Dear Aventine Readers,

Over the summer hopes were raised — and then dashed — that one of the great quests of science had been reached: the discovery of a superconductor that does not require extremely low temperatures to function.  Superconductors don’t often enjoy moments in the sun, so this month we dive into why this discovery would have been world-changing and explore the century-plus slog (now energized) in this ongoing and mythic pursuit. 

Also in this issue: small modular nuclear reactors, or SMRs. Some people believe they could be the answer to the current logjam in creating more nuclear power. Others don’t think they’ll ever become commercially viable. Read what five experts in the field say, below. 

And if you only read one thing, skip ahead to our summary of a lovely MIT Technology Review story about how a startup in the U.K. figured out how to use heat generated by personal computers to power home hot water tanks.

Thanks for reading,

Danielle Mattoon
Executive Director, Aventine

The early reports sounded too good to be true. In July, a group of researchers from South Korea claimed to have discovered a superconductor — a material that conducts electricity without energy loss — that worked at room temperature and pressure. It sent the usually complex and esoteric topic of superconductivity crashing into newspaper headlines around the world.  After decades of hope and disappointment, some people thought that the holy grail of physics — and a potential solution to our energy prayers — had finally arrived.   

It hadn’t. In the weeks following those first ebullient headlines, research labs refuted the evidence supporting the claims around the newly discovered superconductor LK-99. But in the meantime, global interest had been sparked: Why the hoopla and headlines? Why the bated breath around replicating the results?  What would it have meant if LK-99 had been the real deal?

To answer some of these questions, this month we’re focusing on superconductors — what they are, what they could be and why, next time you see them in the headlines, it could mean that the world is about to change. 

To begin, let’s start with regular conductors — materials that allow an electrical charge to pass through them. These are typically metals, with copper being the most common and effective. But in materials like copper, energy is lost as an electrical current passes through it; electrons bounce around the material’s crystal structure and with each collision, some of their energy is lost and converted into heat. 

Superconductors, on the other hand, allow electrons to pass freely through their structure, allowing for what’s known as lossless power transmission. Many superconductors can also carry extraordinarily high volumes of electrical current, which makes them suitable for building very strong electromagnets. The rub is that most superconductors in use today — the most prevalent being the compounds niobium–tin, discovered in 1954, and niobium–titanium, discovered in 1962 — can operate only at temperatures below -250 degrees Celsius.  Achieving such low temperatures requires large, complex and expensive cooling systems that use liquid helium. Superconductors that can operate at higher temperatures, meanwhile, can only do so under high pressure, a state that is even more difficult to achieve. This means that superconductors are currently expensive, require complex engineering to make use of and are reserved for only a few, very specific tasks, like MRI machines, particle accelerators and experimental nuclear fusion devices. 

So what would it mean if there were a superconductor that could operate in temperature and air pressure ranges that didn’t require extraordinary and costly measures  to maintain? 

For starters, lossless power transmission would create energy savings around the world that in the U.S. alone would equal the amount of energy created by several nuclear power plants.  Such a material would also reduce the operational cost of devices that currently use low-temperature superconductors, like MRI machines, and make more efficient the devices that currently rely on traditional electromagnets, like wind turbines. It could unlock new technologies such as nuclear fusion by making it far easier to build the components such complex technologies require. In short, such a discovery would significantly reduce the amount of energy required to do just about anything, unleash untapped potential in countless industries and open doors to entirely new sources of energy creation.  “There’s lots of ways … high-temperature superconductors can enter technology, [or bring ] new technologies into the mix, that can really help us with some of these incredibly ambitious [climate] targets that we've got,” said Susannah Speller, a professor of materials science at the University of Oxford. 

But LK-99 wasn’t the real deal, and the slog to discover higher-temperature superconductors continues. Experts who spoke with Aventine explained that progress is erratic, often serendipitous, and fraught with engineering headaches, but they also expressed great excitement about the near future of superconductors, united in the belief that such materials will indeed change the world as we know it.  

The majority of practically useful superconductors are concocted in labs, and often by accident. Superconductivity was first observed by chance in 1911 by the Dutch physicist Heike Kamerlingh Onnes, who noticed the phenomenon in wires made from mercury held at -269 degrees Celsius. Since then, researchers have gradually found families of superconductors that operate at higher temperatures, but “mostly through serendipity,” said Richard Greene, a professor in the department of physics at the University of Maryland, whose work has focused on high-temperature superconductors. In the 1980s, a family known as the cuprates  — based around copper oxides — was discovered; they superconduct at around -140 degrees Celsius. In the last decade, a family known as the hydrides — united by the presence of hydrogen in their structure — has emerged, which can superconduct at temperatures as at around -20 degrees Celsius.

But while the hydrides may superconduct at highly desirable temperatures, they only do so at pressures about a million times greater than atmospheric pressure — making them essentially unusable. “Getting to low temperature is much easier than getting to high pressure in terms of practical applications,” explained Speller. Meanwhile, the cuprates — many of which could be cooled by liquid nitrogen, which is “actually really easy and cheap to work with” compared to liquid helium, said Speller  —  have proven stubbornly difficult for other reasons. “They have the mechanical properties of something like a teacup, and [that’s a problem because] we need to make them in lengths of thousands of kilometres, where we can bend them and make them into magnets or bind them into cables,” she said.

The latter point gets to a fundamental issue with all newly discovered superconducting materials. “There's absolutely no guarantee that a room-temperature superconductor that we find would be technologically relevant,” said Kyle Shen, a professor of physical sciences at Cornell University. “It might just be a curiosity.” Even a promising new material could take years or decades to be put into practical use according to the experts Aventine spoke to. 

Perhaps the most popular focus of study among those searching for room-temperature superconductors is the hydride family. While those that superconduct at high temperatures currently do so only at extraordinarily high pressure, there is hope that a hydride might be discovered that doesn't have this requirement, Greene said. Or, he added, it may be possible to engineer a material whose structure at the atomic level gives rise to localized high pressures that result in similar effects.

There are other approaches. Some groups, including Shen’s, are borrowing engineering techniques from the world of semiconductors to craft new materials at the atomic scale, rather than growing large crystals of material, which is how most superconductors have historically been created. By carefully laying down single sheets of atoms on top of one another, Shen explained, it’s possible to tune the materials to exhibit certain properties — such as unusually high strains in interatomic bonds — that may give rise to superconductivity. Still other researchers take a more traditional approach, said Greene, using understanding of existing materials, condensed matter theory and — to some extent — informed hunches about what might work to develop new materials. Sometimes they work; often they don’t. “It's a bit like looking for a needle in a haystack,” said Speller.

Despite all this complexity, spirits run high among those working in the field. “The superconducting community is very energized at the moment because they see their technology being of importance,” said Chris Grovenor, a professor of materials at the University of Oxford. In particular, Speller remarked, demand for existing superconductors is soaring as private companies and research institutions push forward with nuclear fusion experiments. That’s expected to drive down the cost of what are still highly expensive superconducting materials, in turn making them more widely available for different applications — and potentially attracting further investment in superconductivity research.

In the face of  all the challenges, excitement about finding the elusive room-temperature, ambient pressure superconductor doesn’t seem to dim. “We hold out hope,” said Shen.

Subscribe

Subscribe to our newsletter and be kept up to date on upcoming Aventine projects

Brain-to-speech decoders set new records. For anyone who has lost the ability to speak — such as those who have experienced a stroke or developed the motor neuron disease amyotrophic lateral sclerosis (ALS) —  the most advanced technological solutions are currently able to provide brain-to-computer communication of  just 18 words per minute. (Typical speech is around 160 words per minute.) Now, two research groups — one from Stanford University, the other from UC San Francisco — have independently published papers describing brain-computer interfaces that can render intended speech from brain signals at a rate of 62 and 78 words per minute respectively.  Both groups achieved this breakthrough thanks to AI systems trained to identify brain signals associated with words and certain sounds.  (Currently available technology identifies brain signals associated with handwriting movements.)  Researchers who weren’t involved in the research told Science that the advance puts the technology “within striking range” of becoming a commercially viable product. (Spectrum, Nature)

How we might detect consciousness in AI. Ever since ChatGPT burst onto the scene with its uncanny ability to imitate the way people write, there’s been increasing interest in whether AI can achieve sentience — a topic we explore in depth in episode 7 of our podcast, Humans vs. Machines. That episode features Blake Lemoine, the Google engineer who was fired for insisting that the large language model he was working on, known as LaMDA, was sentient, citing the machine’s ability to produce new and novel language and its “inner life and self-awareness” as evidence. Such criteria are subjective and hard to prove. And they did not convince our skeptical host, Gary Marcus.  But Marcus doesn’t think sentience is forever off the table, nor do many others. In fact, as Nature reports, to prepare for such an eventuality, a team of 19 neuroscientists, philosophers and computer scientists  published a checklist of sorts that could be used to determine consciousness in AI, based on six neuroscience models used to describe consciousness in humans, including how people direct attention and update their beliefs based on new information. The more an AI system behaves in ways that match the models, the researchers suggest, the more likely it is that the system is conscious. This work is by no means the last word on the topic, but it does represent a new level of rigor in the way we think about sentience emerging in AI. And if you’re wondering: According to the work, there’s only “a relatively weak case” that AI agents like ChatGPT and LaMDA demonstrate consciousness.

A new energy-saving AI chip. Artificial intelligence is computationally demanding: companies like OpenAI spend millions of dollars running calculations on thousands of chips to train cutting-edge algorithms, and consume huge quantities of electricity in the process. But a new chip developed by IBM and described in the journal Nature hints at a potential alternative future. The new memory chip can be configured so that calculations that would usually require passing information from memory to a processor are instead all performed on the chip. This makes it 14 times more energy efficient at performing operations than conventional chips. (IBM attempts to provide a simplified explanation of exactly how it works if you’re interested in reading more.) The sticking point: All modern computational infrastructure is digital, so to take advantage of this sort of advance would require new forms of hardware and software architecture, Hechen Wang, a research scientist at Intel Labs, told IEEE Spectrum. But given the proliferation of AI, the investment might still be worth it.

Since the 1980s, development of nuclear power plants has been nearly stagnant in the Western world. Despite the technology’s potential as a constant and almost limitless source of clean energy, social controversies surrounding nuclear safety and waste have limited technological advances and plant development. Just 54 operating nuclear power plants in the United States provide about 18 percent of the country’s energy, according to data from the U.S. Energy Information Administration. As the number of new nuclear projects has declined, each of them has become increasingly expensive to develop. That is in part because with fewer projects materials often become more expensive, and also because increased regulatory and safety expectations have added costs. The result has been nuclear projects that reliably blow through their projected budgets and construction timelines.

But over the last decade, designs for smaller, easier to construct and cheaper nuclear power sources have gone from fantasy to near reality. Theoretically, small modular nuclear reactors, often known as SMRs, offer a solution to the huge, unpredictable budgets and timelines required to build  larger power plants. This leads to more certainty about the cost of construction and more flexibility about where plants could be built and how they could be integrated into the power grid. Experts in the field say that reactors could be constructed from modular, factory-made parts that could be assembled like a Lego set. They would produce somewhere around 300-500 megawatts of power, less than half of the one gigawatt produced by most current plants. Because of the lower power output, their designs could be more useful for flexible power grids with variable needs (especially those that depend wholly on solar and wind, where energy supply can be unpredictable) and also make them naturally safer, alleviating public fears induced by accidents at Fukushima, Chernobyl and Three Mile Island. While four of these SMRs are in late stages of construction in Russia, China and Argentina, the United States and other Western countries have been slower to embrace the technology. Only one SMR design has been approved by regulators in the United States.

The barriers to adoption of these reactors are still significant. Some researchers told Aventine that they are skeptical that there is enough demand to justify the factory production of SMRs and that the cost of running one will only become competitive with wind and solar if SMRs are widely deployed. Another obstacle is that even though their smaller size and lower building costs make SMRs a viable vehicle for private investment, the long history of public funding of the technology means that few experts in the field of nuclear power are equipped to design and market a product to the private market. Public and government safety concerns about new, untried reactor designs have also kept most SMR plans from achieving regulatory approval. 

Aventine asked five SMR experts for their perspectives on the barriers and incentives that will determine the future of the industry. Their comments are edited for clarity and brevity.

1. Ukraine. Faced with limited resources, the arms industry in Ukraine has been forced to adopt innovations that even MacGyver would be proud of. The Economist reports that a cottage industry of testing and building bombs has sprung up in the nation — one that makes liberal use of 3D printing and software modeling. So-called candy bombs are designed on computers and their casings 3D printed, before being filled with explosives such as C4 — still plentiful, even if conventional bombs aren’t — and shrapnel. One team of amateurs has built 30,000 of the bombs in just four months. Their designs are sometimes informed by computer modeling to determine the lethality of different shrapnel types, according to one source who spoke to The Economist.

2. U.K. Computers and laptops can generate an annoying amount of heat, especially inside stuffy offices on hot days.  But crank up the number of chips a computer uses and amp up the kinds of tasks it’s asked to perform and the heat that’s pumped out quickly escalates — to the point at which it might actually be useful. A startup in Godalming, U.K. — called Heata and funded by the British government agency Innovate UK — has developed a small server that can be installed inside people's homes. While it performs cloud computing operations for Heata’s corporate customers by connecting to a home’s WiFi network, it also pumps waste heat into its domestic customers' hot water tanks. So far, the company has installed 80 such units in homes in the U.K.. MIT Technology Review reports that each server can save 1 ton of carbon dioxide equivalent per year and reduce a homeowner’s energy bills by $300 over the same period by providing about 80% of an average household’s daily hot water.

3. Everywhere. For the last decade, Nature reports, a group of people with type 1 diabetes have been doing what medical device companies wouldn’t: writing code that turns their insulin pumps into autonomous systems that can accurately predict how much insulin a person needs and administer it. Under normal conditions the pumps require regular manual programming based on a user’s diet, exercise and so on. The DIY coding takes data from the constant glucose monitors these people already wear to control their ongoing insulin dosing. Clinical trials have shown that the technology is effective, and this year the FDA granted clearance for one of these open-source algorithms to be developed for wider use. Commercial devices now exist that provide similar functionality, but are still less advanced than some of the more cutting-edge versions of the DIY devices, which are increasingly operating almost autonomously, according to Nature.

The Ethics of AI on the Battlefield, from MIT Technology Review
4,150 words, 18 minutes

Killer robots may, for now at least, remain a thing of sci-fi nightmares. But artificial intelligence is increasingly making its way into military hardware, which could help everyone from foot soldiers to military officials make life-or-death decisions. It could be a gunsight that provides a sniper with automated target identification, or a smart war room that prompts a general to dispatch a rocket attack on a newly identified fleet of enemy vehicles. The question, though, is how should humans use that information to make decisions? What happens if they follow the AI’s advice and it’s wrong? Conversely, what happens if a human errs and the delay results in civilian death? Can — or should — AI even grapple with these kinds of issues at all? There are no simple answers, but this essay from MIT Technology Reviews explores these quandries and more.

What OpenAI Really Wants, from Wired
9,400 words, 40 minutes

OpenAI soared from being an esoteric nonprofit research lab to a $29 billion company defining the bleeding edge of artificial intelligence in just seven years. If you’re fascinated to find out how that happened, then this story by Steven Levy — arguably the tech reporter with the best access in Silicon Valley — is essential reading. Levy relays conversations with OpenAI leadership (including co-founder and CEO Sam Altman), Elon Musk, and Microsoft CEO Satya Nadella, among others. It’s a surprisingly breezy read and it gets at an important truth: The company, originally founded on lofty academic ambition and a desire to ensure that AI was developed safely has gradually morphed into another Silicon Valley behemoth. However much its leaders argue it hasn’t lost its way, it’s hard to take their words without a large pinch of salt.

The Promise and Peril of Gene Therapy for Eternal Youth, from Proto.Life
3,300 words, 14 minutes

There’s a world envisioned by some startups in which you hit a certain point in your life and can simply pause the aging process through the use of gene therapy. Still others suggest that you might be able to hit rewind if you’ve aged further than you’d like. While this was little more than a pipe dream a decade ago, gene therapy increasingly seems to be presenting a path toward dramatically slowing the aging process, if not stopping it altogether. But as this piece from Proto.Life describes, there’s plenty to think about before we all embrace immortality, from the side effects and sky-high costs of current approaches to the societal upheaval that could be wrought by a population that lives forever (or at least too long).