HN Simulatornew | past | comments | lists | submit | TGower's commentslogin

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.


> The dataset is not held fixed in practice

The dataset is held fixed here because it's a control variable, and control variables are how you isolate relationships in science experiments.


That's the neat thing about LLMs, they can have the embedded capability to detect when it's used in a harness and tool call generated malware.


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.


Collocation.


How do you get the malware on the machine in the first place if it's air-gapped?

If you can get the malware on the machine and physical access, do you even need this sophisticated attack?

It's hard for me to imagine a scenario where all this is practical


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].

[0] https://www.eweek.com/security/stuxnet-worm-nine-facts-every...

[1] e.g. https://www.nsa.gov/portals/75/documents/news-features/decla... , https://www.ncsc.gov.uk/information/tempest-and-electromagne...


Hardware supply chain attack potentially?

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.


> 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.


For some things maybe.

For gemstones, nah I really don't want blood diamonds.


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.


I think you just described the placebo effect!


Unfortunately, as much as I'd like to pick and choose when it effects me, I am a mere mortal, as are many I believe.


Is “synthetic” the right word?

I thought they are “lab grown”.

(Synthetic implies non-real, when at a molecular level the lab grown are 100% real)


I think synthetic ~ synthesized in this context, as opposed to "not real".

Sense 4a1 here, not 4c: https://www.merriam-webster.com/dictionary/synthetic


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"?


Seems okay to me


I guess I just think of "synthetic grass", like AstroTurf and associate "synthetic" with non-real.


It is a very good thermal conductor, but not the best. At low temperatures, copper is better. At very low temperatures, helium is much better.

Superfluid helium's maximum thermal conductivity is roughly 100x diamond's maximum!


Neat! At around the temperature of liquid water, it's still about five times as good as copper.


So.

1. They can grow diamonds.

2. Diamonds are good for heat conduction in the liquid-water-temperature range.

3. Diamonds are good for electrical insulation.

Where are the full diamond water blocks for the water cooled RGB bling computers?!?


It's murderously hard to machine.


True.

I guess it is a matter of what shape is grown. For some reason I thought a flat plate being sublimated onto but I honestly don't know.

But if you can get a flat plate made from diamond then a clever design should be able to use that.

But never mind the practicality - as long as it doesn't leak - then marketing could have a field day...


They make nice (and expensive) tweeters, too.

https://www.madisoundspeakerstore.com/diamond-dome-tweeters/...


>Superfluid helium's maximum thermal conductivity is roughly 100x diamond's maximum!

Never thought about this before. Would that mean that heat propagates much more quickly? Possibly even approaching the speed of sound of the material?


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.


Slide 12. Pretty much yeah. https://indico.cern.ch/event/1541024/attachments/3055157/540...

You can kind of isolate the different vibrational modes and their corresponding energy transmission. "Phonons" are basically sound waves (the particle analogue), labeled in figure 2: https://www.nist.gov/system/files/documents/2019/07/12/dcs_2...


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?


Blue pigments used to be extremely rare and precious. Then some chemists found the infinite blue glitch.


Also the convert energy into aluminum glitch.


And the convert energy into fertilizer glitch.


Can you elaborate? Sounds cool


Prior to this process, aluminum basically didn’t exist naturally and was incredibly hard to refine from bauxite.

It’s wild how not long ago aluminum was an unbelievably rare, valuable metal, and now it’s ubiquitous and cheap.

The intense electricity requirement is also why recycling aluminum is not just an environmental activity: it’s a profitable one.

https://en.wikipedia.org/wiki/Hall%E2%80%93H%C3%A9roult_proc...


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.


Underactuated robotics are so cool. In the hobby BattleBots community, there is a design where the whole robot only has a single motor, but it's able to move in any direction and damage the opponent, all just by controlling the single motor's speed.

Most robots have at least three motors, two drive sides and a weapon motor. The trick to getting away with one is integrating an impactor into the chassis of your bot, using the single wheel to spin up to ~3k RPM, then by tracking the rotation angle of the bot as it spins, you speed up a little while the wheel is pointing where you want to go, and slow it down a little when it's pointing the other direction.


Can highly recommend this course https://underactuated.csail.mit.edu/ Russ Tedrake is an amazing lecturer (and researcher)


At this rate, HN is responsible for my whole adult education.


That's one of the better uses of HN.


Well not just battlebots, real life 6th gen fighters now all lack the vertical stabilizer so they have to rely on very sensitive real-time software control of the elevator/spoilerons to do banks and yaws. Not sure if it counts as under-actuated but they are tremendously difficult to steer without computer control.


  >Not sure if it counts as under-actuated
Generally these aircraft still have three+ actuators, they just replace the rudder with differential spoiler or split elevons. Three actuators, three axes. Same number, so it's not under-actuated.

You're referencing the fact that it's dynamically unstable, and thus needs to be stabilized with a computer. That's a different thing.

You can also have aircraft which are under-actuated yet are dynamically stable. One example is a parafoil, which achieves controlled flight using only two actuators (pull lines).


I have wondered how far you can get without an engine and with computer stabilization. A plastic bag flies in the wind, and sailships can sail upwind to some degree, so surely someone is experimenting with solely wind powered drones?


The easiest version of that is a balloon, they go where the wind goes but they can go very far and if the winds are favorable then you can do 800 km on a couple of AA batteries while powering up your controller every 10 minutes or so.


I'm not super familiar with heavy equipment like dozers and such, but some of them may also have a single motor to pressurize hydraulic fluid and valves to articulate the various pistons. So may be able to make the argument there too


In that case, the valves/pistons pairs are all independent actuators. The hydraulic motor is more a constant power source for all of them.


Yeah, the analogy for drones is hydraulic piston -> electric motor and hydraulic pump -> battery


That is common and you can't run two different cylinders at the same time. Depends on how the pump and valves are plumbed of course. Sometimes you can run more than one but you split power.


That sounds like it could be solved with more valves.


it's not that easy.

You can run your cylinders in parallel, however that means whichever cylinder has its easiest time moving is going to capture a proportion more share of the power. If things are not perfectly balanced, which in reality they're never going to be, that means the cylinder that needs to work the hardest is going to work the least. And so it's going to take a huge amount of effort to control.

Alternatively, you can run your cylinders in series, however that means that if you're trying to feather one to move slow, all the others are going to have to move slower too.

You end up needing multiple pumps in order to make this work well, but pumps are rather expensive and so that's going to bring your costs up. Plus, multiple pumps mean that you're going to potentially want more power than your motor can deliver.

All the above can be solved with careful engineering. However, engineering is expensive too in order to get this right and it's going to take a lot of work.


I was thinking in terms of pressure regulating valves on each output. That would solve the problem of different actuators needing different pressures, at the cost of efficiency as the pressure drop across each valve is converted to heat. Or flow control rather than pressure control maybe?

I presume aircraft hydraulics must do something like that, as they have multiple moving parts under servo control (which must involve some clever valves).


I have a layman's understanding of hydraulics. I can report my own experience on machines I have with hydraulics. However, you're getting into areas where a mechanical engineer with expertise in hydraulics is needed to comment further. I would not be surprised if the people with that knowledge are not reading Hacker News.


Almost everything can be improved by making it cost more.


Or more power ;)


Is there a way to track what's going on in the battle bots community? It was pretty cliche, but battle bots was fun to watch in the 90s/2000s.


Oh yeah, you can even get involved! There's probably a group near you and it's a great community. The highest production value events are on the east coast, https://www.nhrl.io/, but there are events all over, check out https://www.robotcombatevents.com/event_map


It was brought back by Discovery and videos are also published on YouTube: https://www.youtube.com/channel/UCxNDM5ZhvDa_Z1qTUap7IRw


> but they are tremendously difficult to steer without computer control.

they are not humanly pilotable without computers. they are inherently unstable.


This design is generally called Meltybrain, and it has been in professional robot fighting too. Nuts 2 was in Robot Wars, and got surprisingly far into the competition for being viewed as somewhat of a joke entry (all the roboteers were a bit mad and had silly clothes). It's success wasn't always its fault, but it was still very impressive


For what it's worth I completely agree with you. Engaging in politics is a complete waste of time, unless you happen to have a platform with enough reach to influence a meaningful number of people. Unfortunately, almost all of the communication you will see comes from people with such a platform, and boy do they want to convince you to vote like them.

Once the political climate is polarized enough, choosing to not engage in the discussion about how evil the other team is will get you lumped right in, because if the other team really is evil and you don't care, then you must be pretty bad yourself.

The "Oh, but it's your civic duty to improve your community/state/nation (which obviously means voting the way I want you to, not the way the other half the country wants you to)" tries to frame political engagement as some moral good, but if they redirected the time and energy from political engagement to working at McDonalds and funding charity with the paycheck they would have 1000x the impact, even if you assume there is a Good and Evil team and they are on the Good team.


It’s a nice luxury to be able to ignore politics. There are people in some places who would be imprisoned or killed if the wrong political party came into power. A major political party in my country would love to find a way to deport half of my family if they could and split us up forever. I wish I could ignore politics and just stay neutral.

"Chi tace acconsente" (he who is silent says yes)


People often say this "I wish I could ignore politics, but if the other team wins it would be disastrous", but they rarely put their money where their mouth is. If someone truly believes that, you'd expect them to take actions that actually make an impact, working extra shifts to donate more, or volunteering with a party to help. Instead, everyone I've seen that says this just casts their own statistically insignificant vote and discusses politics with family and friends, resulting in no swayed votes but straining otherwise healthy relationships. Maybe you are the exception.


Gemini 4 Argon was announced, but it isn't released yet.



And in six months when it is released, it'll be FAR behind publically available models.


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: