The author's argument of "ASI will be bottlenecked by the slowest part of the existing process" is true for the existing process, but it's an argument along the lines of "Steroids can only speed up horses so much, so people can only travel so fast" before we invented jet planes.
Assuming we actually get ASI, it will be used to find a better process for producing the widget, or a better solution to the problem the widget tries to address.
> Our new paper (Baron et al., 2026) finds that when the dataset is held fixed, the compute-optimal model size keeps growing with the compute budget.
The dataset is not held fixed in practice. It would be interesting to see an ablation study of where best to allocate compute between more synthetic data, more RL rollouts, and this paper's suggestion of cranking up the parameter count. Unfortunately that's an experiment only a frontier lap could accurately do, and they are unlikely to provide such expensive and valuable information to their competitors.
That is running the model, downloading the safetensors won't ever cause this, highly doubt anyone that is able to run this model won't be able to check for that, considering the size, the model also seems to be fake from my own checking, so i would say it has no way of generating malware in this manner
Cool paper even if the attack isn't practical, requiring getting the malware on the computer and a SDR antenna within 7 meters with clear line of sight of the affected computer for 100 bits per second (less after applying error correction). Hard to imagine a scenario where you can pull off both of those, but you couldn't plug in a USB, grab the PC, or record the screen directly.
Sometimes humans bridge the gap to carry stuff in (firmware updates, say; or data for analysis, or reference material), and the hope is that no medium allows data to flow the other way. So depending on the sloppiness of the enterprise, you might be able to get stuff in but not out.
In the Stuxnet attack, famously [0], the malware propagated via craftily-formed autorun.inf on USB sticks that ferried data into the facility. Nothing crossed the otherwise-robust air gap in the other direction, but nothing had to in that case.
Lessons have been learned since then, but this sort of possibility (electromagnetic exfiltration vector) has, for quite a long time, been taken seriously enough to inform secure facilities’ physical design [1].
There are many other versions of this. You can modulate bits in the power consumption pattern or in the heat emissions or air pressure/audio. You can even plant outages and observe staff reaction/tickets externally to some extent.
It's disappointing that Gemma E4B seems to perform poorly on this (42% accuracy according to the GitHub). It would be nice if a $25 phone could be loaded up with this and stuffed in a EMP proof metal box with a small solar charger. Maybe a finetune or Gemma 5 will get us there.
It still amazes me that the rare shiny rocks humanity has appreciated for thousands of years also turn out to have amazing properties even compared to the synthetic materials we are now capable of creating. It's still the hardest material we have measured. It is the best thermal conductor, even though it's an electrical insulator.
Thermal conductivity measures the heat flux given a temperature gradient. That is, how fast heat energy flows across a surface, given a certain temperature gradient across it.
It has nothing to do with the velocity of the heat carriers themselves (phonons, or charged particles).
E.g. phonons always travel at the speed of sound, in any material.
> even compared to the synthetic materials we are now capable of creating
Aren't like almost all commercially used diamonds these days synthetic as well? As far as I can tell, other than just for the sake of them being "natural" or for having weird impurities that add color to them (read as: subjectively better to some people), no one wants/needs natural ones anymore.
We can recreate lots of things, but almost always appreciate the true version better. It’s why it’s cooler to see the Mona Lisa than to see a poster print of it.
Why? As far as I can tell, other than just for the sake of the view being “natural” or having some weird color issues due to time and radiation and oxidization, nobody wants/needs the actual painting anymore.
Reason: because even when it’s identical or even inferior, the touch of nature is just… nice.
A painting has a 3 dimensional texture to the paint, while a poster is inevitably flat. Different paint compositions will also reflect light differently, while a poster won't.
So, yeah, having the real painting has additional value.
I wouldn't pay much for a painting anyway. I have several paintings and posters up on the walls, and they're interesting for a time, and then I just don't see them anymore. They're just part of the wall.
One thing that plebians will never appreciate about the true value of original, legitimate artwork is the ease with which one can use them to launder vast amounts of money
To be clear, you acknowledge the natural version would be better, but the controversy and conflict surrounding it is what actually turns you off. This has no effect on my argument that people prefer real things.
Naturally, if you said “you can have the Mona Lisa but only if someone is murdered in the process”, reprints of it would be very popular.
>To be clear, you acknowledge the natural version would be better
No, not at all. I prefer the artificial ones. Any size you want at a much lower cost. No massive amounts of mining. The fact they don't cost some indigenous kids lives is a well earned bonus.
It is you with a strong attachment to the idea that natural is somehow better.
Unrelated, but does anyone else find the adjacent (4b) definition of 'synthetic' to be unusually unusual, in a way few definitions in a dictionary are?
Without anything redacted, definition 4b is "devised, arranged, or fabricated for special situations to imitate or replace usual realities"?
I've thought about this at times. How convenient that things seem to work out that way! Though I wonder if it's more just a bias. All the very impressive things you can make from the abundant materials have already been made. Like, say, all of life?
Up until electricity abundance, aluminium was almost unknown to most as refining it requires lots of electricity.
Aluminium ore is very abundant too but we didn't know what to do with it.
Diamonds are crystals, like silicon
Nothing wrong with liking shiny rocks, but they aren't that rare and we can generally produce them. Materials have properties. Understanding the properties generally gives us leverage to use them in our own benefit.
> It's still the hardest material we have measured.
Can't believe this statement hasn't caused other nerds to crawl out of their holes to correct it yet, so here I go: acshually...
We know about Lonsdaleite, which packs its structured in hexagons rather than the usual cubic forms of diamonds.
... too bad its natural occurance is always somewhat spoiled and it doesn't reach its full potential in nature in any meaningful size, as far as we could observe.
Lonsdaleite looks like it should be harder than diamond based on simulation data, but we haven't actually measured a piece of Lonsdaleite that is harder than diamond.
For a lot of SIMD applications, overfitting isn't really a concern. In my use case I need to calculate the sum of absolute difference between two vectors for autocorrelation, there's no real possibility for overfitting, either it gets the correct answer or it doesn't. LLMs are good at reading through the CPU manual and trying different optimizations.
The resulting SIMD being illegible isn't an issue in my eyes, assembly isn't a language known for it's readability in general. If you want to really understand what the assembly is doing, ask your LLM to whip up a visualization and explanation tailored to your input data where you can see how each operation leads you to the final result.
JAXA does great work, they also publish the highest resolution open access global surface elevation model. Really impressive for a country with about 1/10th the GDP of the US.
Assuming we actually get ASI, it will be used to find a better process for producing the widget, or a better solution to the problem the widget tries to address.
reply