I’m away from computer so only skimmed the article until I noticed they use the ADG90x RF switch family.
IIRC that family has a single positive supply rail, and they make a nagative rail internally. The advantage is easy of implementation but you also inject some noise depending on switch position. Is this dealt with in the pub?
They do conduct an experiment to control for both switching noise, or other feed-through transmissions. They put two 50 ohm terminators on as the load, and find the receiver gets nothing. They also try cooling down one of the loads in liquid hydrogen, and the transmission resumes.
This is extremely strange. Shouldn't this be impossible by the second law of thermodynamics?
Receiving more energy when a resistor in thermal equilibrium is connected to an antenna seemingly implies that energy gets transferred from an antenna with a resistor to a an antenna without a resistor, even at thermal equilibrium.
Maybe we'll later find it's an experimental subtlety, like the faster than light neutrinos.
I'm not sure, as this paper is very confusing in its style, but I think they try to modulate the load resistance connected to an antenna on the TX side. Since the open has a large real resistance (25k in their case), it's poorly impedance matched to the antenna and most of the noise powwr cannot be transferred to the antenna at 1.4 GHz they are using. 50 ohm load is well matched, hence the max noise power transfer to the antenna. This is all at TX.
I don't see a thermodynamic problem here. Noise modulation with switching is well known in electronics. Switched resistors and their noise is already analyzed. It is often in the context of noise parr of SNR though. They are using this noise modulation to transmit signals, which isn't too novel. I guess the transmit power is very low but this would be horrible for Shannon spectral efficiency because of the need for noise power detection, which cannot be instantenous.
FWIW, the "Significance" blurb at the very top says as much:
> An interesting feature of the system is that all components of the system are at the same physical temperature, but it functions because they have different noise temperatures.
But it should. If you can get energy to flow one way between things at the same physical temperature that seems to violate thermodynamics and allow for perpetual motion.
Put a bowl of water down at room temperature, wait for one molecule to "evaporate", and I think you did it? Unfortunately they are now no longer at the same temperature.
Temperature is an average. You can still take advantage of the fact that there will be much lower and higher temperatures present. Also, what about Peltier effect? Same temperature when it starts.
> the maximum achievable throughput is 26 bps at 1.5 m and goes down to 22 bps at 4.5 m.
And when there are more than one transmitters in real world conditions instead of anechoic chamber?
It's cool and all and if this work is for it's own sake, for the sake of research, no issues. But otherwise, I struggle to think of practical use cases. I grant that my imagination may be deficient.
I think someone from OpenAI said something like "air gapping AI can't (hypothetically) work because it will figure out a way to basically morse code across the airgap". (I think this was in response to people calling them out for not airgapping)
> "air gapping AI can't (hypothetically) work because it will figure out a way to basically morse code across the airgap"
All things well-known for good two decades and written about by x-risk/LessWrong crowd, but to date continued to be dismissed as lunacy.
On one hand it's great to see it demonstrated in practice in concrete terms, on the other hand it's sad we have to basically make the mistakes ourselves, with live and close-to-dangerous systems, because we can't believe the obvious warnings.
I didn't mean they first invented or described the methods, but that they've been consistently presenting a solid argument that these and other methods make airgapping and sandboxing AGI a dangerously flawed idea from the start. They are the serious people who considered this, and to date nobody listens.
IIRC that family has a single positive supply rail, and they make a nagative rail internally. The advantage is easy of implementation but you also inject some noise depending on switch position. Is this dealt with in the pub?
Receiving more energy when a resistor in thermal equilibrium is connected to an antenna seemingly implies that energy gets transferred from an antenna with a resistor to a an antenna without a resistor, even at thermal equilibrium.
Maybe we'll later find it's an experimental subtlety, like the faster than light neutrinos.
I don't see a thermodynamic problem here. Noise modulation with switching is well known in electronics. Switched resistors and their noise is already analyzed. It is often in the context of noise parr of SNR though. They are using this noise modulation to transmit signals, which isn't too novel. I guess the transmit power is very low but this would be horrible for Shannon spectral efficiency because of the need for noise power detection, which cannot be instantenous.
Basically they are misleading when they say "thermal equilibrium", same physical temperature does not imply thermal equilibrium.
> An interesting feature of the system is that all components of the system are at the same physical temperature, but it functions because they have different noise temperatures.
(I assume it's not AI-generated.)
Put a bowl of water down at room temperature, wait for one molecule to "evaporate", and I think you did it? Unfortunately they are now no longer at the same temperature.
Temperature is an average. You can still take advantage of the fact that there will be much lower and higher temperatures present. Also, what about Peltier effect? Same temperature when it starts.
what's described in this paper is not passive, it has a powered amplifier in there which lowers the "noise temperature"
same way a heat pump can move heat between things at the same temperature - inside and outside of house when you start it
usage limit reached, resets at 7:50AM
(I am floored that this works. This is amazing.)
And when there are more than one transmitters in real world conditions instead of anechoic chamber?
It's cool and all and if this work is for it's own sake, for the sake of research, no issues. But otherwise, I struggle to think of practical use cases. I grant that my imagination may be deficient.
I guess your not on Twitter. It's not a coincidence this research paper from 2022 got posted today.
All things well-known for good two decades and written about by x-risk/LessWrong crowd, but to date continued to be dismissed as lunacy.
On one hand it's great to see it demonstrated in practice in concrete terms, on the other hand it's sad we have to basically make the mistakes ourselves, with live and close-to-dangerous systems, because we can't believe the obvious warnings.
https://www.nsa.gov/portals/75/documents/news-features/decla...
the x-risk/LessWrong crowd aren't adding anything to the conversation that seriously people haven't already considered.