It's pretty easy to see the problem here: The Internet is brimming with misinformation, and most large language models are trained on a massive body of text obtained from the Internet.

Ideally, having substantially higher volumes of accurate information might overwhelm the lies. But is that really the case? A new study by researchers at New York University examines how much medical information can be included in a large language model (LLM) training set before it spits out inaccurate answers. While the study doesn't identify a lower bound, it does show that by the time misinformation accounts for 0.001 percent of the training data, the resulting LLM is compromised.

While the paper is focused on the intentional "poisoning" of an LLM during training, it also has implications for the body of misinformation that's already online and part of the training set for existing LLMs, as well as the persistence of out-of-date information in validated medical databases.

Read full article

Comments

Most scientific literature is published in for-profit journals that rely on subscriptions and paywalls to turn a profit. But that trend has been shifting as various governments and funding agencies are requiring that the science they fund be published in open-access journals. The transition is happening gradually, though, and a lot of the historical literature remains locked behind paywalls.

These paywalls can pose a problem for researchers who aren't at well-funded universities, including many in the Global South, which may not be able to access the research they need to understand in order to pursue their own studies. One solution has been Sci-Hub, a site where people can upload PDFs of published papers so they can be shared with anyone who can access the site. Despite losses in publishing industry lawsuits and attempts to block access, Sci-Hub continues to serve up research papers that would otherwise be protected by paywalls.

But what it's serving up may not always be the latest and greatest. Generally, when a paper is retracted for being invalid, publishers issue an updated version of its PDF with clear indications that the research it contains should no longer be considered valid. Unfortunately, it appears that once Sci-Hub has a copy of a paper, it doesn't necessarily have the ability to ensure it's kept up to date. Based on a scan of its content done by researchers from India, about 85 percent of the invalid papers they checked had no indication that the paper had been retracted.

Read full article

Comments

It's difficult to comprehend what 1,000 cubic kilometers of rock would look like. It's even more difficult to imagine it being violently flung into the air. Yet the Yellowstone volcanic system blasted more than twice that amount of rock into the sky about 2 million years ago, and it has generated a number of massive (if somewhat smaller) eruptions since, and there have been even larger eruptions deeper in the past.

All of which might be enough to keep someone nervously watching the seismometers scattered throughout the area. But a new study suggests that there's nothing to worry about in the near future: There's not enough molten material pooled in one place to trigger the sort of violent eruptions that have caused massive disruptions in the past. The study also suggests that the primary focus of activity may be shifting outside of the caldera formed by past eruptions.

Understanding Yellowstone

Yellowstone is fueled by what's known as a hotspot, where molten material from the Earth's mantle percolates up through the crust. The rock that comes up through the crust is typically basaltic (a definition based on the ratio of elements in its composition) and can erupt directly. This tends to produce relatively gentle eruptions where lava flows across a broad area, generally like you see in Hawaii and Iceland. But this hot material can also melt rock within the crust, producing a material called rhyolite. This is a much more viscous material that does not flow very readily and, instead, can cause explosive eruptions.

Read full article

Comments

When fast radio bursts (FRBs) were first detected in 2007, they were a complete enigma. As their name implies, these events involve a very brief eruption of radio emissions and then typically silence, though a few objects appear to be capable of sending out multiple bursts. By obtaining enough data from lots of individual bursts, researchers gradually put the focus on magnetars, versions of neutron stars that have intense magnetic fields.

But we still don't know whether a magnetar is a requirement for an FRB or if the events can be triggered by less magnetized neutron stars as well. And we have little hint of the mechanism that produces the burst itself. Bursts could potentially be produced by an event in the star's magnetic field itself, or the star could be launching some energetic material that subsequently produces an FRB at some distance from the star.

But now, a rare burst has provided indications that FRBs likely originate near the star and that they share a feature with the emissions of pulsars, another subtype of neutron star.

Read full article

Comments

Recently, Congress' Select Subcommittee on the Coronavirus Pandemic released its final report. The basic gist is about what you'd expect from a Republican-run committee, in that it trashes a lot of Biden-era policies and state-level responses while praising a number of Trump's decisions. But what's perhaps most striking is how it tackles a variety of scientific topics, including many where there's a large, complicated body of evidence.

Notably, this includes conclusions about the origin of the pandemic, which the report describes as "most likely" emerging from a lab rather than being the product of the zoonotic transfer between an animal species and humans. The latter explanation is favored by many scientists.

The conclusions themselves aren't especially interesting; they're expected from a report with partisan aims. But the method used to reach those conclusions is often striking: The Republican majority engages in a process of systematically changing the standard of evidence needed for it to reach a conclusion. For a conclusion the report's authors favor, they'll happily accept evidence from computer models or arguments from an editorial in the popular press; for conclusions they disfavor, they demand double-blind controlled clinical trials.

Read full article

Comments

On Monday, Nature released a paper from Google's quantum computing team that provides a key demonstration of the potential of quantum error correction. Thanks to an improved processor, Google's team found that increasing the number of hardware qubits dedicated to an error-corrected logical qubit led to an exponential increase in performance. By the time the entire 105-qubit processor was dedicated to hosting a single error-corrected qubit, the system was stable for an average of an hour.

In fact, Google told Ars that errors on this single logical qubit were rare enough that it was difficult to study them. The work provides a significant validation that quantum error correction is likely to be capable of supporting the execution of complex algorithms that might require hours to execute.

A new fab

Google is making a number of announcements in association with the paper's release (an earlier version of the paper has been up on the arXiv since August). One of those is that the company is committed enough to its quantum computing efforts that it has built its own fabrication facility for its superconducting processors.

Read full article

Comments

On Friday, the US Department of Agriculture (USDA) announced that it would begin a nationwide testing program for the presence of the H5N1 flu virus, also known as the bird flu. Testing will focus on pre-pasteurized milk at dairy processing facilities (pasteurization inactivates the virus), but the order that's launching the program will require anybody involved with milk production before then to provide samples to the USDA on request. That includes "any entity responsible for a dairy farm, bulk milk transporter, bulk milk transfer station, or dairy processing facility."

The ultimate goal is to identify individual herds where the virus is circulating and use the agency's existing powers to do contact tracing and restrict the movement of cattle, with the ultimate goal of eliminating the virus from US herds.

A bovine disease vector

At the time of publication, the CDC had identified 58 cases of humans infected by the H5N1 flu virus, over half of them in California. All but two have come about due to contact with agriculture, either cattle (35 cases) or poultry (21). The virus's genetic material has appeared in the milk supply and, although pasteurization should eliminate any intact infectious virus, raw milk is notable for not undergoing pasteurization, which has led to at least one recall when the virus made its way into raw milk. And we know the virus can spread to other species if they drink milk from infected cows.

Read full article

Comments

2023 was always going to be a hot year, given that warmer El Niño conditions were superimposed on the long-term trend of climate change driven by our greenhouse gas emissions. But it's not clear anybody was expecting the striking string of hot months that allowed the year to easily eclipse any previous year on record. As the warmth has continued at record levels even after the El Niño faded, it's an event that seems to demand an explanation.

On Thursday, a group of German scientists—Helge Goessling, Thomas Rackow, and Thomas Jung—released a paper that attempts to provide one. They present data that suggests the Earth is absorbing more incoming sunlight than it has in the past, largely due to reduced cloud cover.

Balancing the numbers on radiation

Years with strong El Niño conditions tend to break records. But the 2023 El Niño was relatively mild. The effects of the phenomenon are also directly felt in the tropical Pacific, yet ocean temperatures set records in the Atlantic and contributed to a massive retreat in ice near Antarctica. So, there are clearly limits to what can be attributed to El Niño. Other influences that have been considered include the injection of water vapor into the stratosphere by the Hunga Tonga eruption, and a reduction in sulfur emissions due to new rules governing international shipping. 2023 also corresponds to a peak in the most recent solar cycle.

Read full article

Comments

By some measures, AI systems are now competitive with traditional computing methods for generating weather forecasts. Because their training penalizes errors, however, the forecasts tend to get "blurry"—as you move further ahead in time, the models make fewer specific predictions since those are more likely to be wrong. As a result, you start to see things like storm tracks broadening and the storms themselves losing clearly defined edges.

But using AI is still extremely tempting because the alternative is a computational atmospheric circulation model, which is extremely compute-intensive. Still, it's highly successful, with the ensemble model from the European Centre for Medium-Range Weather Forecasts considered the best in class.

In a paper being released today, Google's DeepMind claims its new AI system manages to outperform the European model on forecasts out to at least a week and often beyond. DeepMind's system, called GenCast, merges some computational approaches used by atmospheric scientists with a diffusion model, commonly used in generative AI. The result is a system that maintains high resolution while cutting the computational cost significantly.

Read full article

Comments

We're nearing the end of the year, and there are typically a flood of announcements regarding quantum computers around now, in part because some companies want to live up to promised schedules. Most of these involve evolutionary improvements on previous generations of hardware. But this year, we have something new: the first company to market with a new qubit technology.

The technology is called a dual-rail qubit, and it is intended to make the most common form of error trivially easy to detect in hardware, thus making error correction far more efficient. And, while tech giant Amazon has been experimenting with them, a startup called Quantum Circuits is the first to give the public access to dual-rail qubits via a cloud service.

While the tech is interesting on its own, it also provides us with a window into how the field as a whole is thinking about getting error-corrected quantum computing to work.

Read full article

Comments

In September, Microsoft made an unusual combination of announcements. It demonstrated progress with quantum error correction, something that will be needed for the technology to move much beyond the interesting demo phase, using hardware from a quantum computing startup called Quantinuum. At the same time, however, the company also announced that it was forming a partnership with a different startup, Atom Computing, which uses a different technology to make qubits available for computations.

Given that, it was probably inevitable that the folks in Redmond, Washington, would want to show that similar error correction techniques would also work with Atom Computing's hardware. It didn't take long, as the two companies are releasing a draft manuscript describing their work on error correction today. The paper serves as both a good summary of where things currently stand in the world of error correction, as well as a good look at some of the distinct features of computation using neutral atoms.

Atoms and errors

While we have various technologies that provide a way of storing and manipulating bits of quantum information, none of them can be operated error-free. At present, errors make it difficult to perform even the simplest computations that are clearly beyond the capabilities of classical computers. More sophisticated algorithms would inevitably encounter an error before they could be completed, a situation that would remain true even if we could somehow improve the hardware error rates of qubits by a factor of 1,000—something we're unlikely to ever be able to do.

Read full article

Comments

There's a general consensus that we won't be able to consistently perform sophisticated quantum calculations without the development of error-corrected quantum computing, which is unlikely to arrive until the end of the decade. It's still an open question, however, whether we could perform limited but useful calculations at an earlier point. IBM is one of the companies that's betting the answer is yes, and on Wednesday, it announced a series of developments aimed at making that possible.

On their own, none of the changes being announced are revolutionary. But collectively, changes across the hardware and software stacks have produced much more efficient and less error-prone operations. The net result is a system that supports the most complicated calculations yet on IBM's hardware, leaving the company optimistic that its users will find some calculations where quantum hardware provides an advantage.

Better hardware and software

IBM's early efforts in the quantum computing space saw it ramp up the qubit count rapidly, being one of the first companies to reach the 1,000 qubit count. However, each of those qubits had an error rate that ensured that any algorithms that tried to use all of these qubits in a single calculation would inevitably trigger one. Since then, the company's focus has been on improving the performance of smaller processors. Wednesday's announcement was based on the introduction of the second version of its Heron processor, which has 156 qubits (up from an earlier 133 in Revision 1). That's still beyond the capability of simulations on classical computers, should it be able to operate with sufficiently low errors.

Read full article

Comments

By now, it has been firmly established that the Earth went through a series of global glaciations around 600 million to 700 million years ago, shortly before complex animal life exploded in the Cambrian. Climate models have confirmed that, once enough of a dark ocean is covered by reflective ice, it sets off a cooling feedback that turns the entire planet into an icehouse. And we've found glacial material that was deposited off the coasts in the tropics.

We have an extremely incomplete picture of what these snowball periods looked like, and Antarctic terrain provides different models for what an icehouse continent might look like. But now, researchers have found deposits that they argue were formed beneath a massive ice sheet that was being melted from below by volcanic activity. And, although the deposits are currently in Colorado's Front Range, at the time they resided much closer to the equator.

In the icehouse

Glacial deposits can be difficult to identify in deep time. Massive sheets of ice will scour the terrain down to bare rock, leaving behind loosely consolidated bits of rubble that can easily be swept away after the ice is gone. We can spot when that rubble shows up in ocean deposits to confirm there were glaciers along the coast, but rubble can be difficult to find on land.

Read full article

Comments

How did supermassive black holes end up at the center of every galaxy? A while back, it wasn't that hard to explain: That's where the highest concentration of matter is, and the black holes had billions of years to feed on it. But as we've looked ever deeper into the Universe's history, we keep finding supermassive black holes, which shortens the timeline for their formation. Rather than making a leisurely meal of nearby matter, these black holes have gorged themselves in a feeding frenzy.

With the advent of the Webb Space Telescope, the problem has pushed up against theoretical limits. The matter falling into a black hole generates radiation, with faster feeding meaning more radiation. And that radiation can drive off nearby matter, choking off the black hole's food supply. That sets a limit on how fast black holes can grow unless matter is somehow fed directly into them. The Webb was used to identify early supermassive black holes that needed to have been pushing against the limit for their entire existence.

But the Webb may have just identified a solution to the dilemma as well. It has spotted a black hole that appears to have been feeding at 40 times the theoretical limit for millions of years, allowing growth at a pace sufficient to build a supermassive black hole.

Read full article

Comments

Earlier this week, Robert F. Kennedy, Jr. used a Zoom call to tell his supporters that Donald Trump had promised him "control" of the Department of Health and Human Services (HHS), the federal agency that includes the Centers for Disease Control, Food and Drug Administration, National Institutes of Health, as well as the Department of Agriculture. Given Kennedy's support for debunked anti-vaccine nonsense, this represents a potential public health nightmare.

A few days after, Howard Lutnick, a co-chair of Trump's transition team, appeared on CNN to deny that RFK Jr. would be put in charge of HHS. But he followed that with a long rant in which he echoed Kennedy's spurious claims about vaccines. This provides yet another indication of how anti-vaccine activism has become deeply enmeshed with Republican politics, to the point where it may be just as bad even if Kennedy isn't appointed.

Trump as Kennedy’s route to power

Kennedy has a long history of misinformation regarding health, with a special focus on vaccines. This includes the extensively debunked suggestion that there is a correlation between vaccinations and autism incidence, and it extends to a general skepticism about vaccine safety. That's mixed with conspiracy theories regarding collusion between federal regulators and pharmaceutical companies.

Read full article

Comments

Over the Northern Hemisphere's summer, the world's temperatures hovered near 1.5° C above pre-industrial temperatures, and the catastrophic weather events that ensued provided a preview of what might be expected to be the new normal before mid-century. And the warming won't stop there; our current emissions trajectory is such that we will double that temperature increase by the time the century is out and continue beyond its end.

This frightening trajectory and its results have led many people to argue that some form of geoengineering is necessary. If we know the effects of that much warming will be catastrophic, why not try canceling some of it out? Unfortunately, the list of "why nots" includes the fact that we don't know how well some of these techniques work or fully understand their unintended consequences. This means more research is required before we put them into practice.

But how do we do that research if there's the risk of unintended consequences? To help guide the process, the American Geophysical Union (AGU) has just released guidelines for ensuring that geoengineering research is conducted ethically.

Read full article

Comments

On Thursday, the United Nations' Environmental Programme (UNEP) released a report on what it terms the "emissions gap"—the difference between where we're heading and where we'd need to be to achieve the goals set out in the Paris Agreement. It makes for some pretty grim reading. Given last year's greenhouse gas emissions, we can afford fewer than four similar years before we would exceed the total emissions compatible with limiting the planet's warming to 1.5° C above pre-industrial conditions. Following existing policies out to the turn of the century would leave us facing over 3° C of warming.

The report ascribes this situation to two distinct emissions gaps: between the goals of the Paris Agreement and what countries have pledged to do and between their pledges and the policies they've actually put in place. There are some reasons to think that rapid progress could be made—the six largest greenhouse gas emitters accounted for nearly two-thirds of the global emissions, so it wouldn't take many policy changes to make a big difference. And the report suggests increased deployment of wind and solar could handle over a quarter of the needed emissions reductions.

But so far, progress has been far too limited to cut into global emissions.

Read full article

Comments

Colossal, the company founded to try to restore the mammoth to the Arctic tundra, has also decided to tackle a number of other species that have gone extinct relatively recently: the dodo and the thylacine. Because of significant differences in biology, not the least of which is the generation time of Proboscideans, these other efforts may reach many critical milestones well in advance of the work on mammoths.

Late last week, Colossal released a progress report on the work involved in resurrecting the thylacine, also known as the Tasmanian tiger, which went extinct when the last known survivor died in a zoo in 1936. Marsupial biology has some features that may make de-extinction somewhat easier, but we have far less sophisticated ways of manipulating it compared to the technology we've developed for working with the stem cells and reproduction of placental mammals. But, based on these new announcements, the technology available for working with marsupials is expanding rapidly.

Cane toad resistance

Colossal has branched out from its original de-extinction mission to include efforts to keep species from ever needing its services. In the case of marsupial predators, the de-extinction effort is incorporating work that will benefit existing marsupial predators: generating resistance to the toxins found on the cane toad, an invasive species that has spread widely across Australia.

Read full article

Comments

If you're using a large battery for a specialized purpose—say grid-scale storage or an electric vehicle—then it's possible to tweak the battery chemistry, provide a little bit of excess capacity, and carefully manage its charging and discharging so that it enjoys a long life span. But for consumer electronics, the batteries are smaller, the need for light weight dictates the chemistry, and the demand for quick charging can be higher. So most batteries in our gadgets start to see serious degradation after just a couple of years of use.

A big contributor to that is an internal fragmentation of the electrode materials. This leaves some of the electrode material disconnected from the battery's charge handling system, essentially stranding the material inside the battery and trapping some of the lithium uselessly. Now, researchers have found that, for at least one battery chemistry, it's possible to partially reverse some of this decay, boosting the remaining capacity of the battery by up to 30 percent.

The only problem is that not many batteries use the specific chemistry tested here. But it does show how understanding what's going on inside batteries can provide us with ways to extend their lifespan.

Read full article

Comments