team designed this short and simple hop in an attempt to help understand what happened during Ingenuity's previous flight, which was cut short unexpectedly. That July 22 sortie was supposed to last 136 seconds and feature several complicated maneuvers. However, Ingenuity stayed aloft for just 74 seconds, touching down after something triggered its "flight-contingency program." [...] Imagery from Ingenuity's navigation camera likely got out of sync with its inertial measurement unit, which helps the little chopper determine its position, speed and orientation. This also happened near the end of Ingenuity's sixth flight, back in May 2021. The mission team soon uploaded a software patch to deal with the issue, but that patch apparently couldn't handle what happened on Flight 53, NASA officials said in the statement. "Since the very first flight, we have included a program called 'LAND_NOW' that was designed to put the helicopter on the surface as soon as possible if any one of a few dozen off-nominal scenarios was encountered," Teddy Tzanetos, Ingenuity team lead emeritus at NASA's Jet Propulsion Laboratory in Southern California, said in a statement.

"During Flight 53, we encountered one of these, and the helicopter worked as planned and executed an immediate landing," Tzanetos added.

Google is launching its AI-backed note-taking tool to "a small group of users in the US," the company said in a blog post. Formerly referred to as Project Tailwind at Google I/O earlier this year, the new app is now known as NotebookLM (the LM stands for Language Model). The Verge reports: The core of NotebookLM seems to actually start in Google Docs. ("We'll be adding additional formats soon," the blog post says.) Once you get access to the app, you'll be able to select a bunch of docs and then use NotebookLM to ask questions about them and even create new stuff with them. Google offers a few ideas for things you might do in NotebookLM, such as automatically summarizing a long document or turning a video outline into a script. Google's examples, even back at I/O, seemed primarily geared toward students: you might ask for a summary of your class notes for the week or for NotebookLM to tell you everything you've learned about the Peloponnesian War this semester.

These are the kinds of features you'll hear about in practically any AI product, but Google is hoping that by limiting the underlying model only to the information you've added yourself, it can both improve the model's responses and help mitigate its tendency to confidently lie about everything. (Google's not unique in this idea, either: Dropbox, Mem, Notion, and many others are pursuing similar hyper-specific AI tools of their own.) NotebookLM also has citations built in, which should make it easier to quickly fact-check the automatically generated responses. But Google does warn that NotebookLM might still hallucinate and that the model won't always get it right. It also, of course, depends on the information you provide -- if you wrote down the wrong dates for the Peloponnesian War in class, it can't help you.

Google says that the NotebookLM model only has access to the documents you choose to upload and that your data is neither available to others nor is it used to train new AI models. This is one of the trickiest parts of a product like this: Google is asking users to give their private information to an AI model in exchange for some convenient and useful features, and that tradeoff gets more complicated the more sensitive the information becomes.

Popular Science describes how NASA's Volatiles Investigating Polar Exploration Rover (VIPER) will use a pair of ramps to become the first rover to explore the Moon's south pole when it arrives in late 2024. From the report: "We all know how to work with ramps, and we just need to optimize it for the environment we're going to be in," says NASA's VIPER program manager Daniel Andrews. A VIPER test vehicle recently descended down a pair of metal ramps at NASA's Ames Research Center in California, as seen in the agency's recently published photos, with one beam for each set of the rover's wheels. Because the terrain where VIPER will land -- the edge of the massive Nobile Crater -- is expected to be rough, the engineering team has been testing VIPER's ability to descend the ramps at extreme angles. They have altered the steepness, as measured from the lander VIPER will descend from, and differences in elevation between the ramp for each wheel. "We have two ramps, not just for the left and right wheels, but a ramp set that goes out the back too," Andrews says. "So we actually get our pick of the litter, which one looks most safe and best to navigate as we're at that moment where we have to roll off the lander."

VIPER is a scientific successor to NASA's Lunar Crater Observation and Sensing Satellite, or LCROSS mission, which in 2009 confirmed the presence of water ice on the lunar south pole. "It completely rewrote the books on the moon with respect to water," says Andrews, who also worked on the LCROSS mission. "That really started the moon rush, commercially, and by state actors like NASA and other space agencies." The ice, if abundant, could be mined to create rocket propellant. It could also provide water for other purposes at long-term lunar habitats, which NASA plans to construct in the late 2020s as part of the Artemis moon program.

But LCROSS only confirmed that ice was definitely present in a single crater at the moon's south pole. VIPER, a mobile rover, will probe the distribution of water ice in greater detail. Drilling beneath the lunar surface is one task. Another is to move into steep, permanently shadowed regions -- entering craters that, due to their sharp geometry, and the low angle of the sun at the lunar poles, have not seen sunlight in billions of years. The tests demonstrate the rover can navigate a 15-degree slope with ease -- enough to explore these hidden dark spots, avoiding the need to make a machine designed for trickier descents. "We think there's plenty of scientifically relevant opportunities, without having to make a superheroic rover that can do crazy things," Andrews says.

Developed by NASA Ames and Pittsburgh-based company Astrobotic, VIPER is a square golf-cart-sized vehicle about 5 feet long and wide, and about 8 feet high. Unlike all of NASA's Mars rovers, VIPER has four wheels, not six. "A problem with six wheels is it creates kind of the equivalent of a track, and so you're forced to drive in a certain way," Andrews says. VIPER's four wheels are entirely independent from each other. Not only can they roll in any direction, they can be turned out, using the rover's shoulder-like joints to crawl out of the soft regolith of the kind scientists believe exists in permanently shadowed moon craters. The wheels themselves are very similar to those on the Mars rovers, but with more paddle-like treads, known as grousers, to carry the robot through fluffy regolith. [...] Together with Astrobotic, Andrews and his team have altered the ramps, and they now include specialized etchings down their lengths. The rover can detect this pattern along the rampway, using cameras in its wheel wells. "By just looking down there," the robot knows where it is, he says. "That's a new touch." Andrews is sure VIPER will be ready for deployment in 2024, however many tweaks are necessary. After all, this method is less complicated than a sky crane, he notes: "Ramps are pretty tried and true."

New data (PDF) from the Bureau of Labor Statistics found that one-third of Americans worked from home in 2022, up from a quarter, or 25%, in 2019. The survey also found that Americans working full time from home worked 2.5 hours less a day than Americans at the office. Barron's reports: Overall, the total civilian population worked for an average of 3.23 hours a day in 2022 down from 3.26 hours a day in 2019. The U.S. is 1% lazier. That number, given by the BLS, is the total population. Don't forget, babies don't work. [...] As far as what Americans were doing with the time not spent working, TV watching stayed flat, socializing dropped, and gaming increased. "Economics is complicated, but labor productivity is essentially the basis for economic gains," writes Barron's Al Root. "The economy is measured in dollars, but the dollar is just a unit of account. More output per worker is how living standards improve."

"In a strange way, coming back to work is like an economic stimulus package. If people go back to the office, at a 2019 rate, and work 8.2 hours a day instead of the at-home 5.7 hours a day, the economy has just added roughly 800 million weeks of work, an 8% bump."

"The findings will give management teams some momentum to bring workers back to the office," adds Root.

Carl Zimmer writes in The New York Times: In its first week, a fertilized human egg develops into a hollow ball of 200 cells and then implants itself on the wall of the uterus. Over the next three weeks, it divides into the distinct tissues of a human body. And those crucial few weeks remain, for the most part, a black box. "We know the basics, but the very fine details we just don't know," said Jacob Hanna, a developmental biologist at the Weizmann Institute of Science in Israel. Dr. Hanna and a number of other biologists are trying to uncover those details by creating models of human embryos in the lab. They are coaxing stem cells to organize themselves into clumps that take on some of the crucial hallmarks of real embryos.

This month, Dr. Hanna's team in Israel, as well as groups in Britain, the United States and China, released reports on these experiments. The studies, while not yet published in scientific journals, have attracted keen interest from other scientists, who have been hoping for years that such advances could finally shed light on some of the mysteries of early human development. Ethicists have long cautioned that the advent of embryo models would further complicate the already complicated regulation of this research. But the scientists behind the new work were quick to stress that they had not created real embryos and that their clusters of stem cells could never give rise to a human being. "We do it to save lives, not create it," said Magdalena Zernicka-Goetz, a developmental biologist at the University of Cambridge and the California Institute of Technology, who led another effort. [...]

If scientists can create close, reliable models of embryos, they will be able to run large-scale experiments to test potential causes of pregnancy failures, such as viral infections and genetic mutations. The models could lead to other medical advances too, noted Insoo Hyun, a member of the Harvard Medical School Center for Bioethics who was not involved in the new studies. "Once you get the embryo models in place and you can rely on them, that can be an interesting way to screen drugs that women take when they're pregnant," he said. "That would be an enormous benefit." Dr. Hanna [...] also saw a possibility of using embryo models as a new form of stem-cell treatment for diseases such as cancer.

Trial lawyer Kyle Roche has led an interesting life, according to the New York Times. He once earned $100 million selling bitcoin. He helped win a case against Craig Wright (who claims to be Bitcoin creator Satoshi Nakamoto) through his law firm Roche Freedman. And Roche also founded a startup that lets people bet on the outcome of (civil) lawsuits, "to make access to justice more affordable."

But something very bad for his career happened in January of 2022 when two businessmen flew Roche from Miami to the U.K. to discuss an investment. When he woke up the next morning, Roche said, he felt groggy... The brain fog was odd because he didn't think he'd had all that much to drink. As he flew back to Miami a few days later, Roche couldn't shake the feeling that something was amiss.

Months passed. Then, one day last summer, Roche's world detonated. A website called Crypto Leaks posted two dozen videos of him that had been secretly recorded during his meetings with Villavicencio and Ager-Hanssen. The videos portrayed Roche and his law firm, Roche Freedman, as being in the pocket of one of their crypto clients [Ava Labs]... In other clips, Roche made it sound like his sole concern, even when representing other clients, was to promote Ava Labs' interests...

One after another, companies that Roche Freedman had sued filed motions to disqualify the firm from their cases. In October, the first of those motions succeeded: A federal judge in New York tossed Roche Freedman from a case it had filed against Tether, the operator of the world's most used "stablecoin." Within days, Roche was forced to resign from the law firm he had founded. With his career in tatters, he said, he enrolled in ethics classes and began to see a therapist.
Roche calls the recorded remarks baseless bluster to impress a prospective investor (and alleges in court there are signs of deep fake alterations). While Roche "was felled by his own loose lips and his overly cozy relationship with a client," the Times reports "he also was the victim of an elaborate international setup." On April 3, 2020, Roche Freedman filed lawsuits seeking class-action status against seven issuers of digital coins, alleging they had pumped what amounted to unregistered securities with false statements and then dumped them, leaving retail investors holding the bag... Those suits were just an opening salvo: Sixteen months later, Roche filed his biggest securities fraud case yet. It alleged that a British entrepreneur, Dominic Williams, and entities he controlled had swindled investors out of billions of dollars by aggressively promoting, and then dumping, a digital coin tied to a grandiose plan to revolutionize computing. Williams had boldly proclaimed that his Internet Computer blockchain — a decentralized network of computers powered by a digital token called ICP — would supplant the big cloud services offered by Amazon and Microsoft and become humanity's primary computing platform. But after an initial surge that briefly made it one of the most valuable cryptocurrencies, ICP had plummeted 92% — a collapse that Roche's lawsuit attributed to "massive" selling by Williams and other insiders. (Williams denied the allegations.)
The Times reports that Roche's prospective investor Ager-Hanssen, "in addition to running his venture capital firm, has long had a sideline digging up dirt on behalf of wealthy clients entangled in business disputes in Britain and Scandinavia. On multiple occasions, he has secretly recorded his targets. For example, in a 2014 interview, he recounted how he had snared the adversary of a Swedish financier with a hidden microphone and boasted that he employed former intelligence officers from the CIA, MI6 and Mossad..." Roche believes them because he thinks he knows who hired Ager-Hanssen: Williams, the British entrepreneur who was the target of Roche Freedman's biggest pump-and-dump lawsuit... On May 12, 2022, Williams wrote on Twitter that he was "coming for" his critics. That was the same day the cryptoleaks.info domain name was registered. That was the same day the cryptoleaks.info domain name was registered. Then, on June 9, 2022, the Crypto Leaks website went live. Billing itself as the defender of "the honest crypto community," it posted two reports that aligned with Williams' interests...

The first espoused a complicated theory about the ICP token crash that Williams had previously floated on Twitter. The second attacked the Times for an article it had published about the crash. Williams tweeted a link to that Crypto Leaks report, calling it "Gobsmacking." The Dfinity Foundation, a Swiss nonprofit that Williams created to oversee his blockchain, has since sued the Times for defamation in New York. The Times is seeking to dismiss the suit. The videos of Roche were the crux of Crypto Leaks' third exposé. After they were published, Williams and Dfinity filed a motion to disqualify Roche Freedman as plaintiffs' counsel in the pump-and-dump lawsuit, saying Roche's comments demonstrated "a disregard for the integrity of the judicial system...."

Last month, the judge overseeing the pump-and-dump case granted Williams' motion and disqualified Freedman Normand Friedland as plaintiffs' counsel.

Quanta magazine thinks there's a better alternative to the artificial neural networks (or ANNs) powering AI systems. (Alternate URL) For one, ANNs are "super power-hungry," said Cornelia Fermüller, a computer scientist at the University of Maryland. "And the other issue is [their] lack of transparency." Such systems are so complicated that no one truly understands what they're doing, or why they work so well. This, in turn, makes it almost impossible to get them to reason by analogy, which is what humans do — using symbols for objects, ideas and the relationships between them....

Bruno Olshausen, a neuroscientist at the University of California, Berkeley, and others argue that information in the brain is represented by the activity of numerous neurons... This is the starting point for a radically different approach to computation known as hyperdimensional computing. The key is that each piece of information, such as the notion of a car, or its make, model or color, or all of it together, is represented as a single entity: a hyperdimensional vector. A vector is simply an ordered array of numbers. A 3D vector, for example, comprises three numbers: the x, y and z coordinates of a point in 3D space. A hyperdimensional vector, or hypervector, could be an array of 10,000 numbers, say, representing a point in 10,000-dimensional space. These mathematical objects and the algebra to manipulate them are flexible and powerful enough to take modern computing beyond some of its current limitations and foster a new approach to artificial intelligence...

Hyperdimensional computing tolerates errors better, because even if a hypervector suffers significant numbers of random bit flips, it is still close to the original vector. This implies that any reasoning using these vectors is not meaningfully impacted in the face of errors. The team of Xun Jiao, a computer scientist at Villanova University, has shown that these systems are at least 10 times more tolerant of hardware faults than traditional ANNs, which themselves are orders of magnitude more resilient than traditional computing architectures...

All of these benefits over traditional computing suggest that hyperdimensional computing is well suited for a new generation of extremely sturdy, low-power hardware. It's also compatible with "in-memory computing systems," which perform the computing on the same hardware that stores data (unlike existing von Neumann computers that inefficiently shuttle data between memory and the central processing unit). Some of these new devices can be analog, operating at very low voltages, making them energy-efficient but also prone to random noise.
Thanks to Slashdot reader ZipNada for sharing the article.

ZDNet asks the million-dollar question. "Despite the potential for vast productivity gains from generative AI tools such as ChatGPT or GitHub Copilot, will technology professionals' jobs actually grow more complicated? " People can now pump out code on demand in an abundance of languages, from Java to Python, along with helpful recommendations. Already, 95% of developers in a recent survey from Sourcegraph report they use Copilot, ChatGPT, and other gen AI tools this way.

But auto-generating new code only addresses part of the problem in enterprises that already maintain unwieldy codebases, and require high levels of cohesion, accountability, and security.

For starters, security and quality assurance tasks associated with software jobs aren't going to go away anytime soon. "For programmers and software engineers, ChatGPT and other large language models help create code in almost any language," says Andy Thurai, analyst with Constellation Research, before talking about security concerns. "However, most of the code that is generated is security-vulnerable and might not pass enterprise-grade code. So, while AI can help accelerate coding, care should be taken to analyze the code, find vulnerabilities, and fix it, which would take away some of the productivity increase that AI vendors tout about."

Then there's code sprawl. An analogy to the rollout of generative AI in coding is the introduction of cloud computing, which seemed to simplify application acquisition when first rolled out, and now means a tangle of services to be managed. The relative ease of generating code via AI will contribute to an ever-expanding codebase — what the Sourcegraph survey authors refer to as "Big Code". A majority of the 500 developers in the survey are concerned about managing all this new code, along with code sprawl, and its contribution to technical debt. Even before generative AI, close to eight in 10 say their codebase grew five times over the last three years, and a similar number struggle with understanding existing code generated by others.

So, the productivity prospects for generative AI in programming are a mixed bag.