Yea, technically it gets lighter because when you're writing a bit of data, you're storing a positive charge, which means an electron is removed. But what about magnetic storage medias?
I think this is backwards. Flash memory programs (writes zeros) faster than it erases (writes ones), and it erases in blocks. So I would expect an empty USB drive to be filled with ones, not zeros. Otherwise you would have to erase every "empty" block before the first write.
I get why writing zeros might be faster than writing ones (easier to discharge than charge) but I don’t see why that would mean an empty flash drive would start with all ones. Wouldn’t the manufacturer set it to all zeros as part of manufacturing it?
The truth is, there is no fundamental difference between a one and a zero. Flash is either charged or discharged. Physically 'erasing' means setting all the flash to charged and 'writing' means setting (part of) the flash to discharged. If charged means one or zero is a matter of interpretation. Most commonly, charged is considered a one.
You have it backwards. Programming means moving electrons (negative charge) into the floating gate of the flash transistor. This negative charge blocks the transistor from turning on, a state which is traditionally called "zero". Erasing means removing the trapped charge, which allows the transistor to turn on ("one"). With differential current sensing you could choose either state to be one or zero. I've only worked on NOR flash, so maybe in NAND flash it's the other way around.
Regardless of what you call the states, what makes flash flash (as opposed to EEPROM) is that it's erased in large blocks. So the chips should come out of the factory erased. Whether that means all ones or all zeros is up to the manufacturer.
When I was in college (over a decade ago) I worked in an Apple Store and one of my most memorable stories was trying to help somebody understand that an iPod (Classic I believe) wouldn't get heavier the more albums they stored on it. The customer was clearly not super computer literate, and kept comparing storing albums on an iPod like carrying around tapes or CDs on their Walkman.
They were amazed and delighted when I compared the weight of my iPhone to the weight of one of the store iPods and showed them how much more music was on my iPhone compared to the store iPod.
Teaching people about how technology can improve their lives can really be rewarding.
Of course once they were gone the staff ended up in a debate about the actual technical scientific answer. I don't think we ever came to an agreement on it.
I guess the actual analogy to tapes or CDs would be carrying around a pile of blank tapes vs tapes with recordings on them -- same weight, different amount of music
Actually CDs are bit funny. Wouldn't the weight ever so marginally depend on the record? And actually compared to blanks it would have different weight. At least for commercially pressed ones. As there would be pits in the disc. I'm not sure if you account for all components that is acrylic, aluminium and polycarbonate.
> Wouldn't the weight ever so marginally depend on the record?
This raises an interesting point. It's very undesirable to have long sequences of all 0 or 1 in a digital stream that is being read subject to a lot of analog effects. You want the line to return to a known state. This is why simple serial communication protocol like RS-232 uses a mark bit, to synchronize the clock repeatedly to protect against drift.
It's possible -- and usually preferable -- to implement a digital coding in a way such that the average signal intensity is 50% and there are an approximately equal number of 0s and 1s physically speaking, with frequent crossings, regardless of the actual signal contents. Unsurprisingly, CDs do this on several levels, first permutating the stream into blocks with error correction, and then into a line coding that has regular transitions even if the stream is all 0/1. Though they're not guaranteed to literally have the exact same number of 0 and 1 bits across any content.
For vinyl there would absolutely be a small weight difference, as there is actual physical grooves that gets cut in order for the record player to reproduce the sound you "entered" (more accurately, took away) into the groove.
Imagine a bound notebook, like a student would use in school. At the end of the semester the book is full of writing, but in practical terms it doesn't weigh more than it did when it was blank.
The iPod has a tiny notebook inside that it uses to remember your songs.
The analogy that came to my mind was that if you take a deck of playing cards and lay all the cards out on a table, face down, the table weighs a certain amount (the table's weight, plus weight of the cards). Then, if you flip over five cards at random, you have created enough new information to make a poker hand. But, the weight of the table did not change when you created that information.
Take a 2 TiB flash drive. That’s 1e13 bytes or 1e16 bits. Suppose it’s one electron per bit. That’s approaching 1% of a Coulomb! So you take two flash drives, charge them up, and hold them 1cm apart, and they fly apart with several hundred meganewtons of force! The electrostatic force is strong!
This is, of course, completely wrong. When you write to flash, you are moving electrons around, not adding net electrons. The whole device stays electrically neutral (on average), and any mass change is due solely to minuscule relativistic effects if you add or remove energy.
> I worked on all sorts of problems that involved the Coulomb interaction, and occasionally my proposed solution would be very wrong. The worst kind of wrong was the one that made my advisor remark “What you just created is a Coulomb bomb,” which meant that I had proposed something that wasn’t neutral on the large scale.
Right. The only actual effect would be like, having two charged (or two uncharged? maybe?) elements right next to each other entails some repulsive force, and so it takes additional energy to get to that state. That energy has mass via E=mc^2.
Exactly, I was going to point out the same thing. The flash cells might be charged or discharged of electrons, but the whole flash drive can't have a net charge or else you will certainly feel it. And because no solid/liquid/gaseous matter is exchanged between the computer and flash drive, there are no positive particles (nuclei/protons) moving either.
This is similar to what it means to "charge" a capacitor. It doesn't mean injecting or removing electrons from it, because that would cause a net charge. It actually means shifting electrons from one electrode to the other, separated by an insulated gap. http://amasci.com/emotor/cap1.html
Calling it a shift is not right either. It's not the same electron that leaves one side and makes it way to the other. Electrons injected on one side and removed on the other.
> Do you think a physicist would be less certain that OP is wrong, than how certain OP is that he is right?
This is not the correct application of the effect.
A physicist would have been sure about this particular case being wrong, but would likely be a lot more careful and concise about what he's talking about. Also a physicist is supposed to be more confident, as they are on the far right side of the curve.
Being at the bottom of that curve is reserved for folks like me who rely on common sense but cannot be certain outside of extremely basic cases like knowing that electrons are not leaving the USB en masse.
Also consider that most high density Flash today is made from TLC cells which have 8 distinct Vt levels each. That means in average case (i.e. random data) there is a minimum of 3.5 electrons per cell required.
In reality, much more than 1 electron is required to store a 0 or 1 with any level of reliability. On the order of 10s to 100s of electrons per bit even on leading processes.
What about the mechanical strain caused by the charges? Would the device be more compressed due to the attraction between the opposite charges associated with charged bits? Since the device is in air, it would then displace less air. Even though the device mass would be the same, the weight would be more.
I had this same question for my Tesla when I first got it, but I came to the conclusion charging was just a chemical process that alters the state of things. Still don't know the actual answer.
Re GP: It's a closed system, with a constant number of neutrons, protons, and electrons (electrons flow in and out, but protons don't, so the number of electrons has to stay the same).
The mass being added and removed is the mass created and destroyed when you destroy/create a chemical bond. Bonds store energy, and that energy is converted to mass following e=mc^2.
The claim that bonds store energy is a frequent misconception. Bonds are the result of lost energy compared to free, unbound atoms.
If bonds don't store energy, how does breaking hydrocarbon bonds release energy? It doesn't - it's just that the new bonds formed in CO2 and H2O release more energy than it took to break the hydrocarbon bonds.
Energy is not
not a substance. Is a meaure of an entire configuration of states. Forms of energy is an "interpretation", a metpahor.
C02 + H20 has less energy than CH3 + O2, because C02 + H20 are more stable (= lower energy) bonds.
You can say the energy is in the electron, released as it falls into a lower energy state (bond).
If I walk up a hill to a cliff, am I then in a more energetic state?
No, because it takes energy to push me off the cliff, the ground is a more stable position than the top of the ledge and the appparent energy released by me turning a windmill I hit on the way down is because more energy is released by freefalling off a cliff than by sitting down on top of the ledge?
I guess it depends on the type of battery, with ol' lead batteries you used density meters (hydrometers) to measure charge level, and to more dense liquid corresponded higher level of charge.
What's happening there is sulphur and oxygen is moving out of solution onto the solid lead plates. The mass of the liquid changes, the mass of the battery as a whole stays the same (apart from e=mc^2 as mentioned in a sibling of your comment).
This assumes that the memory there is zeroed entirely, while in reality you only zero some small parts of it to make up the logical structure of whatever filesystem you use. The rest, as in 99.(9)% of the memory, is in undefined state and becomes a zero or a one only when the first write occurs.
While massively out-weighed by the electrons and baryonic matter, a more interesting conjecture is mass-energy-information equivalence. Do your files themselves have mass aside from the electrons and nuclei of the thumb drive? Conservation of information at the quantum level would be an amazing expansion on our current understanding of thermal physics.
Why would we entertain this conjecture? Why is it plausible in the slightest, since it seems to clash with intuition that information is a completely separate thing to mass-energy?
> it seems to clash with intuition that information is a completely separate thing to mass-energy?
How intuitive is the mass-energy equivalence? Before Einstein's era, people might have said that such an idea clashes with the intuition that is a completely separate thing from energy.
I didn’t say all valid ideas must be intuitive, but that if an idea is not intuitive then we should have some other reason to believe it is worth investigating.
A few reasons to consider something for study is either existing empirical evidence or internal consistency of a theory. The reason information conservation and energy/mass/information equivalence are interesting is because there is a possibility for such consistency in a few theories.
Physicists are always looking for equivalences and conserved quantities. They are the basis for modern physics. It's how we discovered that 3 of the 4 known forces were all based on the same equivalent principle. You find a conserved quantity that unifies the two forces.
Possibly one of the greatest feats of this was Maxwell when he unified electrical and magnetic forces by discovering an equivalence between the two that allowed energy to be conserved. This helped us to discover the underlying principles of light itself and describe it as an electromagnetic wave.
Information conservation could prove to unify gravity with the remaining forces. Or not! Or it could break them all apart again. That's just what scientific inquiry is all about.
I don't have the knowledge or smarts to give much detail about it but there's some rather interesting parts of this in Quantum Mechanics and Entropy. Along with that it's also one of the fun little areas of things that may or may not be breaking down with our current models around black-holes and our understanding of gravity.
It's an active area of research because of this because it may point to a way to reconcile the incompatibilities between QM and Relativity to get a Grand Unified Theory. I believe Loop Quantum Gravity ends up with effects around this mass-energy-information/entropy equivalence idea as an emergent property of how it's trying to reconcile QM and Relativity.
Intuition hasn't been a good guide for physics since newton. It's not clear to me that anything physical is separate from the information that describes it though.
Even Newtonian physics can be very non-intuitive. Especially when things get complicated. A lot of experiments done in introductory physics courses revolve around this. Stuff like a feather and a bowling ball falling in a vacuum and throwing a ball on a merry go round. Rotating reference frames are the worst...
Information having energy is well known since Boltzmann, as entropy is measured in J/K. I don't think that information is "equivalent" to energy. (Also, I don't think that mass is equivalent to energy either. Sure, if you add energy, you gain mass and vice versa, but since things (for example things with charge) can't be created or destroyed (not even with black holes or antiparticles), talking about the equivalence of energy and mass is total nonsense.)
Yes, you are right, you can destroy an electron with a positron, but then you destroy the positron too, so you have no information about "the energy of the electron".
To be more concrete, bmacho equation, E = mc^2 + q*6*10^23 J/e holds for everything, every experiment, every theory (standard model, GR). It states that energy is equivalent to mass and charge.
> Also, I don't think that mass is equivalent to energy either.
Unfortunately, your argument here is implausible. There have been tens of thousands of experiments demonstrating mass to energy conversion. It is the basic principle behind nuclear energy production. Mass energy equivalence is necessary to maintain conservation laws and has been proven by empirical observation. The total mass/energy in the system will not change, however the two can be exchanged for each other.
> Information having energy is well known since Boltzmann
Information having mass/energy has not been proven. There have been no experiments to support this argument. It is merely a theory.
> ..as entropy is measured in J/K
Entropy is not a conserved quantity since it will increase within a closed system. The Joules/Kelvin unit comes from statistical mechanics where temperature is defined as a change in entropy per increase in total energy. The entropy used in information theory is different, although related. Entropy in information theory is measured in bits.
> if this was true wouldn't it imply that you could tell the difference between random and encrypted data?
Possibly.
> And if you ascribed meaning to random data it would gain weight...
You would have to create a dictionary with that meaning which would be more information that would add that theorized weight. This is a common problem in compression.
but if information alone is weight then creating that dictionary separately would add to the weight of the exact order of the random data now has meaning.
When you assign a meaning to something, you are merely establishing a relationship between two systems which is information itself. It does not add information to either of the two systems. It merely creates third system with it's own information.
For instance, if you have the state "H" and want to add a meaning to it "High Temperature" you need to create a third piece of information "H" -> "High Temperature"
You are not changing the information in "H" or "High Temperature" you are merely creating a third piece of information which is the lambda.
but isn't that like...everything the only way a hard drive has information is because though language and other technical specs we assign meaning to the things on the hard drive?
Those meanings are stored in billions of human brains which trigger neural networks which cause us to exhibit a behavior. The meanings are not intrinsic to the universe, but are specific to our own knowledge.
Take for instance the word "Red." To a computer this means: 52, 65, 64
The there is a mapping to letters call an ASCII table or UTF-8 encoding. Then there is a mapping from character data to a bitmap containing the letters projection on a screen. Then there is a mapping in your brain that associates that sequence of letters to a neural structure that encodes your memory of the color red. That mapping was created by a teacher that showed you the color red next to those letters.
But in the end, the numbers 52, 65, 64 haven't gained any information. They could still mean anything based on which system interacts with them. An RGB color picker will interpret it as a nice teal color. Without knowing what system to use to interpret their meaning, they are meaningless aside from the numbers themselves. You need the key to find the difference which is information itself.
The information content of a file is approximated by it's compressed size. If you have a 1Gb file of all zeros or all ones it will compress down to almost nothing, but a 1Gb file of random numbers will hardly compress at all.
We don't know. I wonder if the universe uses compression on stuff like that? It's a funny idea, but if information actually is mass/energy, you could measure that mass/energy, and if it was different for the cleared drive... Then you could prove that the universe is self-compressing. That would be a bit mind blowing.
The amount of audio stored on a tape depends on the transport speed chosen. Slower speed means less bandwidth, which means high-frequency signals are lost.
The amount of audio stored on a vinyl record depends on the amplitude of the signal. Loud sounds and thumping bass will increase the spacing between tracks, thus reducing the recording time.
This article is completely wrong. For every electron you place in the floating gate of the transistor, an electron is removed from the bulk or substrate of the device. The net weight is exactly the same, you are only moving electrons within the device.
I was uncertain about this as well, but wouldn't the electron-depletion in the substrate be (slowly) replaced from the ground?
In any case, the charge on the device would be negligible in any direction. It's a ridiculously small amount of charge when compared to the overall device, or even the substrate material of the flash memory cells.
Would be fun to do an experiment though. Maybe in 100 years we'll be able to do that.
Funnily enough, there wouldn't be any "depletion from the ground" in the substrate either. If the charge in the substrate was removed and sent to ground, the charges in the gate would also disappear and be sent to ground simultaneously. In that case, the circuit has lost its stored valu. Ideally, this should never happen, and if it does the circuit is literally broken.
Yep. NOR and NAND flash are constructed differently, but programming/erasure and tunnel injection/release are changes in configuration rather than accumulation/release of charge.
Core memory (ferrite bead), SRAM (4 transistors/bit) and DRAM (1 transistor with a pseudo capacitor/bit) would increase in relativistic weight depending on their state.
The same is true of batteries, ultracaps, or anything that consumes or stores large amounts of energy. m = E/c^2. 1e17 J weighs about 1 kg. That's about the energy of 2 of the large nuclear weapons ever tested.
Let's suppose we have an average largish lake, 100 km^3 or 1e14 kg of water, and we store it at average height of 60 m above a generator. PE=m*g*h. That 6e16 J of potential energy weighs 651g.
> However, you would need to weigh more USB drives than exist on the planet together at once before the difference in weight became easily measurable.
Scales precisions limits are not absolute, but relative to their measuring range, so if a weight change is too small to measure, the solution is better scales, not more stuff.
The interplay of gravitation and information invariably leads to the black hole information paradox [1]
Which suggests that one way to decide is to take two USB drives, one full of information and the other empty and throw each into a separate black hole of identical mass.
Then wait until both black holes evaporate and check if there is any black hole hair left.
My guess would the that the act of inserting or removing the drive from USB port would result in larger weight difference as the connector contact areas rub on each other. Depending on hardness of each side you could be either losing or gaining weight. This is excluding any possibly debris falling out or being stuck in the port.
Disclaimer, only a fun thought exercise, I know very little about these things.
82 comments
[ 2.9 ms ] story [ 165 ms ] threadHow the cell presents "voltage" has to do with exactly the wiring of the source and drain on the flash cell, not the basic structure.
The weight of the drive varies with the sum of the charged cells.
Regardless of what you call the states, what makes flash flash (as opposed to EEPROM) is that it's erased in large blocks. So the chips should come out of the factory erased. Whether that means all ones or all zeros is up to the manufacturer.
They were amazed and delighted when I compared the weight of my iPhone to the weight of one of the store iPods and showed them how much more music was on my iPhone compared to the store iPod.
Teaching people about how technology can improve their lives can really be rewarding.
Of course once they were gone the staff ended up in a debate about the actual technical scientific answer. I don't think we ever came to an agreement on it.
This raises an interesting point. It's very undesirable to have long sequences of all 0 or 1 in a digital stream that is being read subject to a lot of analog effects. You want the line to return to a known state. This is why simple serial communication protocol like RS-232 uses a mark bit, to synchronize the clock repeatedly to protect against drift.
It's possible -- and usually preferable -- to implement a digital coding in a way such that the average signal intensity is 50% and there are an approximately equal number of 0s and 1s physically speaking, with frequent crossings, regardless of the actual signal contents. Unsurprisingly, CDs do this on several levels, first permutating the stream into blocks with error correction, and then into a line coding that has regular transitions even if the stream is all 0/1. Though they're not guaranteed to literally have the exact same number of 0 and 1 bits across any content.
The iPod has a tiny notebook inside that it uses to remember your songs.
Take a 2 TiB flash drive. That’s 1e13 bytes or 1e16 bits. Suppose it’s one electron per bit. That’s approaching 1% of a Coulomb! So you take two flash drives, charge them up, and hold them 1cm apart, and they fly apart with several hundred meganewtons of force! The electrostatic force is strong!
This is, of course, completely wrong. When you write to flash, you are moving electrons around, not adding net electrons. The whole device stays electrically neutral (on average), and any mass change is due solely to minuscule relativistic effects if you add or remove energy.
> I worked on all sorts of problems that involved the Coulomb interaction, and occasionally my proposed solution would be very wrong. The worst kind of wrong was the one that made my advisor remark “What you just created is a Coulomb bomb,” which meant that I had proposed something that wasn’t neutral on the large scale.
This is similar to what it means to "charge" a capacitor. It doesn't mean injecting or removing electrons from it, because that would cause a net charge. It actually means shifting electrons from one electrode to the other, separated by an insulated gap. http://amasci.com/emotor/cap1.html
Ironically, that makes a Dunning-Kruger Effect effect.
https://gwern.net/doc/iq/2020-gignac.pdf
Do you think a physicist would be less certain that OP is wrong, than how certain OP is that he is right?
This is not the correct application of the effect.
A physicist would have been sure about this particular case being wrong, but would likely be a lot more careful and concise about what he's talking about. Also a physicist is supposed to be more confident, as they are on the far right side of the curve.
Being at the bottom of that curve is reserved for folks like me who rely on common sense but cannot be certain outside of extremely basic cases like knowing that electrons are not leaving the USB en masse.
(Also, 2 TiB is just 2e12 bytes, and 2e12 bytes only equals 1.6e13 bits, it’s log2(8) that’s 3, but that doesn’t really affect your conclusion.)
In reality, much more than 1 electron is required to store a 0 or 1 with any level of reliability. On the order of 10s to 100s of electrons per bit even on leading processes.
What about the mechanical strain caused by the charges? Would the device be more compressed due to the attraction between the opposite charges associated with charged bits? Since the device is in air, it would then displace less air. Even though the device mass would be the same, the weight would be more.
Re GP: It's a closed system, with a constant number of neutrons, protons, and electrons (electrons flow in and out, but protons don't, so the number of electrons has to stay the same).
The mass being added and removed is the mass created and destroyed when you destroy/create a chemical bond. Bonds store energy, and that energy is converted to mass following e=mc^2.
If bonds don't store energy, how does breaking hydrocarbon bonds release energy? It doesn't - it's just that the new bonds formed in CO2 and H2O release more energy than it took to break the hydrocarbon bonds.
Energy is not not a substance. Is a meaure of an entire configuration of states. Forms of energy is an "interpretation", a metpahor.
C02 + H20 has less energy than CH3 + O2, because C02 + H20 are more stable (= lower energy) bonds.
You can say the energy is in the electron, released as it falls into a lower energy state (bond).
If I walk up a hill to a cliff, am I then in a more energetic state?
No, because it takes energy to push me off the cliff, the ground is a more stable position than the top of the ledge and the appparent energy released by me turning a windmill I hit on the way down is because more energy is released by freefalling off a cliff than by sitting down on top of the ledge?
or Yes, for exactly the same reason?
https://pubs.aip.org/aip/sci/article/2022/9/091111/2849001/A...
How intuitive is the mass-energy equivalence? Before Einstein's era, people might have said that such an idea clashes with the intuition that is a completely separate thing from energy.
Physicists are always looking for equivalences and conserved quantities. They are the basis for modern physics. It's how we discovered that 3 of the 4 known forces were all based on the same equivalent principle. You find a conserved quantity that unifies the two forces.
Possibly one of the greatest feats of this was Maxwell when he unified electrical and magnetic forces by discovering an equivalence between the two that allowed energy to be conserved. This helped us to discover the underlying principles of light itself and describe it as an electromagnetic wave.
Information conservation could prove to unify gravity with the remaining forces. Or not! Or it could break them all apart again. That's just what scientific inquiry is all about.
It's an active area of research because of this because it may point to a way to reconcile the incompatibilities between QM and Relativity to get a Grand Unified Theory. I believe Loop Quantum Gravity ends up with effects around this mass-energy-information/entropy equivalence idea as an emergent property of how it's trying to reconcile QM and Relativity.
Intuition hasn't been a good guide for physics since newton. It's not clear to me that anything physical is separate from the information that describes it though.
To be more concrete, bmacho equation, E = mc^2 + q*6*10^23 J/e holds for everything, every experiment, every theory (standard model, GR). It states that energy is equivalent to mass and charge.
Unfortunately, your argument here is implausible. There have been tens of thousands of experiments demonstrating mass to energy conversion. It is the basic principle behind nuclear energy production. Mass energy equivalence is necessary to maintain conservation laws and has been proven by empirical observation. The total mass/energy in the system will not change, however the two can be exchanged for each other.
> Information having energy is well known since Boltzmann
Information having mass/energy has not been proven. There have been no experiments to support this argument. It is merely a theory.
> ..as entropy is measured in J/K
Entropy is not a conserved quantity since it will increase within a closed system. The Joules/Kelvin unit comes from statistical mechanics where temperature is defined as a change in entropy per increase in total energy. The entropy used in information theory is different, although related. Entropy in information theory is measured in bits.
And if you ascribed meaning to random data it would gain weight...
Possibly.
> And if you ascribed meaning to random data it would gain weight...
You would have to create a dictionary with that meaning which would be more information that would add that theorized weight. This is a common problem in compression.
For instance, if you have the state "H" and want to add a meaning to it "High Temperature" you need to create a third piece of information "H" -> "High Temperature"
You are not changing the information in "H" or "High Temperature" you are merely creating a third piece of information which is the lambda.
Take for instance the word "Red." To a computer this means: 52, 65, 64
The there is a mapping to letters call an ASCII table or UTF-8 encoding. Then there is a mapping from character data to a bitmap containing the letters projection on a screen. Then there is a mapping in your brain that associates that sequence of letters to a neural structure that encodes your memory of the color red. That mapping was created by a teacher that showed you the color red next to those letters.
But in the end, the numbers 52, 65, 64 haven't gained any information. They could still mean anything based on which system interacts with them. An RGB color picker will interpret it as a nice teal color. Without knowing what system to use to interpret their meaning, they are meaningless aside from the numbers themselves. You need the key to find the difference which is information itself.
A lot of "Ifs" though.
They were fixated on much storage in bytes it must have to store the music
The amount of audio stored on a vinyl record depends on the amplitude of the signal. Loud sounds and thumping bass will increase the spacing between tracks, thus reducing the recording time.
In any case, the charge on the device would be negligible in any direction. It's a ridiculously small amount of charge when compared to the overall device, or even the substrate material of the flash memory cells.
Would be fun to do an experiment though. Maybe in 100 years we'll be able to do that.
Core memory (ferrite bead), SRAM (4 transistors/bit) and DRAM (1 transistor with a pseudo capacitor/bit) would increase in relativistic weight depending on their state.
The same is true of batteries, ultracaps, or anything that consumes or stores large amounts of energy. m = E/c^2. 1e17 J weighs about 1 kg. That's about the energy of 2 of the large nuclear weapons ever tested.
Let's suppose we have an average largish lake, 100 km^3 or 1e14 kg of water, and we store it at average height of 60 m above a generator. PE=m*g*h. That 6e16 J of potential energy weighs 651g.
> However, you would need to weigh more USB drives than exist on the planet together at once before the difference in weight became easily measurable.
Scales precisions limits are not absolute, but relative to their measuring range, so if a weight change is too small to measure, the solution is better scales, not more stuff.
Which suggests that one way to decide is to take two USB drives, one full of information and the other empty and throw each into a separate black hole of identical mass.
Then wait until both black holes evaporate and check if there is any black hole hair left.
[1] https://en.wikipedia.org/wiki/Black_hole_information_paradox
Disclaimer, only a fun thought exercise, I know very little about these things.
Good encryption algos will have an average of 50% 1s and 0s regardless of the plaintext data.