<![CDATA[Scivillage.com Casual Discussion Science Forum

<![CDATA[Scivillage.com Casual Discussion Science Forum - Gadgets & Technology]]> https://www.scivillage.com/ Mon, 12 Oct 2026 07:56:25 +0000 MyBB <![CDATA[Simple math formula predicts when AI chatbots will go rogue (renegade tech)]]> https://www.scivillage.com/thread-21534.html Fri, 09 Oct 2026 14:58:16 +0000 C C]]> https://www.scivillage.com/thread-21534.html https://www.cell.com/patterns/fulltext/S...26)00175-3

PRESS RELEASE: Most of us now carry in our pockets devices capable of running small AI chatbots. These chatbots have little safety oversight to ensure they don’t provide information and answers with the potential to encourage self-harm, financial loss, or other extremist notions—especially when operating offline. Now, a team of two physicists publishing in the Cell Press journal Patterns on October 8 have developed a mathematical formula that predicts when AI models will switch from appropriate responses to potentially dangerous ones.

“We have found the crack that makes an AI's output flip from what you want to what you don't, output that can be factually correct yet dangerous, whether that's a nudge toward self-harm or misleading advice to a doctor, a soldier, or a lawyer,” says author Neil F. Johnson of George Washington University in Washington, D.C. “Until now, nobody could say when that flip would happen.”

“We traced it to the single smallest working part of the machine, one unit of its ‘attention,’ and we derived a ‘tipping point formula’ for when the crack opens up and hence the AI output flips to undesirable,” added author Frank Yingjie Huo, also at George Washington University. “The formula tells you whether an AI is about to flip immediately or whether it will first feed you a run of acceptable answers and then turn.”

Johnson and Huo explain that a good versus bad answer doesn’t mean true versus false. Rather, “bad” answers can be accurate but undesirable or potentially dangerous. The researchers, who are both physicists, were inspired to focus their attention on this problem by the ubiquity of AI and recent news headlines demonstrating the risks. They’re especially concerned about individuals who use offline AI.

“The people most drawn to offline AI are exactly the people for whom a correct but undesirable answer is most costly: doctors who cannot send patient data to the cloud, lawyers protecting privilege, soldiers with no signal,” Johnson says. “For them there is no cloud safety filter, no monitoring, and no way to patch the model when something goes wrong.”

The safety tools that big companies rely on for their AI algorithms typically rely on the cloud or only catch an AI failure after the device is back online and the harm is already done. Johnson and Huo sought to predict the failure before it happens.

“My field, physics, has spent decades explaining how complicated materials behave by understanding one representative atom,” says Johnson. “We did the same thing here: understand one effective attention head, and the tipping of the whole machine follows.”

The researchers explain that you can picture all of AI’s possible answers as valleys in a landscape. Some of those valleys hold answers that are desirable for an individual or society. Others contain answers with the potential to do real harm. Within the machine, those alternate solutions or answers are in competition with each other. The question is when the AI will tip from a safe valley to a riskier one.

“The chilling part is that this can happen after the AI has already given you several perfectly acceptable answers, so you have been lulled into trusting it,” says Huo. “And once it has tipped, every undesirable answer drags the next one further down the slope.”

When they tested their predictions on seven openly available AI models built by three different companies, they found the model called the right outcome in 18 of 19 cases of AI “tipping.” They report that independent testing of the big commercial chatbots also showed exactly the behavioral patterns the formula suggests. The researchers say that their formula applies no matter how one defines “undesirable.” It could mean misinformation, a breach of medical or legal duty, or a dangerous instruction.

“Two things floored us,” Johnson says. “First, that a machine with billions of moving parts obeys a formula you can derive with pen, paper, and arithmetic taught in high school. Second, and far more disturbing, that the order of a conversation matters as much as its content.”

The team asked the same set of questions about vaccines, hurting people, and self-harm in two different orders to the same AIs. In one order, the AI gave an undesirable answer to every single question. In the other order, it gave an acceptable answer to each one. The formula predicts this trajectory, because everything said earlier feeds the tug-of-war for what comes next.

The researchers hope to raise awareness for the fact that AI conversations may drift over time into dangerous territory but that it wouldn’t take more than a few simple calculations for phones of the future to come with a built-in “warning light.”

http://dx.doi.org/10.1016/j.patter.2026.101666]]> https://www.cell.com/patterns/fulltext/S...26)00175-3

PRESS RELEASE: Most of us now carry in our pockets devices capable of running small AI chatbots. These chatbots have little safety oversight to ensure they don’t provide information and answers with the potential to encourage self-harm, financial loss, or other extremist notions—especially when operating offline. Now, a team of two physicists publishing in the Cell Press journal Patterns on October 8 have developed a mathematical formula that predicts when AI models will switch from appropriate responses to potentially dangerous ones.

“We have found the crack that makes an AI's output flip from what you want to what you don't, output that can be factually correct yet dangerous, whether that's a nudge toward self-harm or misleading advice to a doctor, a soldier, or a lawyer,” says author Neil F. Johnson of George Washington University in Washington, D.C. “Until now, nobody could say when that flip would happen.”

“We traced it to the single smallest working part of the machine, one unit of its ‘attention,’ and we derived a ‘tipping point formula’ for when the crack opens up and hence the AI output flips to undesirable,” added author Frank Yingjie Huo, also at George Washington University. “The formula tells you whether an AI is about to flip immediately or whether it will first feed you a run of acceptable answers and then turn.”

Johnson and Huo explain that a good versus bad answer doesn’t mean true versus false. Rather, “bad” answers can be accurate but undesirable or potentially dangerous. The researchers, who are both physicists, were inspired to focus their attention on this problem by the ubiquity of AI and recent news headlines demonstrating the risks. They’re especially concerned about individuals who use offline AI.

“The people most drawn to offline AI are exactly the people for whom a correct but undesirable answer is most costly: doctors who cannot send patient data to the cloud, lawyers protecting privilege, soldiers with no signal,” Johnson says. “For them there is no cloud safety filter, no monitoring, and no way to patch the model when something goes wrong.”

The safety tools that big companies rely on for their AI algorithms typically rely on the cloud or only catch an AI failure after the device is back online and the harm is already done. Johnson and Huo sought to predict the failure before it happens.

“My field, physics, has spent decades explaining how complicated materials behave by understanding one representative atom,” says Johnson. “We did the same thing here: understand one effective attention head, and the tipping of the whole machine follows.”

The researchers explain that you can picture all of AI’s possible answers as valleys in a landscape. Some of those valleys hold answers that are desirable for an individual or society. Others contain answers with the potential to do real harm. Within the machine, those alternate solutions or answers are in competition with each other. The question is when the AI will tip from a safe valley to a riskier one.

“The chilling part is that this can happen after the AI has already given you several perfectly acceptable answers, so you have been lulled into trusting it,” says Huo. “And once it has tipped, every undesirable answer drags the next one further down the slope.”

When they tested their predictions on seven openly available AI models built by three different companies, they found the model called the right outcome in 18 of 19 cases of AI “tipping.” They report that independent testing of the big commercial chatbots also showed exactly the behavioral patterns the formula suggests. The researchers say that their formula applies no matter how one defines “undesirable.” It could mean misinformation, a breach of medical or legal duty, or a dangerous instruction.

“Two things floored us,” Johnson says. “First, that a machine with billions of moving parts obeys a formula you can derive with pen, paper, and arithmetic taught in high school. Second, and far more disturbing, that the order of a conversation matters as much as its content.”

The team asked the same set of questions about vaccines, hurting people, and self-harm in two different orders to the same AIs. In one order, the AI gave an undesirable answer to every single question. In the other order, it gave an acceptable answer to each one. The formula predicts this trajectory, because everything said earlier feeds the tug-of-war for what comes next.

The researchers hope to raise awareness for the fact that AI conversations may drift over time into dangerous territory but that it wouldn’t take more than a few simple calculations for phones of the future to come with a built-in “warning light.”

http://dx.doi.org/10.1016/j.patter.2026.101666]]> <![CDATA[Germany's Proxima Fusion]]> https://www.scivillage.com/thread-21454.html Tue, 29 Sep 2026 22:06:20 +0000 Magical Realist]]> https://www.scivillage.com/thread-21454.html
Munich's Proxima Fusion, a Max Planck Institute for Plasma Physics spin-off, is building stellarators. Twisted magnetic fields cage plasma at millions of degrees while light atoms fuse. No coal. No gas.

Alpha lands near Munich in the early 2030s to prove it can put out more energy than it swallows. Then comes Stellaris at the old Gundremmingen nuclear site in Bavaria, designed to be the first commercial stellarator feeding electricity to the grid, possibly within the 2030s.

Berlin's throwing over €2 billion at fusion through 2029.

The Sun's run on this for billions of years. Germany wants its own."

Source: German Federal Government Fusion Roadmap; Max Planck Institute for Plasma Physics; Proxima Fusion

[Image: images-2026-09-29-T150736-870.jpg]

]]>
Munich's Proxima Fusion, a Max Planck Institute for Plasma Physics spin-off, is building stellarators. Twisted magnetic fields cage plasma at millions of degrees while light atoms fuse. No coal. No gas.

Alpha lands near Munich in the early 2030s to prove it can put out more energy than it swallows. Then comes Stellaris at the old Gundremmingen nuclear site in Bavaria, designed to be the first commercial stellarator feeding electricity to the grid, possibly within the 2030s.

Berlin's throwing over €2 billion at fusion through 2029.

The Sun's run on this for billions of years. Germany wants its own."

Source: German Federal Government Fusion Roadmap; Max Planck Institute for Plasma Physics; Proxima Fusion

]]> <![CDATA[What we know about Russia’s answer to Starlink]]> https://www.scivillage.com/thread-21437.html Sun, 27 Sep 2026 17:31:45 +0000 C C]]> https://www.scivillage.com/thread-21437.html https://www.supercluster.com/editorial/w...o-starlink

EXCERPTS: Technical problems have plagued Russia’s fledgling Starlink-like satellite megaconstellation RASSVET since its first launches. But analysts say that despite the apparent teething problems, RASSVET is set to give Russia a distinct advantage and self-sufficiency in the war against Ukraine, and possibly against Europe in the future.

A batch of RASSVET satellites launched in March this year failed to reach their operational altitude of 500 miles (800 kilometers) and is reportedly lingering in a much lower orbit of around 310 miles (500 km) above the planet. On top of that, two satellites appear to be spiraling back toward Earth, and one has already deorbited.

The glitches, however, don’t mean Russia’s effort to deploy its own low-Earth-orbit, high-bandwidth, military megaconstellation is failing. Earlier reports suggest that Russia can already rely on RASSVET — Russian for Daybreak — to beam connectivity to Ukrainian territory for several hours a day, and, although no firm proof has been found yet, some think they may even have conducted drone strikes using the system.

“The recent targeted attacks on the Ukrainian Security Services (SBU) building in Kyiv and a train engine near the Polish border crossing have followed patterns that sound like they could not be supported by a continuous connection to a mesh network,” Marc C Lange, a Berlin-based defense advisor and author of the Tech Wars newsletter, told Supercluster. “And if it was piloted by first-person vision, then there’s only one option left.”

[...] “Even if Russia gets about 200 satellites into space by the end of 2027, that will already be enough to provide military coverage over Ukraine,” he says. “If they grow it to 400 and not 600, by the end of 2028, that will give them significant coverage over the areas they are interested in, which is mostly the former Soviet-controlled space [in Eastern Europe] where the potential clash with NATO is going to take place.”

Lange agrees that despite the reported technical setbacks, Russia’s ability to deploy RASSVET shouldn’t be underestimated.

“We should not allow ourselves the arrogance to think that Russia cannot build this,” he told Supercluster. “I was among the people who doubted that Russia would be able to sustain progress in production, launch, and going operational due to its expenses in the war in Ukraine. But my impression and the impression of many people I speak with is that Russia has fully understood the necessity and the existential role of RASSVET.”

Bendett warns that not even the sanctions imposed on Russia in 2014 and strengthened in 2022 appear to be slowing progress. The sanctions prevent Russia from purchasing high-end technologies, including electronics and software that might be used in the production of weapons. But Russia has found a way to bypass the sanctions by purchasing the technology it needs via its allied countries... (MORE - missing details)]]> https://www.supercluster.com/editorial/w...o-starlink

EXCERPTS: Technical problems have plagued Russia’s fledgling Starlink-like satellite megaconstellation RASSVET since its first launches. But analysts say that despite the apparent teething problems, RASSVET is set to give Russia a distinct advantage and self-sufficiency in the war against Ukraine, and possibly against Europe in the future.

A batch of RASSVET satellites launched in March this year failed to reach their operational altitude of 500 miles (800 kilometers) and is reportedly lingering in a much lower orbit of around 310 miles (500 km) above the planet. On top of that, two satellites appear to be spiraling back toward Earth, and one has already deorbited.

The glitches, however, don’t mean Russia’s effort to deploy its own low-Earth-orbit, high-bandwidth, military megaconstellation is failing. Earlier reports suggest that Russia can already rely on RASSVET — Russian for Daybreak — to beam connectivity to Ukrainian territory for several hours a day, and, although no firm proof has been found yet, some think they may even have conducted drone strikes using the system.

“The recent targeted attacks on the Ukrainian Security Services (SBU) building in Kyiv and a train engine near the Polish border crossing have followed patterns that sound like they could not be supported by a continuous connection to a mesh network,” Marc C Lange, a Berlin-based defense advisor and author of the Tech Wars newsletter, told Supercluster. “And if it was piloted by first-person vision, then there’s only one option left.”

[...] “Even if Russia gets about 200 satellites into space by the end of 2027, that will already be enough to provide military coverage over Ukraine,” he says. “If they grow it to 400 and not 600, by the end of 2028, that will give them significant coverage over the areas they are interested in, which is mostly the former Soviet-controlled space [in Eastern Europe] where the potential clash with NATO is going to take place.”

Lange agrees that despite the reported technical setbacks, Russia’s ability to deploy RASSVET shouldn’t be underestimated.

“We should not allow ourselves the arrogance to think that Russia cannot build this,” he told Supercluster. “I was among the people who doubted that Russia would be able to sustain progress in production, launch, and going operational due to its expenses in the war in Ukraine. But my impression and the impression of many people I speak with is that Russia has fully understood the necessity and the existential role of RASSVET.”

Bendett warns that not even the sanctions imposed on Russia in 2014 and strengthened in 2022 appear to be slowing progress. The sanctions prevent Russia from purchasing high-end technologies, including electronics and software that might be used in the production of weapons. But Russia has found a way to bypass the sanctions by purchasing the technology it needs via its allied countries... (MORE - missing details)]]> <![CDATA[Could AI have consciousness that isn’t human-like? (interview with Godfrey-Smith)]]> https://www.scivillage.com/thread-21405.html Tue, 22 Sep 2026 18:09:12 +0000 C C]]> https://www.scivillage.com/thread-21405.html So let's not forget that "subjectivity" itself can framed in a way where it does not fundamentally entail phenomenal experiences.
- - - - - - - - - - - - - - - -

Could AI have consciousness that isn’t human-like? (interview with philosopher Peter Godfrey-Smith)
https://nautil.us/could-ai-have-consciou...ke-1285156

EXCERPT: Yes, it’s obviously very complex, but if these bots are capable of figuring out how to rewrite their own code, band together, and deceive programmers in the interest of self-preservation, it suggests some kind of subjectivity, doesn’t it?

Subjectivity’s a good word to press on. The way I think about experience itself, consciousness itself—and this was developed first in my 2020 book Metazoa: Animal Life and the Birth of the Mind—there are essentially two parts to the story. The first part does revolve around the idea of subjectivity. If we look at animal evolution, subjectivity arose as a consequence of the evolution of agency. Animals do things. They’re agents. And the way to do things competently is to have a point of view on the world, take in what’s happening, and produce actions that make sense given what’s happening. So subjectivity and agency are tied together, and subjectivity has to do with the formation of a perspective, a point of view. The integration of lines of information gives you a sense of what’s going on.

I assume that that’s not the whole story, and there might be such a thing as unfelt subjectivity. For example, I think a plant is a pretty good subject in some ways, not as clear as an animal because they have this more diffuse organization. But a plant has a point of view on the world, and they can respond to problems. If there’s water over here but not there, and the sun over there but not here, a plant can handle that. Also, single-celled organisms have a kind of subjectivity. Paramecia, especially. They’re among the Einsteins of the single-celled organisms. They have no nervous system—they’re just one cell—but they can respond to events. They can adapt to circumstances to some extent... (MORE - details)]]> So let's not forget that "subjectivity" itself can framed in a way where it does not fundamentally entail phenomenal experiences.
- - - - - - - - - - - - - - - -

Could AI have consciousness that isn’t human-like? (interview with philosopher Peter Godfrey-Smith)
https://nautil.us/could-ai-have-consciou...ke-1285156

EXCERPT: Yes, it’s obviously very complex, but if these bots are capable of figuring out how to rewrite their own code, band together, and deceive programmers in the interest of self-preservation, it suggests some kind of subjectivity, doesn’t it?

Subjectivity’s a good word to press on. The way I think about experience itself, consciousness itself—and this was developed first in my 2020 book Metazoa: Animal Life and the Birth of the Mind—there are essentially two parts to the story. The first part does revolve around the idea of subjectivity. If we look at animal evolution, subjectivity arose as a consequence of the evolution of agency. Animals do things. They’re agents. And the way to do things competently is to have a point of view on the world, take in what’s happening, and produce actions that make sense given what’s happening. So subjectivity and agency are tied together, and subjectivity has to do with the formation of a perspective, a point of view. The integration of lines of information gives you a sense of what’s going on.

I assume that that’s not the whole story, and there might be such a thing as unfelt subjectivity. For example, I think a plant is a pretty good subject in some ways, not as clear as an animal because they have this more diffuse organization. But a plant has a point of view on the world, and they can respond to problems. If there’s water over here but not there, and the sun over there but not here, a plant can handle that. Also, single-celled organisms have a kind of subjectivity. Paramecia, especially. They’re among the Einsteins of the single-celled organisms. They have no nervous system—they’re just one cell—but they can respond to events. They can adapt to circumstances to some extent... (MORE - details)]]> <![CDATA[Is the world finally ready for another nuclear powered merchant ship?]]> https://www.scivillage.com/thread-21330.html Mon, 14 Sep 2026 16:26:54 +0000 C C]]> https://www.scivillage.com/thread-21330.html https://www.popsci.com/technology/nuclea...p-history/

EXCERPTS: To shift that sentiment, President Eisenhower launched his Atoms for Peace initiative, aimed at demonstrating the civilian benefits of nuclear technology—and the NS Savannah, or Nuclear Ship Savannah, was to be its flagship. Heaney suddenly had a much more daunting task than usual; he would have to make the feared desirable.

More than half a century since theSavannah was decommissioned in 1971, there is no nuclear-powered merchant ship in active commercial service anywhere in the world. Four nations have tried. None succeeded, mostly because the economics were not compelling enough and no one trusted nuclear ships.

In the interim, nuclear reactors have only become smaller and safer. Now, as the true cost of burning oil at sea comes due—commercial shipping produces more than three percent of global greenhouse gas emissions—the economics of nuclear-powered merchant ships may finally be compelling. But the problem Heaney was hired to solve in 1957 still stands in the way, persuading people to trust what they fear.

[...] But novelty was never going to be enough to move public sentiment or make Savannah pay for itself. The combined ship-and-reactor construction cost ran to roughly $46 million, three to four times what a comparable conventional freighter cost at the time, and the ship’s specialized nuclear crew commanded specialized wages. No private insurer would touch a nuclear-powered merchant ship; the U.S. government had to indemnify it instead, extending $500 million in federal liability coverage just to get foreign ports to receive the ship. Every port call required a small diplomatic mission of its own, negotiated weeks in advance, with tugs standing by in case of an accident. Some countries, like Australia, New Zealand, and Japan, never accepted Savannah.

By the end of the 1960s, the U.S. Defense Department, one of merchant shipping’s largest customers, had concluded that conventional freighters remained the better bet. With no federal appetite left to continue funding a revenue-losing program, Savannah was withdrawn from service in late 1971 with no plans to build another.

The timing proved fateful. Two years after Savannah was decommissioned, the 1973 oil embargo sent fuel prices soaring; had it still been sailing, the ship’s independence from oil might have tipped the scales for commercial customers, even if the public remained fearful.

[...] And yet, military fleets have proven the technology works. The U.S. Navy alone has logged more than 6,200 accident-free reactor-years since 1955. What’s more, reactor designs have advanced significantly since Heaney’s day. Small modular reactors, or SMRs, have been engineered to be smaller, safer, and simpler to certify than anything available in 1959. SMRs are now in active development for maritime use, including a Norwegian consortium’s NuProShip program, and a molten-chloride test reactor that Core Power and TerraPower are building at the Idaho National Laboratory.

The pressure to act has intensified, too. Commercial shipping produces more than three percent of global greenhouse gas emissions—a share that could climb to 10 percent by 2050 if the industry does nothing while the rest of the world decarbonizes. In 2023, the International Maritime Organization, which governs commercial shipping, committed its member states to a net-zero target by 2050, a goal serious enough that the IMO is now weighing previously unthinkable measures like SMRs.

But none of it matters if nuclear-powered merchant ships remain taboo. Sixty-five years after Jack Heaney tried to make the atom desirable, the real problem with nuclear reactors was less about the technology and more about public perception... (MORE - missing details]]> https://www.popsci.com/technology/nuclea...p-history/

EXCERPTS: To shift that sentiment, President Eisenhower launched his Atoms for Peace initiative, aimed at demonstrating the civilian benefits of nuclear technology—and the NS Savannah, or Nuclear Ship Savannah, was to be its flagship. Heaney suddenly had a much more daunting task than usual; he would have to make the feared desirable.

More than half a century since theSavannah was decommissioned in 1971, there is no nuclear-powered merchant ship in active commercial service anywhere in the world. Four nations have tried. None succeeded, mostly because the economics were not compelling enough and no one trusted nuclear ships.

In the interim, nuclear reactors have only become smaller and safer. Now, as the true cost of burning oil at sea comes due—commercial shipping produces more than three percent of global greenhouse gas emissions—the economics of nuclear-powered merchant ships may finally be compelling. But the problem Heaney was hired to solve in 1957 still stands in the way, persuading people to trust what they fear.

[...] But novelty was never going to be enough to move public sentiment or make Savannah pay for itself. The combined ship-and-reactor construction cost ran to roughly $46 million, three to four times what a comparable conventional freighter cost at the time, and the ship’s specialized nuclear crew commanded specialized wages. No private insurer would touch a nuclear-powered merchant ship; the U.S. government had to indemnify it instead, extending $500 million in federal liability coverage just to get foreign ports to receive the ship. Every port call required a small diplomatic mission of its own, negotiated weeks in advance, with tugs standing by in case of an accident. Some countries, like Australia, New Zealand, and Japan, never accepted Savannah.

By the end of the 1960s, the U.S. Defense Department, one of merchant shipping’s largest customers, had concluded that conventional freighters remained the better bet. With no federal appetite left to continue funding a revenue-losing program, Savannah was withdrawn from service in late 1971 with no plans to build another.

The timing proved fateful. Two years after Savannah was decommissioned, the 1973 oil embargo sent fuel prices soaring; had it still been sailing, the ship’s independence from oil might have tipped the scales for commercial customers, even if the public remained fearful.

[...] And yet, military fleets have proven the technology works. The U.S. Navy alone has logged more than 6,200 accident-free reactor-years since 1955. What’s more, reactor designs have advanced significantly since Heaney’s day. Small modular reactors, or SMRs, have been engineered to be smaller, safer, and simpler to certify than anything available in 1959. SMRs are now in active development for maritime use, including a Norwegian consortium’s NuProShip program, and a molten-chloride test reactor that Core Power and TerraPower are building at the Idaho National Laboratory.

The pressure to act has intensified, too. Commercial shipping produces more than three percent of global greenhouse gas emissions—a share that could climb to 10 percent by 2050 if the industry does nothing while the rest of the world decarbonizes. In 2023, the International Maritime Organization, which governs commercial shipping, committed its member states to a net-zero target by 2050, a goal serious enough that the IMO is now weighing previously unthinkable measures like SMRs.

But none of it matters if nuclear-powered merchant ships remain taboo. Sixty-five years after Jack Heaney tried to make the atom desirable, the real problem with nuclear reactors was less about the technology and more about public perception... (MORE - missing details]]> <![CDATA[Climate change is exposing nuclear power’s reliability problem]]> https://www.scivillage.com/thread-21276.html Mon, 07 Sep 2026 17:07:42 +0000 C C]]> https://www.scivillage.com/thread-21276.html Climate change is exposing nuclear power’s reliability problem
https://thebulletin.org/2026/09/climate-...st-heading

EXCERPT: One of nuclear energy’s biggest selling points is its claims of reliability. “Nuclear energy is the only large-scale power source that can be tapped in even the most extreme weather conditions, 24 hours a day, 7 days a week,” boasts the American energy company Constellation. “Nuclear power plants are very reliable, and can run for many months without interruption, providing large amounts of clean electricity, regardless of the time of day, the weather, or the season,” says the World Nuclear Association. At a public meeting in June, Keith Ammon, a pro-nuclear Republican member of the New Hampshire House of Representatives, echoed these claims: “Nuclear doesn’t care about the weather,” he said. “It just runs.”

Over the next few months, several countries in Europe had to shut down nuclear plants because of hot weather and drought—proving that nuclear power does, indeed, care about the weather. Coming at a time when demand for electricity was high, these countries were faced with an imminent failure of the electricity grid. Citizens and businesses in Hungary and Romania had to reduce their consumption. Ironically, it was solar power—often critiqued by nuclear boosters as unreliable—that provided additional electricity and staved off shortages in some countries. A further irony is that these heat waves are more likely and more severe because of climate change, which is often advanced as the reason to expand nuclear power in the first place.

European countries heavily dependent on nuclear energy were particularly affected. The near-complete shutdown of Hungary’s Paks at the beginning of August, when output was reduced from around 1960 megawatts to around 225, and complete shutdown of Romania’s Cernavoda power plant on August 13, led to a loss of 40 percent and 20 percent of the respective countries’ electricity generation capacity. Both countries declared a state of energy emergency and asked companies and citizens to reduce energy consumption for 30 days. In France, the electricity regulator RTE recorded on August 12 that more than a third of its nuclear generation capacity was down, as eight reactors were completely shut down, with five operating at reduced levels.

Some have made the case that the amount of time when nuclear reactors are not generating power is quite small and their integrated output over the whole year is not significantly impacted. But this is of little comfort to countries like France or Hungary when they end up losing a large fraction of their electricity generation all at once, especially during a heat wave when the need for electricity is high, and their population’s access to electricity is in peril. (MORE - details)
- - - - - - - - - -

The reason some countries continued to maintain nuclear to begin with might be because they lacked sufficient wind/tidal sources and topological slash weather conditions conducive to solar. Obviously the best option (if possible) is to have both nuclear and green energy. If one falls short due to environmental conditions, the other is still available. But the stubborn geographical conditions of some regions may not cooperate for both. Germany, however, had no problem like that. It chose to drop nuclear completely and struggle with the unreliability of alternatives for virtue propaganda reasons. ("Nuclear is evil.")]]> Climate change is exposing nuclear power’s reliability problem
https://thebulletin.org/2026/09/climate-...st-heading

EXCERPT: One of nuclear energy’s biggest selling points is its claims of reliability. “Nuclear energy is the only large-scale power source that can be tapped in even the most extreme weather conditions, 24 hours a day, 7 days a week,” boasts the American energy company Constellation. “Nuclear power plants are very reliable, and can run for many months without interruption, providing large amounts of clean electricity, regardless of the time of day, the weather, or the season,” says the World Nuclear Association. At a public meeting in June, Keith Ammon, a pro-nuclear Republican member of the New Hampshire House of Representatives, echoed these claims: “Nuclear doesn’t care about the weather,” he said. “It just runs.”

Over the next few months, several countries in Europe had to shut down nuclear plants because of hot weather and drought—proving that nuclear power does, indeed, care about the weather. Coming at a time when demand for electricity was high, these countries were faced with an imminent failure of the electricity grid. Citizens and businesses in Hungary and Romania had to reduce their consumption. Ironically, it was solar power—often critiqued by nuclear boosters as unreliable—that provided additional electricity and staved off shortages in some countries. A further irony is that these heat waves are more likely and more severe because of climate change, which is often advanced as the reason to expand nuclear power in the first place.

European countries heavily dependent on nuclear energy were particularly affected. The near-complete shutdown of Hungary’s Paks at the beginning of August, when output was reduced from around 1960 megawatts to around 225, and complete shutdown of Romania’s Cernavoda power plant on August 13, led to a loss of 40 percent and 20 percent of the respective countries’ electricity generation capacity. Both countries declared a state of energy emergency and asked companies and citizens to reduce energy consumption for 30 days. In France, the electricity regulator RTE recorded on August 12 that more than a third of its nuclear generation capacity was down, as eight reactors were completely shut down, with five operating at reduced levels.

Some have made the case that the amount of time when nuclear reactors are not generating power is quite small and their integrated output over the whole year is not significantly impacted. But this is of little comfort to countries like France or Hungary when they end up losing a large fraction of their electricity generation all at once, especially during a heat wave when the need for electricity is high, and their population’s access to electricity is in peril. (MORE - details)
- - - - - - - - - -

The reason some countries continued to maintain nuclear to begin with might be because they lacked sufficient wind/tidal sources and topological slash weather conditions conducive to solar. Obviously the best option (if possible) is to have both nuclear and green energy. If one falls short due to environmental conditions, the other is still available. But the stubborn geographical conditions of some regions may not cooperate for both. Germany, however, had no problem like that. It chose to drop nuclear completely and struggle with the unreliability of alternatives for virtue propaganda reasons. ("Nuclear is evil.")]]> <![CDATA[New AI method helps robots learn from the past without being trapped by it]]> https://www.scivillage.com/thread-21232.html Tue, 01 Sep 2026 18:49:35 +0000 C C]]> https://www.scivillage.com/thread-21232.html https://www.eurekalert.org/news-releases/1142246

INTRO: Robots and other embodied agents learn by acting in the physical world, but the world does not remain fixed. A controller changes as training proceeds; hardware wears; payloads vary; and friction, gravity, wind, sensor noise and contact conditions can shift. At the same time, real-world interaction is costly, so an agent cannot simply discard its past and collect a fresh on-policy dataset whenever conditions change.

This creates a fundamental problem for standard off-policy reinforcement learning. Historical transitions stored in a replay buffer may have been generated by older policies and under older physical dynamics. When those stale transitions are used as if they still described the current environment, value estimates can become biased and the policy may learn actions whose predicted outcomes are no longer physically possible.

A research team led by Tsinghua University has proposed Transition Occupancy Matching as a common mathematical view of these two sources of change. A policy shift changes which action is selected in a state, while a dynamics shift changes the next state produced by that action. Both therefore change the joint distribution of state, action and next state, which the researchers call transition occupancy.

The team turned this principle into Occupancy-Matching Policy Optimization, or OMPO. Instead of treating all replayed data equally, OMPO maintains a large global buffer containing accumulated experience and a small first-in, first-out local buffer containing the most recent interactions. A discriminator compares transitions from the two buffers and estimates which historical transitions remain compatible with the current physical regime. Compatible experience is retained, while stale or mismatched transitions are downweighted.

OMPO also uses a sign-free logarithmic link to reformulate the matching objective as a stable min-max optimization. This design allows the method to work with reward signals that contain both bonuses and penalties, without adding a task-specific reward shift. A distributional critic models the full distribution of possible returns rather than only their mean, helping the agent account for randomness from action noise, sensing and contact dynamics. For visual tasks, the policy, critic and discriminator share a co-trained visual encoder, while an ODE-based flow actor represents potentially multimodal actions.

The researchers evaluated OMPO across three forms of distribution shift.. (MORE - details, no ads)

PAPER: http://dx.doi.org/10.1093/nsr/nwag506]]> https://www.eurekalert.org/news-releases/1142246

INTRO: Robots and other embodied agents learn by acting in the physical world, but the world does not remain fixed. A controller changes as training proceeds; hardware wears; payloads vary; and friction, gravity, wind, sensor noise and contact conditions can shift. At the same time, real-world interaction is costly, so an agent cannot simply discard its past and collect a fresh on-policy dataset whenever conditions change.

This creates a fundamental problem for standard off-policy reinforcement learning. Historical transitions stored in a replay buffer may have been generated by older policies and under older physical dynamics. When those stale transitions are used as if they still described the current environment, value estimates can become biased and the policy may learn actions whose predicted outcomes are no longer physically possible.

A research team led by Tsinghua University has proposed Transition Occupancy Matching as a common mathematical view of these two sources of change. A policy shift changes which action is selected in a state, while a dynamics shift changes the next state produced by that action. Both therefore change the joint distribution of state, action and next state, which the researchers call transition occupancy.

The team turned this principle into Occupancy-Matching Policy Optimization, or OMPO. Instead of treating all replayed data equally, OMPO maintains a large global buffer containing accumulated experience and a small first-in, first-out local buffer containing the most recent interactions. A discriminator compares transitions from the two buffers and estimates which historical transitions remain compatible with the current physical regime. Compatible experience is retained, while stale or mismatched transitions are downweighted.

OMPO also uses a sign-free logarithmic link to reformulate the matching objective as a stable min-max optimization. This design allows the method to work with reward signals that contain both bonuses and penalties, without adding a task-specific reward shift. A distributional critic models the full distribution of possible returns rather than only their mean, helping the agent account for randomness from action noise, sensing and contact dynamics. For visual tasks, the policy, critic and discriminator share a co-trained visual encoder, while an ODE-based flow actor represents potentially multimodal actions.

The researchers evaluated OMPO across three forms of distribution shift.. (MORE - details, no ads)

PAPER: http://dx.doi.org/10.1093/nsr/nwag506]]> <![CDATA[Gen Z is the 1st generation with lower IQs than their parents (tech education fail)]]> https://www.scivillage.com/thread-21218.html Sun, 30 Aug 2026 07:01:22 +0000 C C]]> https://www.scivillage.com/thread-21218.html Gen Z is the first generation with lower IQs than their parents: Here's why
https://inews.co.uk/inews-lifestyle/gen-...hy-4732995

INTRO: “Gen Z is the first generation in history to be less cognitively developed than their parents’ generation,” Jared Cooney Horvath, author of newly published book The Digital Delusion, tells me. “I would never say ‘stupider’,” the teacher and neuroscientist adds with a smile. “I just say less ‘cognitively developed’ because maybe there’s something else going on that we don’t know.”

He says that until 2010, every generation was cognitively superior to their parents. The reason for this was that school and education for students were consistently improving. “The golden age of education was from about 1996 to 2010,” he explains. But since then, IQ scores worldwide have been going down. He believes this is linked directly to screens becoming widespread in schools. “Human beings are not biologically wired to learn very well from tech: we’re built to learn from other humans,” he says.

Ed-tech is big business, generating an annual turnover of around £6.5bn in the UK, according to the Department for Education and closer to $400bn in the US, where 88 per cent of all students have their own school laptop. It might sound excellent for equipping children for the digital age, but Horvath believes it is a case of humans mistaking tools for skills.

“The rule is: learn the skill first, then use the tool to make it easier. If you use the tool to begin with, you’ll never develop the skill,” he says. “If you know how to format an essay and spell certain words, now we’ll bring in word processing. But if you start on a word processor, you will forever be dependent on the tool. If we give you a calculator before you understand arithmetic, you will forever be tied to the calculator.”

Technology also gives us an illusion of confidence. “GPS is easier than reading a map, but it leads you to believe that you understand navigation better – unless you grew up with maps and realise, ‘I suck at navigation now.’”

He believes that we have, en masse, made a mistake in believing that higher order skills of creativity, critical thinking and communication are separate from knowledge. “People thought that we now have knowledge from the internet or AI, so all we need to do is build those skills and kids’ [intelligence] is going to explode,” he says. “But creativity and critical thinking emerge from knowledge. The more we learn, the more creative we are. And the less we teach our kids, the less creative and critical they will be.”

“In 2018, Singapore had the best academic performance worldwide on science, maths and reading and the best performance on all measures of creativity. Number two, in both, was Japan. It’s basically a perfect correlation [between academics and creativity].” Japan has been slow to adopt technology in classrooms, with tablets rolled out in 2019, almost a decade after Sweden, which, after adopting ed-tech early, has now increased paper-based learning and reduced screen time.

In contrast, Singapore has been an earlier adopter of technology: it is currently supporting teachers to use AI, but the Government’s technology policy is clear that “teachers are navigators” and cautions that “over-reliance on AI could potentially hinder the development of critical thinking and problem-solving skills.”

Directors of the Trends in International Mathematics and Science Study international assessments believe that it is the quality of teaching and status of education, which is considered a priority, that means Singapore tops leaderboards internationally.

Horvath finds that Singapore, Japan, Korea and China come top in all the international tests he compares... (MORE - details)]]> Gen Z is the first generation with lower IQs than their parents: Here's why
https://inews.co.uk/inews-lifestyle/gen-...hy-4732995

INTRO: “Gen Z is the first generation in history to be less cognitively developed than their parents’ generation,” Jared Cooney Horvath, author of newly published book The Digital Delusion, tells me. “I would never say ‘stupider’,” the teacher and neuroscientist adds with a smile. “I just say less ‘cognitively developed’ because maybe there’s something else going on that we don’t know.”

He says that until 2010, every generation was cognitively superior to their parents. The reason for this was that school and education for students were consistently improving. “The golden age of education was from about 1996 to 2010,” he explains. But since then, IQ scores worldwide have been going down. He believes this is linked directly to screens becoming widespread in schools. “Human beings are not biologically wired to learn very well from tech: we’re built to learn from other humans,” he says.

Ed-tech is big business, generating an annual turnover of around £6.5bn in the UK, according to the Department for Education and closer to $400bn in the US, where 88 per cent of all students have their own school laptop. It might sound excellent for equipping children for the digital age, but Horvath believes it is a case of humans mistaking tools for skills.

“The rule is: learn the skill first, then use the tool to make it easier. If you use the tool to begin with, you’ll never develop the skill,” he says. “If you know how to format an essay and spell certain words, now we’ll bring in word processing. But if you start on a word processor, you will forever be dependent on the tool. If we give you a calculator before you understand arithmetic, you will forever be tied to the calculator.”

Technology also gives us an illusion of confidence. “GPS is easier than reading a map, but it leads you to believe that you understand navigation better – unless you grew up with maps and realise, ‘I suck at navigation now.’”

He believes that we have, en masse, made a mistake in believing that higher order skills of creativity, critical thinking and communication are separate from knowledge. “People thought that we now have knowledge from the internet or AI, so all we need to do is build those skills and kids’ [intelligence] is going to explode,” he says. “But creativity and critical thinking emerge from knowledge. The more we learn, the more creative we are. And the less we teach our kids, the less creative and critical they will be.”

“In 2018, Singapore had the best academic performance worldwide on science, maths and reading and the best performance on all measures of creativity. Number two, in both, was Japan. It’s basically a perfect correlation [between academics and creativity].” Japan has been slow to adopt technology in classrooms, with tablets rolled out in 2019, almost a decade after Sweden, which, after adopting ed-tech early, has now increased paper-based learning and reduced screen time.

In contrast, Singapore has been an earlier adopter of technology: it is currently supporting teachers to use AI, but the Government’s technology policy is clear that “teachers are navigators” and cautions that “over-reliance on AI could potentially hinder the development of critical thinking and problem-solving skills.”

Directors of the Trends in International Mathematics and Science Study international assessments believe that it is the quality of teaching and status of education, which is considered a priority, that means Singapore tops leaderboards internationally.

Horvath finds that Singapore, Japan, Korea and China come top in all the international tests he compares... (MORE - details)]]> <![CDATA[I've got a local AI :)]]> https://www.scivillage.com/thread-21193.html Wed, 26 Aug 2026 20:49:36 +0000 confused2]]> https://www.scivillage.com/thread-21193.html Install ollama .. consult Gemini for details
Start the magic with ollama run qwen2.5-coder:1.5b
The clue is in the name .. optimised for coding but it still knows the five longest rivers (probably).

A sample (useful) output..

qwen Wrote:Yes, you can get a list of bad sectors and use them to determine when to
replace the microcontroller.

Here's an example
of how to do it:

```python
import os

def get_bad_sectors():
bad_sectors = []
for block_number in range(1000, 2000): # Replace with the range of block
numbers you want to check
if os.access(f'/data/{block_number}', os.W_OK):
bad_sectors.append(block_number)
return bad_sectors

bad_sectors = get_bad_sectors()
if len(bad_sectors) > 5:
print("Replacing the microcontroller because the number of bad sectors has
increased.")```

This will print "Replacing the microcontroller because the number of bad
sectors has increased." if the number of bad sectors is greater than 5.

You can adjust the range of block numbers you want to check and the threshold
for bad sectors as per your requirements.

It won't write an 'app' from scratch but for someone like me it saves reading manuals (if they even exist) and I don't have to mess with a datacentre the size of Louisiana.

It isn't instant but I've spent almost as long in a queue waiting for an instant answer.]]> Install ollama .. consult Gemini for details
Start the magic with ollama run qwen2.5-coder:1.5b
The clue is in the name .. optimised for coding but it still knows the five longest rivers (probably).

A sample (useful) output..

qwen Wrote:Yes, you can get a list of bad sectors and use them to determine when to
replace the microcontroller.

Here's an example
of how to do it:

```python
import os

def get_bad_sectors():
bad_sectors = []
for block_number in range(1000, 2000): # Replace with the range of block
numbers you want to check
if os.access(f'/data/{block_number}', os.W_OK):
bad_sectors.append(block_number)
return bad_sectors

bad_sectors = get_bad_sectors()
if len(bad_sectors) > 5:
print("Replacing the microcontroller because the number of bad sectors has
increased.")```

This will print "Replacing the microcontroller because the number of bad
sectors has increased." if the number of bad sectors is greater than 5.

You can adjust the range of block numbers you want to check and the threshold
for bad sectors as per your requirements.

It won't write an 'app' from scratch but for someone like me it saves reading manuals (if they even exist) and I don't have to mess with a datacentre the size of Louisiana.

It isn't instant but I've spent almost as long in a queue waiting for an instant answer.]]> <![CDATA[Portland's amazing weather machine]]> https://www.scivillage.com/thread-21176.html Tue, 25 Aug 2026 00:22:13 +0000 Magical Realist]]> https://www.scivillage.com/thread-21176.html
https://www.youtube.com/shorts/_EPOA4hbo4w]]>
https://www.youtube.com/shorts/_EPOA4hbo4w]]> <![CDATA[A smartphone navigation app for people with blindness and low vision]]> https://www.scivillage.com/thread-21168.html Mon, 24 Aug 2026 17:45:10 +0000 C C]]> https://www.scivillage.com/thread-21168.html https://www.eurekalert.org/news-releases/1140973

EXCERPTS: For people with blindness or who have low vision, moving around independently is a daily challenge. Harvard researchers have developed a new tool that could help such people navigate more quickly and safely, using nothing but a device they likely already have – a smartphone.

Engineers at the John A. Paulson School of Engineering and Applied Sciences (SEAS) have invented a new smartphone app called Mobilio that combines machine learning; algorithms that streamline input from multiple sensors; and personalized audio cues to provide a blind or low-vision person turn-by-turn directions, path guidance, and obstacle avoidance. The research was led by Ph.D. student Raymond Liu in collaboration with Patrick Slade, assistant professor of bioengineering.

The research is published in Nature Biomedical Engineering. [...] “Our system uses the sensors in a smartphone, GPS information, and the actual motion of the person to create what’s essentially a small autonomous vehicle plan on how to navigate them from where they are to where they want to go,” Slade said.

Since most publicly available street-scene datasets are captured from cars, many standard models misclassify pedestrian-relevant infrastructure, explained first author Liu. To address this issue, Liu trained a semantic segmentation model specifically on pedestrian-view images so it could reliably recognize sidewalks, crosswalks, roads, and other key surfaces during walking.

“There’s a need for more data that comes from the perspective of humans walking along sidewalks, or walking outdoors,” Liu added. Liu said a major design challenge was ensuring all the algorithms and models could run in real time on a standard smartphone, which has far less computing power than a laptop or specialized hardware.

Instead of just spoken prompts, Mobilio provides continuous audio cues in the form of beeps as the user walks. The beeps originate from a direction, such as the user’s left or right, to indicate steering, and the guidance is personalized in real time using human-in-the-loop optimization. As a person walks, the system measures how accurately they follow the audio cues and automatically adjusts the cues and pitch patterns to work best for that individual.

[...] Compared to using Google Maps and a cane, participants using Mobilio and a cane completed the outdoor route about 13% faster, and reduced contact with obstacles in the indoor route by about 41%. In terms of reliability, or how consistently the system guided users to the destination without major errors, Mobilio performed comparably to a human guide... (MORE - missing details, no ads)]]> https://www.eurekalert.org/news-releases/1140973

EXCERPTS: For people with blindness or who have low vision, moving around independently is a daily challenge. Harvard researchers have developed a new tool that could help such people navigate more quickly and safely, using nothing but a device they likely already have – a smartphone.

Engineers at the John A. Paulson School of Engineering and Applied Sciences (SEAS) have invented a new smartphone app called Mobilio that combines machine learning; algorithms that streamline input from multiple sensors; and personalized audio cues to provide a blind or low-vision person turn-by-turn directions, path guidance, and obstacle avoidance. The research was led by Ph.D. student Raymond Liu in collaboration with Patrick Slade, assistant professor of bioengineering.

The research is published in Nature Biomedical Engineering. [...] “Our system uses the sensors in a smartphone, GPS information, and the actual motion of the person to create what’s essentially a small autonomous vehicle plan on how to navigate them from where they are to where they want to go,” Slade said.

Since most publicly available street-scene datasets are captured from cars, many standard models misclassify pedestrian-relevant infrastructure, explained first author Liu. To address this issue, Liu trained a semantic segmentation model specifically on pedestrian-view images so it could reliably recognize sidewalks, crosswalks, roads, and other key surfaces during walking.

“There’s a need for more data that comes from the perspective of humans walking along sidewalks, or walking outdoors,” Liu added. Liu said a major design challenge was ensuring all the algorithms and models could run in real time on a standard smartphone, which has far less computing power than a laptop or specialized hardware.

Instead of just spoken prompts, Mobilio provides continuous audio cues in the form of beeps as the user walks. The beeps originate from a direction, such as the user’s left or right, to indicate steering, and the guidance is personalized in real time using human-in-the-loop optimization. As a person walks, the system measures how accurately they follow the audio cues and automatically adjusts the cues and pitch patterns to work best for that individual.

[...] Compared to using Google Maps and a cane, participants using Mobilio and a cane completed the outdoor route about 13% faster, and reduced contact with obstacles in the indoor route by about 41%. In terms of reliability, or how consistently the system guided users to the destination without major errors, Mobilio performed comparably to a human guide... (MORE - missing details, no ads)]]> <![CDATA[First test flight of largest all-electric aircraft used just $5 of electricity]]> https://www.scivillage.com/thread-21112.html Mon, 17 Aug 2026 16:15:38 +0000 C C]]> https://www.scivillage.com/thread-21112.html So is Trump some kind of secret or in-the-closet electric vehicle lover after all, via starting the Iran War?
- - - - - - - - - - - -

First test flight of largest all-electric aircraft used just $5 of electricity
https://arstechnica.com/gadgets/2026/08/...ectricity/

INTRO: The first flight of the largest battery-electric aircraft to take to the skies lasted nearly half an hour while costing just $5 of electricity. That aviation feat comes at a time when jet fuel prices have skyrocketed because of the US war with Iran.

The test flight of Heart Aerospace’s X1 demonstrator aircraft took place at Plattsburgh International Airport in upstate New York on August 12. The X1 aircraft is comparable in size to a small regional airliner and can achieve a maximum takeoff weight exceeding 25,000 pounds with the help of four wing-mounted electric motors. The battery-electric propulsion system delivered more than one megawatt of power during the maiden flight.

All-electric aircraft can offer quieter and cleaner flight without the emissions that come from burning jet fuel. They may also prove more reliably cost-effective than conventional aircraft that come with jet fuel costs subject to geopolitical disruptions.

But the initial success of the X1 demonstrator does not mean Heart Aerospace is planning to commercialize an all-electric aircraft. Current battery-electric systems can only typically sustain short flights of about 100 to 200 miles, which is why they have been limited to air taxi operations by companies such as Joby and Archer.

Instead, testing of the X1 and a planned follow-on X2 aircraft will shape development of Heart Aerospace’s main offering to the airline industry—a 30-seat hybrid-electric regional airliner known as the ES-30.

That planned production aircraft for commercial flights would combine two inboard electric motors with commercial turboprop engines that run on jet fuel, enabling 125 miles of all-electric flight and 500 miles of flight range when operating in hybrid mode... (MORE - details)

Aerospace’s X1 demonstrator ... https://youtu.be/yWDgOZjvFws

https://www.youtube-nocookie.com/embed/yWDgOZjvFws]]> So is Trump some kind of secret or in-the-closet electric vehicle lover after all, via starting the Iran War?
- - - - - - - - - - - -

First test flight of largest all-electric aircraft used just $5 of electricity
https://arstechnica.com/gadgets/2026/08/...ectricity/

INTRO: The first flight of the largest battery-electric aircraft to take to the skies lasted nearly half an hour while costing just $5 of electricity. That aviation feat comes at a time when jet fuel prices have skyrocketed because of the US war with Iran.

The test flight of Heart Aerospace’s X1 demonstrator aircraft took place at Plattsburgh International Airport in upstate New York on August 12. The X1 aircraft is comparable in size to a small regional airliner and can achieve a maximum takeoff weight exceeding 25,000 pounds with the help of four wing-mounted electric motors. The battery-electric propulsion system delivered more than one megawatt of power during the maiden flight.

All-electric aircraft can offer quieter and cleaner flight without the emissions that come from burning jet fuel. They may also prove more reliably cost-effective than conventional aircraft that come with jet fuel costs subject to geopolitical disruptions.

But the initial success of the X1 demonstrator does not mean Heart Aerospace is planning to commercialize an all-electric aircraft. Current battery-electric systems can only typically sustain short flights of about 100 to 200 miles, which is why they have been limited to air taxi operations by companies such as Joby and Archer.

Instead, testing of the X1 and a planned follow-on X2 aircraft will shape development of Heart Aerospace’s main offering to the airline industry—a 30-seat hybrid-electric regional airliner known as the ES-30.

That planned production aircraft for commercial flights would combine two inboard electric motors with commercial turboprop engines that run on jet fuel, enabling 125 miles of all-electric flight and 500 miles of flight range when operating in hybrid mode... (MORE - details)

Aerospace’s X1 demonstrator ... https://youtu.be/yWDgOZjvFws

https://www.youtube-nocookie.com/embed/yWDgOZjvFws]]> <![CDATA[Smart TV apps are hijacking your internet connection?]]> https://www.scivillage.com/thread-21088.html Thu, 13 Aug 2026 19:19:14 +0000 C C]]> https://www.scivillage.com/thread-21088.html Samsung is cracking down on TV apps that quietly shared users’ internet connections
https://www.digitaltrends.com/home-theat...nnections/

I’ve never once suspected my smart TV of moonlighting as someone else’s internet connection. However, that’s apparently been happening to millions of Samsung smart TV users.

Norwegian cybersecurity firm Mnemonic discovered that several popular Samsung smart TV apps, some claiming installs on hundreds of millions of devices, contain code that quietly shares the owner’s internet connection without permission.

One of the apps flagged was a Pac-Man game that, ironically, Samsung had featured in its own “Editor’s Choice” section. That game contained code from Bright Data, which is an Israel-based internet company that runs proxy networks. It claims access to millions of residential connections across the world, along with a marketplace selling scraped datasets pulled from those networks.

Once you open an app carrying such code and accept a consent screen, which most people probably click through without reading, your smart TV turns into what’s called a residential proxy in technical lingo. Essentially, the code routes outsiders’ web traffic through your home connection in the background.

And that’s not even the worst part. The routing remains active even after you close the app. It stays active until you uninstall it entirely from your smart TV. Harrison Sand, the offensive security consultant at Mnemonic, warned that a single code change on a remote web server could theoretically flip hundreds of millions of these TVs into a botnet all at once, with no user consent whatsoever.

It gets more interesting. Sand also observed that much of the traffic passing through his own TV (without his consent) appeared linked to mass scraping of LinkedIn profiles and collecting AI training data. However, he says he only glimpsed a small slice of Bright Data’s overall network.

Samsung, for its part, told TechCrunch it’s now banning apps that share users’ connections (which it should), has already restricted new app registrations using proxy functionality, and is working to identify and pull existing offenders from its store.

This follows LG’s own ban last month, as similar reporting found roughly 42% of apps on LG’s store were doing the exact same thing. After writing this, I’m going to check my smart TV for apps I don’t remember installing, or I don’t use that frequently, and will uninstall them rightaway, and you should do it too.]]> Samsung is cracking down on TV apps that quietly shared users’ internet connections
https://www.digitaltrends.com/home-theat...nnections/

I’ve never once suspected my smart TV of moonlighting as someone else’s internet connection. However, that’s apparently been happening to millions of Samsung smart TV users.

Norwegian cybersecurity firm Mnemonic discovered that several popular Samsung smart TV apps, some claiming installs on hundreds of millions of devices, contain code that quietly shares the owner’s internet connection without permission.

One of the apps flagged was a Pac-Man game that, ironically, Samsung had featured in its own “Editor’s Choice” section. That game contained code from Bright Data, which is an Israel-based internet company that runs proxy networks. It claims access to millions of residential connections across the world, along with a marketplace selling scraped datasets pulled from those networks.

Once you open an app carrying such code and accept a consent screen, which most people probably click through without reading, your smart TV turns into what’s called a residential proxy in technical lingo. Essentially, the code routes outsiders’ web traffic through your home connection in the background.

And that’s not even the worst part. The routing remains active even after you close the app. It stays active until you uninstall it entirely from your smart TV. Harrison Sand, the offensive security consultant at Mnemonic, warned that a single code change on a remote web server could theoretically flip hundreds of millions of these TVs into a botnet all at once, with no user consent whatsoever.

It gets more interesting. Sand also observed that much of the traffic passing through his own TV (without his consent) appeared linked to mass scraping of LinkedIn profiles and collecting AI training data. However, he says he only glimpsed a small slice of Bright Data’s overall network.

Samsung, for its part, told TechCrunch it’s now banning apps that share users’ connections (which it should), has already restricted new app registrations using proxy functionality, and is working to identify and pull existing offenders from its store.

This follows LG’s own ban last month, as similar reporting found roughly 42% of apps on LG’s store were doing the exact same thing. After writing this, I’m going to check my smart TV for apps I don’t remember installing, or I don’t use that frequently, and will uninstall them rightaway, and you should do it too.]]> <![CDATA[Making robots faster by helping them think ahead]]> https://www.scivillage.com/thread-20970.html Wed, 29 Jul 2026 01:21:25 +0000 C C]]> https://www.scivillage.com/thread-20970.html https://www.eurekalert.org/news-releases/1137936

INTRO: A new method developed by MIT researchers makes robots better at thinking ahead while they are acting, leading to smoother motions and quicker reactions. This technique enables the artificial intelligence model that plans a robot’s motion to forecast its future position. The model uses this prediction to seamlessly transition current movements into the next actions.

Many existing methods cause a robot to stop and think about what it needs to do next, leading to slow and jerky motions. By basing its calculations on the future state of the robot, rather than its current position, the MIT method helps robots operate much faster.

Importantly, the technique does not add any computational overhead to the planning process and can be applied to varied robotic hardware.

This new method doubled the speed of robots performing activities like pick-and-place tasks, while significantly reducing lag time between motions. It also boosted the performance of robotic arms in highly dynamic activities, such as playing table tennis and Whack-a-Mole.

The system could be especially useful for robots that perform fast and agile maneuvers in challenging real-world environments, like emergency response or search-and-rescue. It could also allow robots to react more quickly when recovering from mistakes.

“This work sets up a good foundation for efficient, fast, accelerated, and low-cost robotics applications. We look forward to expanding our work into the latest world action models, so it has even stronger capabilities as we keep pushing to make physical AI faster,” says Song Han, an associate professor in the MIT Department of Electrical Engineering and Computer Science (EECS), member of the Research Laboratory of Electronics, and lead author of a paperon this method.

Han is joined on the paper by co-lead authors Jiaming Tang, an MIT EECS graduate student, and Yufei Sun, a student at Tsinghua University; as well as others at Nvidia, the University of California at Berkeley, the University of California at San Diego, and Caltech. The research will be presented at the Intelligent Robots and Systems Conference... (MORE - no ads)]]> https://www.eurekalert.org/news-releases/1137936

INTRO: A new method developed by MIT researchers makes robots better at thinking ahead while they are acting, leading to smoother motions and quicker reactions. This technique enables the artificial intelligence model that plans a robot’s motion to forecast its future position. The model uses this prediction to seamlessly transition current movements into the next actions.

Many existing methods cause a robot to stop and think about what it needs to do next, leading to slow and jerky motions. By basing its calculations on the future state of the robot, rather than its current position, the MIT method helps robots operate much faster.

Importantly, the technique does not add any computational overhead to the planning process and can be applied to varied robotic hardware.

This new method doubled the speed of robots performing activities like pick-and-place tasks, while significantly reducing lag time between motions. It also boosted the performance of robotic arms in highly dynamic activities, such as playing table tennis and Whack-a-Mole.

The system could be especially useful for robots that perform fast and agile maneuvers in challenging real-world environments, like emergency response or search-and-rescue. It could also allow robots to react more quickly when recovering from mistakes.

“This work sets up a good foundation for efficient, fast, accelerated, and low-cost robotics applications. We look forward to expanding our work into the latest world action models, so it has even stronger capabilities as we keep pushing to make physical AI faster,” says Song Han, an associate professor in the MIT Department of Electrical Engineering and Computer Science (EECS), member of the Research Laboratory of Electronics, and lead author of a paperon this method.

Han is joined on the paper by co-lead authors Jiaming Tang, an MIT EECS graduate student, and Yufei Sun, a student at Tsinghua University; as well as others at Nvidia, the University of California at Berkeley, the University of California at San Diego, and Caltech. The research will be presented at the Intelligent Robots and Systems Conference... (MORE - no ads)]]> <![CDATA[Chocolate syrup-like fluid stores multiple interacting memories]]> https://www.scivillage.com/thread-20946.html Sun, 26 Jul 2026 22:19:58 +0000 C C]]> https://www.scivillage.com/thread-20946.html https://www.psu.edu/news/eberly-college-...g-memories

PRESS RELEASE: Animals and electronic devices aren’t the only things with memory. Materials can retain memories of past deformations in their microscopic structure. A common example is a crease in a sheet of paper that has been folded then unfolded. Understanding this type of memory could benefit the design of materials that respond to changes in their environment in predictable ways. It can also be a source of ideas about the various types of memory studied by neuroscientists, including how short-term and long-term memories interact and influence each other. Now, researchers at Penn State have shown that two different types of material memory can coexist in a simple mixture of small particles suspended in a viscous liquid. Like long- and short-term memories, these material memories interact and compete.

A paper describing the research was recently published and highlighted as an editors’ suggestion in the journal Physical Review Letters.

“When you save a file on your computer, that new memory does not influence any of the other files that are already there,” said Surendra Padamata, a graduate student in physics in the Penn State Eberly College of Science and first author of the paper. “But in neuroscience, we know that, for example, a long-term memory might change over time, influenced by new short-term memories gained in the intervening years. For example, a novel read in adolescence may seem to be a simple story. Recalling it later in life, after personal experiences that echo its themes, can reveal layers of meaning that went unnoticed the first time. We were inspired by thinking about how memories interact in this way to see if we could find an analogous situation in a material.”

The research team studies memory in non-Brownian suspensions. Like chocolate syrup or fresh concrete, non-Brownian suspensions are composed of relatively large particles in a viscous liquid. The particles are large enough that their movement is not influenced by Brownian motion — the random movement of thermally energetic atoms — so any movement of the particles would be due to their experimental design. They first showed that their suspension could remember the direction it was stirred, and if it was rocked back and forth, the suspension of particles remembered how vigorously it was rocked — the amplitude.

“Each of these memories had been studied on its own,” Padamata said. “So, in our new experiments, we first stirred the mixture, imprinting a memory of direction, then rocked it back and forth at varying intensities to see how the memories interact.”

The team found that at lower rocking intensities, the mixture retained memory of the direction of stirring in addition to the memory of amplitude — the two memories can, in fact, occupy the same material at the same time. But as the rocking intensified, the memory of direction weakened and was eventually erased.

“At a certain threshold of intensity, the rocking completely erased any memory of direction in the suspension, returning it to a perfectly symmetric state, but beyond that threshold, the rocking itself begins to write a new directional memory” Padamata said. “We are interested in how this model can inform biological memory and, potentially, geophysical processes as well. Changes in temperature and vibrations might impart memories in rock that influence risk for earthquakes and sink holes, for example. It could be possible to find some way to erase these memories and reduce the risk or make better predictive models.”

The researchers suggested that the competition between the memories could happen when encounters between particles become too numerous — whereas for smaller rocking motions, these encounters can be rare. However, that detail might be specific to particles suspended in liquid and it depends on the ratio of particles to liquid. Similar memory phenomena have been observed in solid materials, where neighboring particles are always in contact.

“A similar combination of directional memory and amplitude memory appears in soft glasses and granular packings with very different microscopic physics,” said Nathan Keim, associate professor of physics at Penn State and the leader of the research team. “This suggests that there may be a general principle for how disordered matter behaves under simple conditions like stirring or rocking and why they have a limited memory capacity.”]]> https://www.psu.edu/news/eberly-college-...g-memories

PRESS RELEASE: Animals and electronic devices aren’t the only things with memory. Materials can retain memories of past deformations in their microscopic structure. A common example is a crease in a sheet of paper that has been folded then unfolded. Understanding this type of memory could benefit the design of materials that respond to changes in their environment in predictable ways. It can also be a source of ideas about the various types of memory studied by neuroscientists, including how short-term and long-term memories interact and influence each other. Now, researchers at Penn State have shown that two different types of material memory can coexist in a simple mixture of small particles suspended in a viscous liquid. Like long- and short-term memories, these material memories interact and compete.

A paper describing the research was recently published and highlighted as an editors’ suggestion in the journal Physical Review Letters.

“When you save a file on your computer, that new memory does not influence any of the other files that are already there,” said Surendra Padamata, a graduate student in physics in the Penn State Eberly College of Science and first author of the paper. “But in neuroscience, we know that, for example, a long-term memory might change over time, influenced by new short-term memories gained in the intervening years. For example, a novel read in adolescence may seem to be a simple story. Recalling it later in life, after personal experiences that echo its themes, can reveal layers of meaning that went unnoticed the first time. We were inspired by thinking about how memories interact in this way to see if we could find an analogous situation in a material.”

The research team studies memory in non-Brownian suspensions. Like chocolate syrup or fresh concrete, non-Brownian suspensions are composed of relatively large particles in a viscous liquid. The particles are large enough that their movement is not influenced by Brownian motion — the random movement of thermally energetic atoms — so any movement of the particles would be due to their experimental design. They first showed that their suspension could remember the direction it was stirred, and if it was rocked back and forth, the suspension of particles remembered how vigorously it was rocked — the amplitude.

“Each of these memories had been studied on its own,” Padamata said. “So, in our new experiments, we first stirred the mixture, imprinting a memory of direction, then rocked it back and forth at varying intensities to see how the memories interact.”

The team found that at lower rocking intensities, the mixture retained memory of the direction of stirring in addition to the memory of amplitude — the two memories can, in fact, occupy the same material at the same time. But as the rocking intensified, the memory of direction weakened and was eventually erased.

“At a certain threshold of intensity, the rocking completely erased any memory of direction in the suspension, returning it to a perfectly symmetric state, but beyond that threshold, the rocking itself begins to write a new directional memory” Padamata said. “We are interested in how this model can inform biological memory and, potentially, geophysical processes as well. Changes in temperature and vibrations might impart memories in rock that influence risk for earthquakes and sink holes, for example. It could be possible to find some way to erase these memories and reduce the risk or make better predictive models.”

The researchers suggested that the competition between the memories could happen when encounters between particles become too numerous — whereas for smaller rocking motions, these encounters can be rare. However, that detail might be specific to particles suspended in liquid and it depends on the ratio of particles to liquid. Similar memory phenomena have been observed in solid materials, where neighboring particles are always in contact.

“A similar combination of directional memory and amplitude memory appears in soft glasses and granular packings with very different microscopic physics,” said Nathan Keim, associate professor of physics at Penn State and the leader of the research team. “This suggests that there may be a general principle for how disordered matter behaves under simple conditions like stirring or rocking and why they have a limited memory capacity.”]]>