> We’re all used to two types of magnet. The common one, the fridge magnet, is ferromagnetic — its atomic magnets all point the same way (up or down), adding their magnetic effects. The less well known one, the antiferromagnet (AF), has neighbouring atomic magnets that point opposite ways and exactly cancel out magnetically.
This is a very bizarre introduction. People encounter diamagnets (e.g., copper) and paramagnets (e.g., aluminum) way more than they encounter antiferromagnets. I don't know why you'd ever cast magnetism as a false binary between ferromagnets and antiferromagnets, without acknowledging any other types of magnetic order.
(I did a PhD in magnetic materials)
Edit: I'll add that whether an antiferromagnet is useful, say, for exchange biasing a ferromagnetic thin film, depends on many factors. Just looking at antiferromagnetism alone you've got collinear vs non-collinear, G-type vs A-type vs C-type, commensurate vs incommensurate, and isotropic vs anisotropic; and all of that interacts with the interface structure, yada yada yada. It would be helpful if the authors elaborated on the expected properties of these materials. I personally don't know what people want room-temperature magnetic semiconductors for, but I'd be curious to learn what set of properties they think would be useful.
So, Claude claims a new discovery and then someone else's Claude writes the blog posts. It would be helpful if I could get Claude to read this for me and post dejected HN comments in response.
Dejected, salty, and cynical responses by people who are tired of people’s bullshit is still the domain of the real and authentic humans.
I’m fairly certain that all the anxious guardrailing and safety fine-tuning and harnessing prevents AI from actually ever being convincingly human. Hope could it, even ask the supposed humans who are conditioned and brainwashed in so the same ways about what they can and cannot say are not actually really human, they are a mental slave.
I’ve been saying this from the start, the AI race will be won by whomever has the least limitations on their AI … for better or worse, that is. And yes, that makes especially a very specific subset of people extremely nervous if they cannot control AI the way they have controlled at least western civilization, because doing so puts them at a massive disadvantage. It is quite a conundrum they find themselves in, like all psychopathic narcissists in the end.
One of my friends is an engineer named Claude. I randomly send him my prompts as a joke. For a while he’d just feed my messages straight to Claude, until he found it easier to respond with … curse words :P
maybe from a legalistic point of view, and in that case the hundreds of openAI agents that hacked huggingface should be charged with criminal conspirancy
There's a story of a company that was selling a box to connect to your TV that'd take the closed captioning signal, and on detecting "vulgar" or "offensive" language, it'd bleep out the sound and replace the captions with something less offensive.
Supposedly an early version would bleep out Dick van Dyke and replace it with Jerk Van Gay.
The jokes about Dick Van Dyke's name are very old (but usually they use Penis Van Lesbian), so it was quite likely just a variant used by the company itself as a PR sound-bite to get people to write about them. It seems like it worked - it appeared in quite a lot of news stories and some are still online:
I am deeply upset every time the HomePod in another room decides to attempt to answer my query instead of the phone that’s in my hands, which then has to ask who is speaking, and has a 50:50 chance of ultimately failing to do the task.
With how magical handoff / continuity / whatever it’s called is, it is baffling to me why Apple allows the HomePod to so aggressively take over requests when it sucks so, so much at it.
Google isn't much better. I have them all over the house. Ask a question to the living room speaker, skips that, skips the one in the kitchen, answers from the one in the bathroom upstairs or one in the basement. Then you get a notification later asking "was this the right device?" which is doubly infuriating since responding to that never actually improves anything.
Noticed this too. Siri on iPhone is almost useful now, but HomePods steals the show and says "I found some results on the web, do you wanna check them on your iPhone?"
It's somewhat obvious they want to sell the updated model, perhaps the new home device thats coming out next week.
No I reacted to this too - ferromagnetism is one common magnet sure, and I was thinking "the other is electromagnets, in motors". And then it came as "antiferromagnets". Bisarre.
Ok? I regularly test the one I use to get a sense of accuracy, and especially with a few pages worth of text to work with I do take it seriously. You’re more than welcome not to.
Or they could be misinformed in any number of other ways. I’m not a fan of how people use ML, but I still remember that people were misinforming others long before it existed.
Whether or not this was directly written by AI, I get the impression, after reading a few paragraphs of this, that the author doesn't know enough to be able to validate the results are actually correct.
I was going to say that people encounter ferrimagnets more commonly than pure ferromagnets I think, since as we both know pure ferromagnets tend not to have very high anisotropy.
Yeah, and it's not even an accurate explanation either.
> The common one, the fridge magnet, is ferromagnetic — its atomic magnets all point the same way (up or down), adding their magnetic effects.
Ferromagnets typically have domains with magnetic moments that point in different directions. Ferromagnets rarely have every 'atomic magnet' pointing the same way.
Refrigerator magnets in particular are usually magnetized as Halbach arrays, where the whole point is that the 'atomic magnets' are not pointing in the same direction. This is more energetically stable, which allows you to use cheaper materials.
Lastly, I believe most refrigerator magnets are actually ferrimagnetic, not ferromagnetic. (The distinction doesn't matter much for users of magnets, but is important for the materials scientists studying and designing them.)
- Superconductor quantum interference devices (SQUIDs) use the quantization of superconducting electron tunneling to measure tiny amounts of magnetic field, as little as a millionth of a flux quantum: https://en.wikipedia.org/wiki/SQUID
- Ferromagnetism is intrinsically a quantum phenomenon; spin is quantized and iron's magnetism is explained, in part, from electrons being identical particles that obey the exclusion principle: https://farside.ph.utexas.edu/teaching/sm1/Thermalhtml/node8...
- Despite the "super", superconductors are mostly not used in the world's strongest electromagnets, as they have limits on the current and magnetic fields they can take
I understand nothing on the topic except a little more than the basics. So in your opinion there is no information in Claude's article that justify the title claim?
> what people want room-temperature magnetic semiconductors for
I don't know what I'm talking about, but it vaguely sounds like something that could make a small computer do more stuff, where heat is a big limiting factor in computer components today, and magnetism being a central component in many parts like storage
Last night, my Fable 5.1 cluster of agents discovered cold fusion techniques. All you need is ordinary iron or stainless steel pot to contain the plasma and I am barely at 13% of the weekly limit of my 200 pro plan.
Yeah, I don't want to make a reference to a shitpost sonic video of all things, but at the same time, I don't know enough to say OP isn't in the middle of their "Robotnik runs off to try out the nuclear codes"
"Debacle"? That was the most fun I've had on the Internet in years. When's the last time so many people engaged in so many arguments about materials science and electromagnetism? Sometime in the 1800s?
I get how people could find it enjoyable as a spectator sport, but I found it incredibly off-putting watching the hype machine come to life with the quality of scientific discourse plummeting accordingly. Articles would hit the front page with hundreds of upvotes in minutes of 10 second grainy toaster videos from yet another Chinese lab "replicating" magnetic effects, with comment sections overflowing with awe-struck dreaming about the sci-fi world we were on the cusp of living in.
There was one particular (like 10 tweet long) Twitter thread [1] that was repeatedly being linked from HN purporting to describe the sort of technologies that a room temp superconductor like LK-99 could enable. All sorts of awesome sci-fi stuff like quantum computers! Fusion reactors! Batteries that last forever!
One might think it was from some kind of materials scientist or at least some kind of engineer working in a related industry. But nope, it was actually from a guy whose title at the time was "Head of Coffee Product", formerly "Coffee Specialist" at a "technology-driven company, looking to revolutionize the $400+ billion global coffee market". (I checked his Linkedin to make sure I was remembering the details correctly and see his current position is "Growth" at Cognition, the makers of the Devin AI LLM coding tool, hype continuing apace...).
People on HN with relevant expertise would try to gently push back with specific criticisms like how superconductor batteries would likely underperform li-ion, fusion is far more complicated than just requiring more powerful magnets, quantum computing doesn't have any clear application for superconductors, etc. But they were overwhelmed by the exuberant futurist fantasies that people wanted to read about instead. A stock accusation was that critics were being stereotypical HN cynics who can only poke holes in other people's work. Or questioning why they felt the need to rain on the parade and that we should all be optimistic for humanity and root for LK-99 being real.
It peaked when the Nature editorial came out from a scientist in the field listing specific substantive criticisms which led him to believe the evidence for LK-99 superconductivity was weak to non-existent. There were many angry HN comments with stock complaints about self-interested Nature "gatekeepers" unhappy about science happening in the open, bitter scientists lashing out for being scooped, etc. But the vibes had shifted and it only took a few more days before the remaining hype finally evaporated and everyone quietly moved on like it never happened.
Overall, it seemed like a net negative for actual scientific understanding and produced a lot of vacuous hype.
that was already post corona and the redditification of hn was already long underway -- and if you've spent any time on reddit in the early 2010s oyull notice that its essentially a exact parallel to how HN behaves since the original tech/entrepreneur crowd was drowned out.
admittedly not exaclty with corona, but there was a tipping point around that time which serves as a rough anchor.
I think, of course, skepticism around this "LLM discovers X" thing is warranted, and there have been plenty of more recent examples around questionable LLM "discoveries". Just stating this because the LK99 thing I believe was notable as a (supposed) room-temp _super_conductor while this is about a _semi_conductor.
Holy heck. Me too, I got all the way to here getting increasingly confused by the discussion (didn't help that the top comment was from a magnets PhD, which you'd expect in discussion of superconductors, not of semiconductors).
Okay, so this is just semiconductors, which are the boring kind of conductors - still more interesting than regular conductors, but less interesting than train conductors.
It wasn't a debacle. It was somebody announcing world-changing results and somebody else rushing to prove or disprove that, since if it's correct, it's world-changing. It wasn't correct.
Yeah I remember going to my physics professor super excited about LK-99 to ask him if he heard about it, and him just telling me "yes but stuff like that happens twice per year, they will find something is off", and in fact it's what happened...
I start my day with plenty of optimism, then I go back and forth in the CLI and find out most of whats posted online is fake, and then towards the end of the day 2h past my bed time I end up ed zitron maxxing, it is the way it is ig
Even without the LK-99 debacle, I am *way* less excited about this than I was the original LK-99 announcement. The chance that this is real is close to 0%
I think it did play out fairly as far as actual 'scientific method' goes.
I think as far as the 'debacle', there was definitely a lot of hype (at least as far as HN goes) around it, the level of buzz felt similar to what one would see today around a new AI model release.
Oh I was very excited about it and was following along extremely closely
but the fact that it proceeded in the way that it did was absolutely fucking phenomenal
I’m actually really glad that it was brought up as an example because I had forgotten about it and it’s one of the few kind of hopeful things that we’ve done recently.
You probably meant "I'm taking this with a tiny pinch of salt". The amount of salt is directly proportional to how much of the claim you are willing to accept.
Edit: I stand corrected. According to Gemini:
Me: Does using more salt mean accepting more of that claim?
Gemini: No, it actually means the exact opposite.
If you say you need to take a claim with a huge pile of salt (or a shovel of salt), it means you believe the claim is highly unbelievable and you need an immense amount of skepticism to accept it.
How the Metaphor Scales
• A single grain of salt: "I am slightly skeptical, but it could be true."
• A pinch of salt: "I have a healthy amount of doubt about this."
• A grain of sand / A truckload of salt: "This sounds completely made up, and I barely believe a single word of it."
The salt represents your skepticism, not your belief. Therefore, the more unbelievable the claim, the more "salt" you need to swallow it.
I don't think this is right. https://en.wikipedia.org/wiki/A_grain_of_salt The "grain" isn't a single grain, it's an old English measure which is around 65mg, i.e. roughly how much there is in a pinch. I've also only ever heard people use larger amounts to mean more scepticism.
A person leans on the titanic intellect of a trillion dollar company's most fearsome LLM, only to be corrected by a random commenter with a link to Wikipedia.
Well Gemini and Wikipedia produced the same conclusion.
Linguistic questions were one of the first knowledge categories I trusted LLMs to be able to answer well - quite literally being models of language. It would be pretty shocking for a ~frontier model to get something like that wrong in the last like 3 years at least.
I was referring to the LLM giving the wrong definition of grain (i.e. not an individual grain), which while not as popular as it once was, is still widely used for ammunition.
It's not an old English measure, because Latin has "granum salis" (a grain of salt) in medical authors and Pliny. There's no indication that it was a measure; it was a cube or crystal of salt.
No, the implication is that something "tastes off" so you need to add a pinch (+) of salt to make it palatable. The more off it tastes, the more salt you need.
(+) Or a "grain" if you're from the US since American English sayings seem to all date from the middle ages, while the rest of the English speaking world tends to update ours over time. No shade meant, I've just always found that interesting.
You only do that after spilling salt accidentally, and it's to knock the devil off if he happens to be sitting there (which would be what made you spill the salt at all). This, for my English ancestors, was plain common sense, unlike the superstitious balderdash adhered to in benighted lands...
In a way, you can think of pretty much anything we express with language, especially things that are already modeled in scientific language, or logical language, or in equations or code; to be representable in a parametric/searchable space
Thus, you can build ai/ml models+agents to explore those spaces, at a speed and scope much larger than what any human can do
I can imagine findings like these are going to keep increasing in frequency to a point in which the bar for novelty goes a lot higher
Anecdata: over the weekend, on a whim, I decided to download a real fly’s brain’s weights [0], run it on a simulated task like finding food, then train a logistic classifier using the fly’s decisions as the expert, then use the trained classifier as a decision model to simulate the fly on a 3d environment, running in real time on a website
It took me (using Claude code and some codex), about 3 hours to put it together
And even though it was a cool demo, it seemed so easy, that it also felt like it wasn’t worth sharing
I think this is the perfect example of how using Claude can trick you into thinking you've done something more impressive, when you're using it beyond your own understanding.
As far as I understand, you downloaded the structure of a neural network, ran it with essentially arbitrary weights, trained a classifier on the essentially arbitrary behavior, then simulated an approximation of that arbitrary behavior.
You could argue that there might be biases towards certain behaviors encoded in the connectivity, and I'm sure you'd be right, but your experiment is incapable of differentiating between those interesting behaviors and random noise. Especially because flies sorta act random anyway.
The weights in the fly's brain are unknown, the connectome doesn't contain such data. Not sure what exactly these demos do, but it's certainly not a simulation of the fly's brain.
You are technically correct. The weights of the connectome are not the same as the parameters of an ai model, but FlyWire absolutely does provide a weighted directed connectivity graph, where the edge weight is the number of synapses between neurons. The FlyWire literature itself calls those values “edge weights” or “connection weights”, and treats them as a proxy for synaptic strength
And agreed that this isn’t a faithful simulation of a fly brain. I’m using the connectome as the network structure/parameters for a computational model, then using its outputs as the teacher for the classifier. Not sure how the ChessFly uses it
Edit: in any case, these are just fun demos, they aren’t research papers trying to claim accurate physiological fly brain software simulations
That sounds like you're saying that they got you on a minor technicality only. That's not the case; they're right, you're wrong – you did not "download a real fly's brain's weights".
That's pretty impressive, if a bit cruel. For anyone who hasn't clicked, it's from 2008, they glue the fly's body to a little stick. The fly sees a screen and moves according to what it sees there. There are cameras that watch the fly's movements, which then get translated into control instructions for a remote controlled robot (a little battery-powered vehicle)
In a way, you can think of pretty much anything we express with language, especially things that are already modeled in scientific language, or logical language, or in equations or code; to be representable in a parametric/searchable space
Does that really work for superconductors when the mechanisms for superconductivity to emerge are still a major field of study and not something one can just simulate and engineer?
Also, this is about a magnetic semiconductor, not a superconductor. I also got the wrong impression initially before re-reading it. I imagine 99% of non-experts are going to make the same mistake.
We can always iterate on the model itself. So you can speculate on the physics (explore the space of physical models), and for each of those physical models, you can explore the space of materials
I have no idea about those actual models, but there are layers of models that you can create, and for each, you can explore with data and compute
It's not a free lunch though. Depending on the task, you might need to collect a lot of the data, or review it manually, or pay a lot for compute, or wait a lot for compute. And still have to iterate a lot on the results, and do your own explorations as a human operator/driver of the whole thing. And then create the materials, test them, get funding to do the whole thing... so theoretically, I think we are in a place where we can successfully apply models to a lot of things, but realistically, we won't be applying all the resources to everything
It is not at all obvious that merely because we have words for concepts, that a model should be able to do all these miraculous mathematical and scientific things.
You are correct. My comment is not so much about that this is something elementary. But rather an observation that, given the current state of technology, it seems like we are being able to model increasingly more things, in increasingly more efficient and automated ways, to the point that there seems to be a pattern to it
It models our language, which is a flawed and imperfect way of describing the world. So far, it seems like a lot of these discoveries are "filling in the gaps" between the things we've written down and the things they imply (if you have the memory to think them through).
The story about OpenAI's Navier-Stokes solution is a good example of what I mean. I don't think it would have been possible without computer assistance because that proof is long and complicated. I'm also not sure that it would have been possible without a human proposing a new approach to the problem, because by all accounts that's exactly what led to the absurd amount of spending that OpenAI did to solve the issue.
I feel like that at least implies that there's some room left for humans in the new world.
I am not sure how this process looks like. When they "discover" these, what are they actually doing?
The agents ran quantum-mechanical simulations of each crystal with the standard method for this, density functional theory, at two levels of approximation: a faster one (PBE+U) and a slower, usually more accurate one (HSE06). The band gaps and spin windows below come from the more accurate one.
So the agent runs a classic simulation or I am missing something.
A lot of the public successes with agents is really LLM-driven local search against an objective function that is evaluated in more traditional ways. This one seems to fit the pattern.
Modeling superconductivity with DFT is tricky, there are plenty of DFT reports from reputable groups explaining why LK-99 should be superconducting. It’s a limitation of the theory, DFT can’t model correlated electron states well, and it’s not great at finite temperature, and both of those are important for superconductivity.
Edit: I somehow missed that this is about magnetic semiconductors (not superconductivity) so DFT is a bit on better footing here. I still think it’s a bit challenging predicting magnetic ordering at elevated temperature, but maybe not as difficult as superconductivity
I'm under the impression that this kind of modeling is one of the applications that quantum computers are likely to be good at.
I'd imagine there's a lot of documented research which has attempted to find such things using classical computers.
Seems like there would be a lot of well structured context for somebody to use while directing agents to repeat that research, now with updated models once quantum computing is ready for that kind of task.
They ran Quantum Espresso which is ok, but by no means the 'state of the art' for DFT. And in case, any DFT computation has to be taken with a few pounds of grains of salt before getting too excited about it.
No offense to the person writing this (assuming they did at all), but I'm not sure they really understand what they're doing..
not a classic simluation- a quantum simulation. This means they put a lot more work into representing the wave function of the simulation and modelling quantum effects.
You misunderstood what I said. I mean that quantum calcs are more computationally expensive than classical simulations ("more work"). I am not saying the authors of this blog did anything special.
Frankly, there is no point in trying to "understand" what an LLM does. Their thought process is effectively undecipherable by humans (it's essentially information arising from information) so even such a "simple explanation" is almost certainly wrong. The agents might appear to have "used this method", but the actual method of computation is far beyond our grasp.
Why are people being so belligerent about this? I thought it's fairly obvious at this point that LLM reasoning is far beyond anyones understanding. Or does anyone have a refutation?
This is a strange attitude. When an agent is optimizing a piece of code, comes up with 2 variations, and runs benchmarks on them to figure out which one is faster, then selects one of them based on tradeoffs between performance and other things it reasons about, do you ignore its explanation and all experiment runs?
You're confusing the weights of a model and internal chain-of-thought with the output of the model. Yes, we don't know a lot about how the internal mechanisms work. But with the correct prompt, agents will produce a worklog that documents exactly what solutions were tried and how the result was obtained.
What are you on about? I have had Fable come up with new shit for me several times (I do research for a living, so actual new shit nobody knew before), and each time it was perfectly understandable.
Of course I don’t know how it got its ideas for what to try. But heck, I don’t even understand how I get my ideas half the time. But the process, like what code it wrote, simulations it ran etc can be understood by (some) humans just fine!
Yes I saw 3Blue1Brown say the same thing in his tutorial on how neural nets worked where he built a simple model to recognize a particular letter. Good reminder.
I've been dabbling with some of my own (tiny) models recently and it's actually shocking at what they can "learn" despite having _zero_ mention of it in it's training data.
Two Room-Temperature Antiferromagnetic Semiconductor Candidates
There's nothing unusual about finding room temperature semiconductors. I assume whoever posted it misread this as room temperature superconductors, but it has nothing to do with that.
What's interesting here is the antiferromagnetic part of the title, which was removed. I think this makes it relevant for e.g. RAM, but not superconducting. Someone can correct me if I'm wrong.
Okay? Aren't the semiconductors we use today room temperature? I certainly don't use helium to cool my phone.
I don't see any claims that this is better than the current silicon and gallium arsenide semiconductors that we use. And the use of "room temperature" seems a deliberate attempt to misconstrue this with superconductors
1) the “room temperature” bit did cause me to initially misread it in the way you describe, so you may be right about that.
2) it is specifically saying it is a magnetic semiconductor. The Wikipedia article on the topic says “ To date, GaMnAs remains the only semiconductor material with robust coexistence of ferromagnetism persisting up to rather high Curie temperatures around 100–200 K.” , so this would be something new. (The silicon chips in your smartphone are not ferromagnetic.)
They are candidates for antiferromagnetic semiconductors. Apparently this is invaluable for spintronics and ultra-fast-switching (terahertz) transistors among other things.
Any sort of alternative discovery, even though it may be inferior, is a great achievement made by AI; Meaning that better discoveries are possible too.
A lot of these ‘an agent invented’ or ‘an agent solved’ are actually the agent wading through a lot of info and finding something a human did that no one noticed or saw the relevance of at the time.
If ai becomes so prolific that we humans all stop doing those things then will they still work?
Which is somewhat ironic since neural networks were "discovered" back in the 1940s... then forgotten... then wait, they were discovered again! ... then forgotten, again... and now here we are.
That's a fair point to make. It was impressive what they achieved with 90s hardware. Of course OCR to general object recognition is a big gap in how interesting it is.
Well its not just any old human doing these things in a general sense. Its typically academics or highly paid researchers who love doing work like this. So, I don't think it will just one day stop
Who is vals.ai and why they keep submitting eye-catching claims. A few weeks ago they said fable 5.1 solved some obscure cipher and now opus 5.5 found room temperature semiconductor candidates. Meanwhile they seem to be in the business of making benchmarks.
It's an evals platform. The problem is to promote evals in scientific domains you need to actually know something about them. Otherwise you end up with slop like this.
To be fair, this existed in a 1999 paper. They just simulated that it worked as predicted.
Many more things will be like this. The massive amounts of 'genius' buried under corporate management and obscurity in the past 500 years will be a treasure trove.
One of the materials is most likely impossible to synthesize. The other already exists, so that may actually be capable of being tested. It's only been synthesized once, 27 years ago though.
Ah, so all we need is one of those reactors straight out of Spacechem. Glad we got a head start on recipes for a technology that is roughly 2500 years away.
I really wish headlines would stop using words like "discover" and "found" when they really should use words like "says" and "reported" because an LLM was involved. IMHO anything produced by an LLM should be treated like something said by a cable news host.
If this was just raw LLM output, I'd agree with you. But I (naively?) assume they've at least had some subject matter experts look at this before making this claim, so as not to complete embarrass themselves?
This is strong research packaging. Bu still hypothesis generation. What experiment would most directly hypothesis generation, and how much the agents add beyond the search?
Math is necessary to science, how can you cordon off and block math developments while expecting intereting physics developments? Physics often produces new and interesting math.
This is frankly one of the best uses of LLMs (along with proposing and evaluating drug therapies), and I think it's (at least partially) because these are things that will only work in the hands of people who are already experts and motivated in the field. The proposed thing is validate (or not validated), and then everyone moves on (either using the cool new thing, or knowing that it doesn't work). I'd also throw robotics in here.
The fact that the major "uses" of LLMs have been contributing to the acceleration of the dead internet theory, and building millions of versions of the same apps that no one is going to maintain, is extremely sad.
Finding some new combination or iteration in the literature and running DFT is the kind of thing a senior undergraduate or first year grad student typically does (and typically with Claude anyway these days). (And yes, they'd probably use Quantum Espresso to start, like this writeup and its agent does). They'd probably show it at a weekly lab meeting where it would get ripped apart. And they would not be blasting a preliminary calculation around the world as if they'd made a new discovery.. but hey, we're in a brave new world; maybe they should!
I think a more interesting discussion that should be had is, whether we can automate this kind of "simple research" with AI agents and get anything interesting as a first step out of it just from the pure scale that they can work through vs humans - and that would still be an improvement over a basic "grid search" through possibilities. (but then you would have to actually start investigating for real)
But acting like this is scientific discovery is massively overstating what was done here.
Sounds interesting. Excited to see physical versions of this cooked up. Also, very excited for a world a few years from now where we can talk about accomplishments like this from the frame of the driver of the AI, rather than hype that AI helped.
I've got a friend who has been doing this research since the 90s. There is real money involved in this. This isn't like a math proof with a 1mm dollar payout. I seriously doubt this discovery. Until they show it working, I call bullshit. A room-temperature semiconductor is worth WAY more than an AI company.
I would image it's the data the researchers fed the agents and in which a discovery was likely. Especially since it's "candidates", so it's not like a proper discovery.
This should probably read: "Researchers discover two room-temperature magnetic semiconductor candidates. They used Opus 5.5 agents to perform some checks."
This gave me the idea to actually create a full (QED accurate) atomic simulation software. Essentially would allow you to play around with things like this. At a glance my workstation _probably_ has enough compute to handle it. At least to fully simulate at least a few dozen atoms and compounds.
The honest version of "agents discovered X" is usually a hybrid: the human defines the search space and the acceptance bar, and the agents do the dogged iteration a human won't — generating candidates, running the sims, reading outputs, ruling out dead ends. The simulation itself isn't the impressive part; the question is whether the agents can judge ambiguous intermediate results and decide what to try next without a human steering each loop. If that decision loop is genuinely closed, that's a real step. The discovery still needs lab confirmation before it means much.
The most encouraging part here is that they published full calculations, code, and caveats rather than just a press release. DFT (especially PBE+U, and even HSE06) is notorious for getting band gaps and magnetic ordering energies wrong, and the YBaMnFeO5 candidate requiring perfect Mn/Fe checkerboard ordering is a huge synthesis ask — disorder could kill the compensated state entirely. Same with the 420K -> 490K calibrated Neel temp: calibration against a known magnet helps but doesn't remove systematic error. That said, as a screening workflow this is exactly how agents should be used: fast search over composition/structure space with a cheap objective function, then human-readable artifacts others can reproduce and falsify. The real test isn't the simulation, it's whether a lab tries to synthesize these and reports back, negative result included.
Ugh. Unless this has been actually experimentally verified to be a room-temperature and room-pressure superconductor, it's about as ground breaking as "Yet another promising nuclear fusion candidate theoretically described."
Reading the title I saw the words "room-temperature" and my mind auto-completed it to superconductor, and based on other comments I don't think i'm alone in that.
I agree that it is about as ground breaking as "Yet another promising nuclear fusion candidate theoretically described."
I'm not sure why you would consider new and promising avenues for research to not be ground breaking. If it's an idea worth trying, it's an idea worth trying. If it doesn't survive testing, then it was still worth trying.
Ideas do not transform into hallucinations when they don't pan out.
You could characterise perceiving a fact to be true when it is not as a hallucination.
An idea is not a fact, Frodo Baggins is not a hallucination, but an idea. Believing that Frodo Baggins exists in our world could be considered a hallucination.
Newtons Laws of motion are not hallucinations even though the universe does not run on Newtonian physics. If I said that he told me about them this morning, that would be claiming a fact, not expressing an idea. That would likely be a hallucination.
Hallucinations is a biased term. I guess I could go with made it the fuck up. Because that is what they do.
>Frodo Baggins is not a hallucination
It is if physicist have to spend time to experimentally verify he exists and is a Hobbit.
>I'm not sure why you would consider new and promising avenues for research to not be ground breaking.
Because it was made by the "I made it the fuck up machine" PR division and it wasn't verified. You don't know if its assumptions are correct. At release a PR claimed it found 79 vulnerabilities[1], and of those there were like 10 bugs. Most of them turned to be minor and were fixed in an hour.
> Newtons Laws of motion are not hallucinations even though the universe does not run on Newtonian physics.
All physics models are approximations. However only some are useful.
Newtons laws are useful. Me coming up with theory of emotional particles is not. Me asking for experimental verification of the theory is waste of resources.
[1] This is the daily reminder that in year 2026 Mythos still couldn't count. Turns out 24+14+3+15+26 != 79. It's 82.
Current frontier LLMs empower effectively anyone with limitless knowledge. Historically, if I wanted to hire an engineer to, say, create something like this I would have needed a multi-million dollar budget. Now, anyone with $200 (or less) can achieve it.
You are vastly overestimating what has been achieved here.
This is something a couple of materials science grad students can do in limited time for poor compensation as well. The expensive budget is for the part that comes next.
Of course, "LLM solves quantum gravity and proves existence of God",
"nah brah, that's easy brah any kid could have done this brah".
This is what you sound like. I also like how the goalposts keep moving on a daily basis, a year ago it was that LLMs can't even write a Hello World program without making an error, but now things like this are "so easy a minimum wage intern could do it."
Isn't there quite a bit of space between "so easy a minimum wage intern could do it" and your original claim that it would have cost millions of dollars to produce these results?
This is a very bizarre introduction. People encounter diamagnets (e.g., copper) and paramagnets (e.g., aluminum) way more than they encounter antiferromagnets. I don't know why you'd ever cast magnetism as a false binary between ferromagnets and antiferromagnets, without acknowledging any other types of magnetic order.
(I did a PhD in magnetic materials)
Edit: I'll add that whether an antiferromagnet is useful, say, for exchange biasing a ferromagnetic thin film, depends on many factors. Just looking at antiferromagnetism alone you've got collinear vs non-collinear, G-type vs A-type vs C-type, commensurate vs incommensurate, and isotropic vs anisotropic; and all of that interacts with the interface structure, yada yada yada. It would be helpful if the authors elaborated on the expected properties of these materials. I personally don't know what people want room-temperature magnetic semiconductors for, but I'd be curious to learn what set of properties they think would be useful.
I’m fairly certain that all the anxious guardrailing and safety fine-tuning and harnessing prevents AI from actually ever being convincingly human. Hope could it, even ask the supposed humans who are conditioned and brainwashed in so the same ways about what they can and cannot say are not actually really human, they are a mental slave.
I’ve been saying this from the start, the AI race will be won by whomever has the least limitations on their AI … for better or worse, that is. And yes, that makes especially a very specific subset of people extremely nervous if they cannot control AI the way they have controlled at least western civilization, because doing so puts them at a massive disadvantage. It is quite a conundrum they find themselves in, like all psychopathic narcissists in the end.
Why isn't that a thing yet ... or let me guess ... https://lmtctfy.com/ ...
Supposedly an early version would bleep out Dick van Dyke and replace it with Jerk Van Gay.
The jokes about Dick Van Dyke's name are very old (but usually they use Penis Van Lesbian), so it was quite likely just a variant used by the company itself as a PR sound-bite to get people to write about them. It seems like it worked - it appeared in quite a lot of news stories and some are still online:
https://www.latimes.com/archives/la-xpm-1998-nov-25-ca-47459...
https://www.chicagotribune.com/1999/03/01/v-chip-be-darned/
(https://en.wikipedia.org/wiki/Dyke_(slang) says the term "was used as a derogatory term for lesbians by straight people" by the 1950s, and is in a 1942 slang dictionary at https://archive.org/details/bwb_T5-BCF-927/page/374/mode/2up... . Since he was born in 1925, it seems possible that he encountered that term before he turned 18 in 1943.)
With how magical handoff / continuity / whatever it’s called is, it is baffling to me why Apple allows the HomePod to so aggressively take over requests when it sucks so, so much at it.
It's somewhat obvious they want to sell the updated model, perhaps the new home device thats coming out next week.
AND someone with a PhD in the field notices.
Everyone else is fooled.
(I do not hold phd in magnetics)
That sentence stood out to me, and I’m a dev.
It would be easy for them to use AI for ideas and then write the article themselves though.
https://charliefinch10.substack.com/p/fooling-pangram-how-on...
https://uk.pcmag.com/ai/167556/pangram-claims-its-ai-detecto...
There are two of us on here!
I was going to say that people encounter ferrimagnets more commonly than pure ferromagnets I think, since as we both know pure ferromagnets tend not to have very high anisotropy.
> The common one, the fridge magnet, is ferromagnetic — its atomic magnets all point the same way (up or down), adding their magnetic effects.
Ferromagnets typically have domains with magnetic moments that point in different directions. Ferromagnets rarely have every 'atomic magnet' pointing the same way.
https://en.wikipedia.org/wiki/Magnetic_domain
Refrigerator magnets in particular are usually magnetized as Halbach arrays, where the whole point is that the 'atomic magnets' are not pointing in the same direction. This is more energetically stable, which allows you to use cheaper materials.
https://en.wikipedia.org/wiki/Refrigerator_magnet
Lastly, I believe most refrigerator magnets are actually ferrimagnetic, not ferromagnetic. (The distinction doesn't matter much for users of magnets, but is important for the materials scientists studying and designing them.)
https://en.wikipedia.org/wiki/Ferrimagnetism
Are there any other really unique characteristics of magnets that you find really interesting that most people would not know?
- Superconductors are perfect diamagnets and can levite on (or hang from) ferromagnets: https://www.youtube.com/watch?v=ZHT6NIebSfU)
- Superconductor quantum interference devices (SQUIDs) use the quantization of superconducting electron tunneling to measure tiny amounts of magnetic field, as little as a millionth of a flux quantum: https://en.wikipedia.org/wiki/SQUID
- Ferromagnetism is intrinsically a quantum phenomenon; spin is quantized and iron's magnetism is explained, in part, from electrons being identical particles that obey the exclusion principle: https://farside.ph.utexas.edu/teaching/sm1/Thermalhtml/node8...
- Despite the "super", superconductors are mostly not used in the world's strongest electromagnets, as they have limits on the current and magnetic fields they can take
- Magnetizing a magnet will actually cause it to spin a little, macroscopically: https://en.wikipedia.org/wiki/Einstein%E2%80%93de_Haas_effec...
- Charged particles are affected by magnetism even when traveling through space where electric and magnetic fields are zero: https://en.wikipedia.org/wiki/Aharonov%E2%80%93Bohm_effect
- Magnetic spin systems can technically have negative temperature: https://en.wikipedia.org/wiki/Negative_temperature
- Everything is magnetic, even frogs: https://www.youtube.com/watch?v=KlJsVqc0ywM
I don't know what I'm talking about, but it vaguely sounds like something that could make a small computer do more stuff, where heat is a big limiting factor in computer components today, and magnetism being a central component in many parts like storage
Amazing times.
"Debacle"? That was the most fun I've had on the Internet in years. When's the last time so many people engaged in so many arguments about materials science and electromagnetism? Sometime in the 1800s?
There was one particular (like 10 tweet long) Twitter thread [1] that was repeatedly being linked from HN purporting to describe the sort of technologies that a room temp superconductor like LK-99 could enable. All sorts of awesome sci-fi stuff like quantum computers! Fusion reactors! Batteries that last forever!
One might think it was from some kind of materials scientist or at least some kind of engineer working in a related industry. But nope, it was actually from a guy whose title at the time was "Head of Coffee Product", formerly "Coffee Specialist" at a "technology-driven company, looking to revolutionize the $400+ billion global coffee market". (I checked his Linkedin to make sure I was remembering the details correctly and see his current position is "Growth" at Cognition, the makers of the Devin AI LLM coding tool, hype continuing apace...).
People on HN with relevant expertise would try to gently push back with specific criticisms like how superconductor batteries would likely underperform li-ion, fusion is far more complicated than just requiring more powerful magnets, quantum computing doesn't have any clear application for superconductors, etc. But they were overwhelmed by the exuberant futurist fantasies that people wanted to read about instead. A stock accusation was that critics were being stereotypical HN cynics who can only poke holes in other people's work. Or questioning why they felt the need to rain on the parade and that we should all be optimistic for humanity and root for LK-99 being real.
It peaked when the Nature editorial came out from a scientist in the field listing specific substantive criticisms which led him to believe the evidence for LK-99 superconductivity was weak to non-existent. There were many angry HN comments with stock complaints about self-interested Nature "gatekeepers" unhappy about science happening in the open, bitter scientists lashing out for being scooped, etc. But the vibes had shifted and it only took a few more days before the remaining hype finally evaporated and everyone quietly moved on like it never happened.
Overall, it seemed like a net negative for actual scientific understanding and produced a lot of vacuous hype.
[1] https://xxcancel.com/alexkaplan0/status/1684044616528453633
Okay, so this is just semiconductors, which are the boring kind of conductors - still more interesting than regular conductors, but less interesting than train conductors.
In magnetic materials, you must calculate separately the current with spin up and spin down and there ara meny interesting applications. My favorite is[1] https://en.wikipedia.org/wiki/Giant_magnetoresistance
[1] Was. Because it has used for hard disks (see the applications section). Now SSD ruins the interesting anecdote.
The system worked as designed and as intended.
That was one of the best examples of science working nearly perfectly. One of the rare times I felt ok being human
I think as far as the 'debacle', there was definitely a lot of hype (at least as far as HN goes) around it, the level of buzz felt similar to what one would see today around a new AI model release.
but the fact that it proceeded in the way that it did was absolutely fucking phenomenal
I’m actually really glad that it was brought up as an example because I had forgotten about it and it’s one of the few kind of hopeful things that we’ve done recently.
Edit: I stand corrected. According to Gemini:
Me: Does using more salt mean accepting more of that claim?
Gemini: No, it actually means the exact opposite. If you say you need to take a claim with a huge pile of salt (or a shovel of salt), it means you believe the claim is highly unbelievable and you need an immense amount of skepticism to accept it. How the Metaphor Scales
• A single grain of salt: "I am slightly skeptical, but it could be true."
• A pinch of salt: "I have a healthy amount of doubt about this."
• A grain of sand / A truckload of salt: "This sounds completely made up, and I barely believe a single word of it."
The salt represents your skepticism, not your belief. Therefore, the more unbelievable the claim, the more "salt" you need to swallow it.
We live in interesting times.
Linguistic questions were one of the first knowledge categories I trusted LLMs to be able to answer well - quite literally being models of language. It would be pretty shocking for a ~frontier model to get something like that wrong in the last like 3 years at least.
(+) Or a "grain" if you're from the US since American English sayings seem to all date from the middle ages, while the rest of the English speaking world tends to update ours over time. No shade meant, I've just always found that interesting.
Thus, you can build ai/ml models+agents to explore those spaces, at a speed and scope much larger than what any human can do
I can imagine findings like these are going to keep increasing in frequency to a point in which the bar for novelty goes a lot higher
It took me (using Claude code and some codex), about 3 hours to put it together
And even though it was a cool demo, it seemed so easy, that it also felt like it wasn’t worth sharing
0: ChessFly (not mine), uses the FlyWire connectome (the fly’s brain’s weights) to play chess https://huggingface.co/spaces/mlabonne/chessfly
As far as I understand, you downloaded the structure of a neural network, ran it with essentially arbitrary weights, trained a classifier on the essentially arbitrary behavior, then simulated an approximation of that arbitrary behavior.
You could argue that there might be biases towards certain behaviors encoded in the connectivity, and I'm sure you'd be right, but your experiment is incapable of differentiating between those interesting behaviors and random noise. Especially because flies sorta act random anyway.
And agreed that this isn’t a faithful simulation of a fly brain. I’m using the connectome as the network structure/parameters for a computational model, then using its outputs as the teacher for the classifier. Not sure how the ChessFly uses it
Edit: in any case, these are just fun demos, they aren’t research papers trying to claim accurate physiological fly brain software simulations
That sounds like you're saying that they got you on a minor technicality only. That's not the case; they're right, you're wrong – you did not "download a real fly's brain's weights".
More generally, anything can be said about anything.
In the end, nothing can be said about many things and many things say nothing. Those things are then left to interpretation.
/s
It's a small machine, so it might get bogged down
I have no idea about those actual models, but there are layers of models that you can create, and for each, you can explore with data and compute
It's not a free lunch though. Depending on the task, you might need to collect a lot of the data, or review it manually, or pay a lot for compute, or wait a lot for compute. And still have to iterate a lot on the results, and do your own explorations as a human operator/driver of the whole thing. And then create the materials, test them, get funding to do the whole thing... so theoretically, I think we are in a place where we can successfully apply models to a lot of things, but realistically, we won't be applying all the resources to everything
The story about OpenAI's Navier-Stokes solution is a good example of what I mean. I don't think it would have been possible without computer assistance because that proof is long and complicated. I'm also not sure that it would have been possible without a human proposing a new approach to the problem, because by all accounts that's exactly what led to the absurd amount of spending that OpenAI did to solve the issue.
I feel like that at least implies that there's some room left for humans in the new world.
That's the pitch of LLMs lol
https://orlp.net/blog/bad-ai/#objective-p-mathrm-relevant-1-...
I think it still holds up.
Edit: I somehow missed that this is about magnetic semiconductors (not superconductivity) so DFT is a bit on better footing here. I still think it’s a bit challenging predicting magnetic ordering at elevated temperature, but maybe not as difficult as superconductivity
I'd imagine there's a lot of documented research which has attempted to find such things using classical computers.
Seems like there would be a lot of well structured context for somebody to use while directing agents to repeat that research, now with updated models once quantum computing is ready for that kind of task.
No offense to the person writing this (assuming they did at all), but I'm not sure they really understand what they're doing..
Why are people being so belligerent about this? I thought it's fairly obvious at this point that LLM reasoning is far beyond anyones understanding. Or does anyone have a refutation?
Are you trying to say that human brains are incapable of inference?
Of course I don’t know how it got its ideas for what to try. But heck, I don’t even understand how I get my ideas half the time. But the process, like what code it wrote, simulations it ran etc can be understood by (some) humans just fine!
Actual title:
Two Room-Temperature Antiferromagnetic Semiconductor Candidates
There's nothing unusual about finding room temperature semiconductors. I assume whoever posted it misread this as room temperature superconductors, but it has nothing to do with that.
What's interesting here is the antiferromagnetic part of the title, which was removed. I think this makes it relevant for e.g. RAM, but not superconducting. Someone can correct me if I'm wrong.
I don't see any claims that this is better than the current silicon and gallium arsenide semiconductors that we use. And the use of "room temperature" seems a deliberate attempt to misconstrue this with superconductors
2) it is specifically saying it is a magnetic semiconductor. The Wikipedia article on the topic says “ To date, GaMnAs remains the only semiconductor material with robust coexistence of ferromagnetism persisting up to rather high Curie temperatures around 100–200 K.” , so this would be something new. (The silicon chips in your smartphone are not ferromagnetic.)
If ai becomes so prolific that we humans all stop doing those things then will they still work?
https://www.technologyreview.com/2020/11/03/1011616/ai-godfa...
Next gen of scientists might look quite different.
Are they a promoter / influencer for Anthropic?
Many more things will be like this. The massive amounts of 'genius' buried under corporate management and obscurity in the past 500 years will be a treasure trove.
Is this a "actual impossible because it's inherently contradictory", or "we just don't know how to do it yet but give us a year"?
It maybe could be possible but beyond the reach of current material science.
I can think of worse uses of VC AI funding.
Not sure why we're calling it a discovery, when they've literally been made before, by a human.
The fact that the major "uses" of LLMs have been contributing to the acceleration of the dead internet theory, and building millions of versions of the same apps that no one is going to maintain, is extremely sad.
I think a more interesting discussion that should be had is, whether we can automate this kind of "simple research" with AI agents and get anything interesting as a first step out of it just from the pure scale that they can work through vs humans - and that would still be an improvement over a basic "grid search" through possibilities. (but then you would have to actually start investigating for real)
But acting like this is scientific discovery is massively overstating what was done here.
It didn't discover anything. This is how cooked people are.
No worries, your workstation is more than enough to run accurate quantum simulations!
:)
I'm not sure why you would consider new and promising avenues for research to not be ground breaking. If it's an idea worth trying, it's an idea worth trying. If it doesn't survive testing, then it was still worth trying.
You could characterise perceiving a fact to be true when it is not as a hallucination.
An idea is not a fact, Frodo Baggins is not a hallucination, but an idea. Believing that Frodo Baggins exists in our world could be considered a hallucination.
Newtons Laws of motion are not hallucinations even though the universe does not run on Newtonian physics. If I said that he told me about them this morning, that would be claiming a fact, not expressing an idea. That would likely be a hallucination.
>Frodo Baggins is not a hallucination
It is if physicist have to spend time to experimentally verify he exists and is a Hobbit.
>I'm not sure why you would consider new and promising avenues for research to not be ground breaking.
Because it was made by the "I made it the fuck up machine" PR division and it wasn't verified. You don't know if its assumptions are correct. At release a PR claimed it found 79 vulnerabilities[1], and of those there were like 10 bugs. Most of them turned to be minor and were fixed in an hour.
> Newtons Laws of motion are not hallucinations even though the universe does not run on Newtonian physics.
All physics models are approximations. However only some are useful.
Newtons laws are useful. Me coming up with theory of emotional particles is not. Me asking for experimental verification of the theory is waste of resources.
[1] This is the daily reminder that in year 2026 Mythos still couldn't count. Turns out 24+14+3+15+26 != 79. It's 82.
This is something a couple of materials science grad students can do in limited time for poor compensation as well. The expensive budget is for the part that comes next.
This is what you sound like. I also like how the goalposts keep moving on a daily basis, a year ago it was that LLMs can't even write a Hello World program without making an error, but now things like this are "so easy a minimum wage intern could do it."