There is no action at a distance in quantum mechanics, that is a laymen's misconception. If there was, it would not be compatible with special relativity, but it is compatible as they are already unified under the framework of quantum field theory. The No-communication theorem is a rather simple proof that shows there is no "sharing at a distance" in quantum mechanics. It is an entirely local theory. The misconception arises from people misinterpreting Bell's theorem which says quantum mechanics is not compatible with a local hidden variable theory, so people falsely conclude it's a nonlocal theory, but this is just false because quantum mechanics is not a hidden variable theory, and so it is not incompatible with locality. It is a local theory. Bell's theorem only shows it is nonlocal if you introduce hidden variables, meaning the theorem is really only applicable to a potential replacement to quantum mechanics and is not even applicable to quantum mechanics itself. It is applicable to things like pilot wave theory, but not to quantum theory.
pcalau12i
Personally I think general knowledge is kind of a useless metric because you're not really developing "intelligence" at that point just a giant dictionary, and of course bigger models will always score better because they are bigger. In some sense training an ANN is kinda like a compression algorithm of a ton of knowledge, so the bigger the parameters the less lossy the compression it is, the more it knows. But having an absurd amount of knowledge isn't what makes humans intelligent, most humans know very little, it's problem solving. If we have a problem solving machine as intelligent as a human we can just give it access to the internet for that information. Making it bigger with more general knowledge, imo, isn't genuine "progress" in intelligence. The recent improvements by adding reasoning is a better example of genuine improvements to intelligence.
These bigger models are only scoring better because they have just memorized so much they have seen similar questions before. Genuine improvements to intelligence and progress in this field come when people figure out how to improve the results without more data. These massive models already have more data than ever human could ever have access to in hundreds of lifetimes. If they aren't beating humans on every single test with that much data then clearly there is something else wrong.
That's just the thing, though, the point I am making, which is that it turns out in practice synthetic data can give you the same effect as original data. In some sense, training an LLM is kind of like a lossy compression algorithm, you are trying to fit petabytes of data into a few hundred gigabytes as efficiently as possible. In order to successfully compress it, it has to lose specifics, so the algorithm only captures general patterns. This is true for any artificial neural network, so if you train another neural network with the data yourself, you will also lose specifics in the training process and end up with a model that only knows general patterns. Hence, if you train a model using synthetic data, the information lost in that synthetic data will be information the AI you are training would lose anyways, so you don't necessarily get bad results.
But yes, when I was talking about synthetic data I had in mind data purely generated from an LLM. Of course I do agree translating documents, OCRing documents, etc, to generate new data is generally a good thing as well. I just disagree with your final statement there that it is critical to have a lot of high-quality original data. The notion that we can keep making AIs better by just giving them more and more data, this method is already plateauing in the industry and showing diminishing returns. ChatGPT 3.5 to 4 was a massive leap but the jump to 4.5, which uses an order of magnitude more compute mind you, is negligible.
Just think about it. Humans are way smarter than ChatGPT and we don't require the energy of a small country and petabytes of all the world's information to solve simple logical puzzles, just a hot pocket and a glass of water. There is clearly an issue in how we are training things and not the lack of data. We have plenty of data. Recent breakthroughs have come in finding more clever ways to use the data rather than just piling on more and more data.
For example, many models have recently adopted reasoning techniques, so rather than simply spitting out an answer it generates an internal dialog prior to generating the answer, it "thinks" about the problem for a bit. These reasoning models perform way better on complex questions. OpenAI first invented the technique but kept it under lock and key, and the smaller company DeepSeek managed to replicate it and made their methods open source for everyone, and then Alibaba put it into their Qwen model in a new model they call QwQ which dropped recently and performs almost as well as ChatGPT 4 on some benchmarks yet can be run on consumer-end hardware with as little as 24GB of VRAM.
All the major breakthroughs happening recently are coming from not having more data but using the data in more clever ways. Just recently a diffusion LLM dropped which creates text output but borrows the same techniques used in image generation, so rather than doing it character-by-character it outputs a random sequence of characters all at once and continually refines it until it makes sense. This technique is used with images because uncompressed images take up megabytes of data while LLM outputs only output a few kilobytes in a response, so it would just be too slow to use the same method for image generation, yet by applying the image generation method to do what LLMs do it makes it produce reasonable outputs faster than any traditional LLM.
This is a breakthrough that just happened, here's an IBM article on it from 3 days ago!
https://www.ibm.com/think/news/diffusion-models-llms
The breakthroughs are really not happening in huge data collection right now. Companies will still steal all your data because big data collection is still profitable to sell to advetisers, but it's not at the heart of the AI revolution right now. That is coming from computer science geniuses who cleverly figure out how to use the data in more effective ways.
We know how it works, we just don’t yet understand what is going on under the hood.
Why should we assume "there is something going on under the hood"? This is my problem with most "interpretations" of quantum mechanics. They are complex stories to try and "explain" quantum mechanics, like a whole branching multiverse, of which we have no evidence for.
It's kind of like if someone wanted to come up with deep explanation to "explain" Einstein's field equations and what is "going on under the hood". Why should anything be "underneath" those equations? If we begin to speculate, we're doing just tha,t speculation, and if we take any of that speculation seriously as in actually genuinely believe it, then we've left the realm of being a scientifically-minded rational thinker.
It is much simpler to just accept the equations at face-value, to accept quantum mechanics at face-value. "Measurement" is not in the theory anywhere, there is no rigorous formulation of what qualifies as a measurement. The state vector is reduced whenever a physical interaction occurs from the reference point of the systems participating in the interaction, but not for the systems not participating in it, in which the systems are then described as entangled with one another.
This is not an "interpretation" but me just explaining literally how the terminology and mathematics works. If we just accept this at face value there is no "measurement problem." The only reason there is a "measurement problem" is because this contradicts with people's basic intuitions: if we accept quantum mechanics at face value then we have to admit that whether or not properties of systems have well-defined values actually depends upon your reference point and is contingent on a physical interaction taking place.
Our basic intuition tells us that particles are autonomous entities floating around in space on their lonesome like little stones or billiard balls up until they collide with something, and so even if they are not interacting with anything at all they meaningfully can be said to "exist" with well-defined properties which should be the same properties for all reference points (i.e. the properties are absolute rather than relational). Quantum mechanics contradicts with this basic intuition so people think there must be something "wrong" with it, there must be something "under the hood" we don't yet understand and only if we make the story more complicated or make a new discovery one day we'd "solve" the "problem."
Einstein once said, God does not place dice, and Bohr rebutted with, stop telling God what to do. This is my response to people who believe in the "measurement problem." Stop with your preconceptions on how reality should work. Quantum theory is our best theory of nature and there is currently no evidence it is going away any time soon, and it's withstood the test of time for decades. We should stop waiting for the day it gets overturned and disappears and just accept this is genuinely how reality works, accept it at face-value and drop our preconceptions. We do not need any additional "stories" to explain it.
The blind spot is that we don’t know what a quantum state IS. We know the maths behind it, but not the underlying physics model.
What is a physical model if not a body of mathematics that can predict outcomes? The physical meaning of the quantum state is completely unambiguous, it is just a list of probability amplitudes. Probability captures the likelihoods of certain outcomes manifesting during an interaction, although quantum probability amplitudes are somewhat unique in that they are complex-valued, but this is to add the additional degrees of freedom needed to simultaneously represent interference phenomena. The state vector is a mathematical notation to capture likelihoods of events occurring while accounting for interference effects.
It’s likely to fall out when we unify quantum mechanics with general relativity, but we’ve been chipping at that for over 70 years now, with limited success.
There has been zero "progress" because the "problem" of unifying quantum mechanics and general relativity is a pseudoproblem. It stems from a bias that because we had success quantizing all the fundamental forces except gravity, then therefore gravity should be quantizable. Since the method that worked for all other forces failed, this being renormalization, all these other theories search for a different way to do it.
But (1) there is no reason other than blind faith to think gravity should be quantized, and (2) there is no direct compelling evidence that either quantum mechanics or general relativity are even wrong.
Also, we can already unify quantum mechanics and general relativity just fine. It's called semi-classical gravity and is what Hawking used to predict that black holes radiate. It makes quantum theory work just fine in a curved spacetime and is compatible with all experimental predictions to this day.
People who dislike semiclassical gravity will argue it seems to make some absurd predictions in under specific conditions we currently haven't measured. But this isn't a valid argument to dismiss it, because until you can actually demonstrate via experiment that such conditions can actually be created in physical reality, then it remains a purely metaphysical criticism and not a scientific one.
If semi-classical gravity is truly incorrect then you cannot just point to it having certain strange predictions in certain domains, you also have to demonstrate it is physically possible to actually probe them and this isn't just a metaphysical quirk of the theory of trying to make predictions to things that aren't physical possible in the first place and thus naturally what it would predict would also be physically impossible.
If you could construct such an experiment and its prediction was indeed wrong, you'd disprove it the very second you turned on the experiment. Hence, if you genuinely think semi-classical gravity is wrong and you are actually following the scientific method, you should be doing everything in your power to figure out how to probe these domains.
But instead people search for many different methods of trying to quantize gravity and then in a post-hoc fashion look for ways it could be experimentally verified, then when it is wrong they go back and tweak it so it is no longer ruled out by experiment, and zero progress has been made because this is not science. Karl Popper's impact on the sciences has been hugely detrimental because now everyone just believes if something can in principle be falsified it is suddenly "science" which has popularized incredibly unscientific methods in academia.
Sorry but both the "measurement problem" and the "unification problem" are pseudoproblems and not genuine scientific problems but both stems from biases on how we think nature should work rather than just fitting the best physical model to the evidence and accepting this is how nature works. Physics is making enormous progress and huge breakthroughs in many fields, but there has been zero "progress" in the fields of "solving the measurement "problem" or quantizing gravity because neither of these are genuine scientific problems.
They have been working at this "problem" for decades now and what "science" has come out of it? String Theory which is only applicable to an anti-de Sitter space despite our universe being a de Sitter space, meaning it only applies to a hypothetical universe we don't live in? Loop Quantum Gravity which can't even reproduce Einstein's field equations in a limiting case? The Many Worlds Interpretation which no one can even agree what assumptions need to be added to be able to mathematically derive the Born rule, and thus there is also no agreed upon derivation? What "progress" besides a lot of malarkey on people chasing a pseudoproblem?
If we want to know how nature works, we can just ask her, and that is the scientific method. The experiments are questions, the results are her answers. We should believe her answers and stop calling her a liar. The results of experimental practice---the actual real world physical data---should hold primacy above everything else. We should set all our preconceptions aside and believe whatever the data tells us. There is zero reason to try and update our theories or believe they are "incomplete" until we get an answer from mother nature that contradicts with our own theoretical predictions.
People always cry about how fundamental physics isn't "making progress," but what they have failed to justify is why it should progress in the first place. The only justification for updating a theory is, again, to better fit with experimental data, but they present no data. They just complain it doesn't fit some bias and preconception they have. That is not science.
Eh, individuals can't compete with corpos not just because they have access to more data but because making progress in AI requires a large team of well-educated researchers and sufficient capital to be able to experiment with vast technology. It's a bit like expecting an individual or small business to be able to compete with smartphone manufacturers. It really is not feasible not simply because smartphone manufacturers are using dirty practices but because producing smartphones requires an enormous amount of labor and capital and simply cannot be physically carried out by an individual.
This criticism might be more applicable to a medium-sized business like DeepSeek that is not really "small" but smaller than the others (and definitely not a single individual) and still big enough to still compete, and we can see they still could compete just fine despite the current situation.
The truth is that both USA and China recognize all purely AI-generated work as de facto public domain. That means anything ChatGPT or whatever spits out, no matter what their licensing says, is absolutely free to use however you wish and you will win in court if they try to stop you. There is a common myth that training AI on synthetic data will always be negative. It's actually only sometimes true if you train the AI on its own synthetic data, but through a process they call "distillation" you can train a less intelligent AI on synthetic data from a more intelligent AI and it will actually improve its performance.
That means any AI made by big companies can be distilled into any other AI to improve its performance. This is because you effectively have access to all the data the big companies have access to but indirectly through the synthetic data their AI can produce. For example, if for some reason you curated the information the AI was trained on so it never encountered the concept of a dog, it simply wouldn't know what a dog is. If it encountered it a lot, it would know what a dog is and could explain it if you asked. Hence, that information is effectively accessible indirectly by simply asking the AI for it.
If you use distillation then you should can make effectively your own clones of any big company's AI model and it's perfectly legal. Not only that, but you can make improvements to it as well. You aren't just cloning models, but you have the power to modify them. during this distillation process.
Imagine if the initial model was trained using a particular technique that is rather outdated and you believe you've invented a new method that if re-trained would produce a smarter AI, but you simply lack access to the original data. What you can instead do is generate a ton of synthetic data from the AI and then train your new AI using the new method on that synthetic data. Your new AI will have access to most of the same information but now trained on a superior technique.
We have seen some smaller companies already take pre-existing models and use distillation to improve them, such as DeepSeek taking the Qwen models and distilling R1 reasoning techniques into them to improve their performance.
On the surface, it does seem like there is a similarity. If a particle is measured over here and later over there, in quantum mechanics it doesn't necessarily have a well-defined position in between those measurements. You might then want to liken it to a game engine where the particle is only rendered when the player is looking at it. But the difference is that to compute how the particle arrived over there when it was previously over here, in quantum mechanics, you have to actually take into account all possible paths it could have taken to reach that point.
This is something game engines do not do and actually makes quantum mechanics far more computationally expensive rather than less.
So usually this is explained with two scientists, Alice and Bob, on far away planets. They’re each in the possession of a particle that is entangled with the other, and in a superposition of state 1 and state 2.
This "usual" way of explaining it is just overly complicating it and making it seem more mystical than it actually is. We should not say the particles are "in a superposition" as if this describes the current state of the particle. The superposition notation should be interpreted as merely a list of probability amplitudes predicting the different likelihoods of observing different states of the system in the future.
It is sort of like if you flip a coin, while it's in the air, you can say there is a 50% chance it will land heads and a 50% chance it will land tails. This is not a description of the coin in the present as if the coin is in some smeared out state of 50% landed heads and 50% landed tails. It has not landed at all yet!
Unlike classical physics, quantum physics is fundamentally random, so you can only predict events probabilistically, but one should not conflate the prediction of a future event to the description of the present state of the system. The superposition notation is only writing down probability amplitudes of the likelihoods of what you will observe (state 1 or state 2) of the particles in the future event that you go to the interact with it and is not a description of the state of the particles in the present.
When Alice measures the state of her particle, it collapses into one of the states, say state 1. When Bob measures the state of his particle immediately after, before any particle travelling at light speed could get there, it will also be in state 1 (assuming they were entangled in such a way that the state will be the same).
This mistreatment of the mathematical notation as a description of the present state of the system also leads to confusing language like "it collapses into one of the states" as if the change in a probability distribution represents a physical change to the system. The mental picture people say this often have is that the particle literally physically becomes the probability distribution prior to measuring it---the particle "spreads out" like a wave according to the probability amplitudes of the state vector---and when you measure the particle, this allows you to update the probabilities, and so they must interpret this as the wave physically contracting into an eigenvalue---it "collapses" like a house of cards.
But this is, again, overcomplicating things. The particle never spreads out like a wave and it never "collapses" back into a particle. The mathematical notation is just a way of capturing the likelihoods of the particle showing up in one state or the other, and when you measure what state it actually shows up in, then you can update your probabilities accordingly. For example, if you the coin is 50%/50% heads/tails and you observe it land on tails, you can update the probabilities to 0%/100% heads/tails because you know it landed on tails and not heads. Nothing "collapsed": you're just observing the actual outcome of the event you were predicting and updating your statistics accordingly.
Any time you do something to the particles on Earth, the ones on the Moon are affected also
The no-communication theorem already proves that manipulating one particle in an entangled pair has no impact at al on another. The proof uses the reduced density matrices of the particles which capture both their probabilities of showing up in a particular state as well as their coherence terms which capture their ability to exhibit interference effects. No change you can make to one particle in an entangled pair can possibly lead to an alteration of the reduced density matrix of the other particle.
I don't think solving the Schrodinger equation really gives you a good idea of why quantum mechanics is even interesting. You also shouldstudy very specific applications of it where it yields counterintuitive outcomes to see why it is interesting, such as in the GHZ experiment.
There is a strange phenomenon in academia of physicists so distraught over the fact that quantum mechanics is probabilistic that they invent a whole multiverse to get around it.
Let's say a photon hits a beam splitter and has a 25% chance of being reflected and a 75% chance of passing through. You could make this prediction deterministic if you claim the universe branches off into a grand multiverse where in 25% of the branches the photon is reflected and in 75% of the branches it passes through. The multiverse would branch off in this way with the same structure every single time, guaranteed.
Believe it or not, while they are a minority opinion, there are quite a few academics who unironically promote this idea just because they like that it restores determinism to the equations. One of them is David Deutsch who, to my knowledge, was the first to publish a paper arguing that he believed quantum computers delegate subtasks to branches of the multiverse.
It's just not true at all that the quantum chip gives any evidence for the multiverse, because believing in the multiverse does not make any new predictions. Everyone who proposes this multiverse view (called the Many-Worlds Interpretation) do not actually believe the other branches of the multiverse would actually be detectable. It is something purely philosophical in order to restore determinism, and so there is no test you could do to confirm it. If you believe the outcome of experiments are just random and there is one universe, you would also predict that we can build quantum computers, so the invention of quantum computers in no way proves a multiuverse.
You people have good luck with this? I haven't. I don't find that you can just "trick" people into believing in socialism by changing the words. The moment if becomes obvious you're criticizing free markets and the rich and advocating public ownership they will catch on.