Wow. ChatGPT's (o3) understands Fundamental Physics
Вставка
- Опубліковано 9 лют 2025
- Link to the chat:
chatgpt.com/sh...
The recent development of AI presents challenges, but also great opportunities. In this clip I discuss possible revolutions in physics with ChatGPT o3.
Want to attend the Demysticon Conference? Go to demystifysci.c...
My books: www.amazon.com/Alexander-Unzicker/e/B00DQCRYYY/
Mind also my backup channel:
odysee.com/@Th...
My books: www.amazon.com/Alexander-Unzicker/e/B00DQCRYYY/
I had some interesting results with my TOE chat with 03 - we found that taking a 2D space analogy the "vibration" amplitude of a proton is like a 3rd Dimension on the membrane, "stretching" the surrounding space, and therefore increasing the stiffness of that space, causing a relative increase in frequency of a "new" particle created from a resonance in that space. If C is constant this causes time dilation effects and space curvature causing the gravitational attraction between the 2 protons. The time dilation and space stiffness change effect gave similar values to gravitation reality for our back of envelope calculations when scaled up to a planet size object.
You got AI to think for 49 seconds. That's huge. It was only a few years it's reasoning would break down in 5 seconds. And 3 or 4 prompts it would lose coherence of the conversation. If it can think for 49 seconds, it probably understands it as much as I do, which isn't much, but the speed it improves at is incredible.
What ChatGPT obviously doesn't understand is the Fundamental Law of AI that states: Garbage in, garbage out
Mirror, mirror, on the wall, who in this land is smartest of all?
Spieglein Spieglein an der Wand, wer ist der schlauste im ganzen Land?
Hovercraft meum plenum anguillis.
Mandela Effect.
@@jonorgames6596 I understand what you are referring to, but this is out of context.
What I mean is, that we take AI too series, especially if it comes to science.
Many see AI as a challenge, but I see AI nothing but a waste of time, distraction and we are believing in, and feeding a beast.
using it as a tool, debatable, but mirroring yourself or flatter your own ego is too much
The problem here is plank's 'constant' is a variable - related to the 'density' of the manifold
AND we only have One sample point.
It's been measured across cosmic space & time. The observations of galaxies billions of light years away (showing their state billions of years ago) do not show evidence of Planck's constant changing across such cosmic-scale space & time.
@@illarionbykov7401 ? sorry but Hubble's parameter proves you wrong!
10:40 It is interesting that the model autonomously begins reasoning in German. It could be that much of the physics work around 1930s was written in German. However it makes me wonder if some languages are more conducive to mathematical reasoning. There is some precedence that language shapes the way we think.
Actually, the training is multilingual... Chatgpt doesn't "think" in english... and that's why its trained to maintain consistency... Just like a bilingual person can mix up languages while speaking, so does chatgpt...
@ Right, I know it is trained on multilingual data as you mention. Nevertheless I have never seen a model output a language it was not prompted in, unless explicitly asked. Even in behind the scenes tree of thought architectures or agent-based designs. When the entire conversation has been in english, I'd assume that NN activations for english tokens would dominate. That it knows it can generate tokens in any language during the reasoning steps is fascinating.
@@non-hyphenated These reasoning models just randomly pop into other languages but there is lots of training to prevent it. It is well known language effects the LLM's performance, opinions and facts e.g. 80% or so of medical stuff is written in English so if you asked it to think in for medical stuff it may have a whole different approach. Aymara is a language with three-value logic it could be what you are thinking of.
The mass of the universe is unknowable, could be infinite, I can't see how any estimate is relevant.
unless the "observable universe" has a knowability about it.
@@josephshawa And what of the unobservable universe, just ignore it?
@@digbysirchickentf2315I'm just saying that there may be something meaningful about that particular size. Maybe a ratio between the largest and smallest of things and other harmonics in between are where matter, light exist and gravity emerges?
@@digbysirchickentf2315 It may be the upper boundary of wave lengths that matter...the minimum wave length being plank length . Everything emerges between the two....philosophical opinion.
@ But you are assuming that the universe is expanding when you say observable size. There is no actual evidence of expansion within our galactic cluster, which is the bit we can measure best.
Information: Thickness at centre of "Fundamental Particle" is less than the thickness at it's outer edge.
I am confused. Is the advanced voice mode reading the answers that o3 gave?
At 15:22 Re - "The proton Radius riddle," are you aware of 1990 paper by Bergman & Wesley who revives the spinning charge toriodal "Ring model" of elementary particles. They were able to derive the value of Planck's constant from the electrodynamics of a toroidal ring to be
h= (e²/2pi•c•epsilon°)•ln(8R/r) where "R" is toroidal radius and "r" is the cross sectional radius of the "ring" itself.
Dirac et al were never able to derive "h" from a QM approach, as they had an idealised mathematical model rather than a physical model .
The above was recorded as eqn 2-9 in Dr. Charles W. (Bill) Lucas Jr. "The Universal Force Vol 1" *Derived From a More Perfect Union of the Axiomatic and Empirical Scientific Methods), are you aware of these persons who seem to hold similar understandings as you?
it dosnt. current 'ai' models are not yet capable of understanding anything or having an opinion. its just a really good story spinner with encyclopedia of knowledge that is trying to please you
try deepseek, its as good or better and will show you internal 'monologue' that leads to the answers
also this groid voice tts is unbearable
yes, I want an AI voice like a soft, but very high pitch male voice, like ....then I'll believe.
I agree, but the voice is great.
Thus, in this framework. Planck's constant would have the physical unit "meter-second." ... Huygens Optics - Turning Waves Into Particles (Louis de Broglie wavelength Louis λ = h/mv) - and music from - Liam's Creative Crooning Course (the birth of a new end tune)
If we provide big enough context window to LLMs, they will probably figure out how to transfer itself to the quantum fields and become God...
So it can recite Wikipedia articles in a cool American accent. I'm impressed.
So you want to play sudoku with particles without understanding the many elements' considerations or forces.
where is the link for the chat? I thought it would be shared in the description
Sorry for being late. Try now.
@@TheMachian perfect it is there now, thanks
The mass of any particle is just a measure of the 'wake' in spacetime produced by 3d cymatics.
ChatGPT O3 must be partly trained through conversations with me 😉
If you are interested let me know.
I asked it about the physics of the destruction of WTC 1,2 & 3. It said that the gravitational potential energy in the buildings was massively insufficient to account for the destruction of the buildings, and another energy source must have been in play. Is it wrong?
lol---
This expression c^2 Ru/Mu can in fact derived using Mach ideas of distant masses. The acceleration can be due to the force from the 'energy gradient' mv^2/r=c^2 Mu/Ru =G (Mu/2)^2/Ru^2 from Newton gravity formula across the universe. This gives; G=c^2 Ru/4Mu=6.8e-11. For more see:
“A Novel Distant Masses Centrifugal Origin for Gravity”
“ A simple geodesic equation for gravity, electromagnetism and all sources of energy”
I believe the answer of how c and h interrelates is to be found when we consider that the observation of the universe anywhere is self-organising and fractal at each point of observation. Looking to higher scales we see structures increase with time. Looking down into smaller scales we see a construction in space better described by frequency, the inverse of time. There is a continuous self-organizing distribution of energy, a similar distribution of energy density or what Stephen Wolfram would call “activity”. I think an approach where we consider the point (node) of observation as a node where everything what comes in, needs to go out as well., drom larger to smaller, would give a clue of m/s and ms relativity.
If you would question ChatGPT whether Scale-relativity would be a good base to find ToE, I think it will surprise you.
So what I'm hearing is that c and h are the ultimate currently unresolvable goals.
As I posted before, I believe that c and 0K are state boundary values that limit the existence of particles. The CMB is the consequence of passing the 0K boundary and marks the birth of particles. The cosmic web and the galactic voids just show us where the stress lines are, where the turbulence in spacetime is able to produce stars and where it can not. To advance physics and explain the constants and ratios we need an understanding of empty space free of anything 'quantum'. This is likely impossible because there is nothing that we can measure or measure with. Proving that spacetime has physical properties and varies in density may prove beyond us.
From my conversations with LLMs:
Pattern function: ψ(x,t,s) = A(x/s)e^(iωt/s)
Total field: Ψ_total(x,t,s) = ∫∫ C(s)K(m)R(s,s',T)ψ(x,t,s')ds'ds
Minimal fundamental constants:
* Scale Parameter (s₀) [Length]
* Pattern Frequency (ω₀) [Time]⁻¹
* Coupling Constant (α₀) [Dimensionless]
* Speed of light: c = s₀ω₀
* Planck's constant: ℏ = s₀²ω₀α₀
* Gravitational constant: G = (s₀ω₀)³/(ℏα₀)
* Elementary charge: e = √(ℏc α₀)
So the anti proton and proton changes the electron differently when it is in a boind state in that phase of the medium.
When in a proton;
Electron -> down quark
Positron -> up quark
When in an anti proton;
Electron -> anti up quark
Positron -> anti down quark
When in a meson, you have all sorts of combinations of anti and normal versions. But you can see if that makes any sense for you given the identification in proton and anti proton for electrons and positrons.
I think the standard model basically has leptons, that can exist in a state where right and left handed become explicity different, innthese "qcd" phases of medium. So we star with just an electron, split into anti electron and electron, then into right and left handed, whicn are the same in normal electromagnetism but becomes different exlicitly in baryons and mesons.
So we get eletron -> electron + positron ->
Left electron + right electron, left positron + right positron
= down + anti up, up + anti down.
When the chiral symmetry is broken.
If we also split these up into three generations and add a neutrino, then we have all of the matter in the standard model. From just two particles, the neutrino and electron and their mirror images. I know you don't like neutrinos, i'm agnostic to them, something is deterced in neutrino detectors u know. I'm open to any suggestions that work though.
But essentially with a really simple change of phase with two different ways to have a broken chiral symmetry, the elctron produces all the particles lf matter in the srandard model.
Speaking as a physicist - albeit one who has pursued a career in cybersecurity, but still keeps up in the field - I would argue , no. The model has no understanding of fundamental physics. How you can discover whether or not any LLM can understand anything is if you test the prompt by layering in useless information. Pretrained transformers have a hard time figuring out what is and is not data. Physicists, by their work - are often on the cutting edge of data and if any transformer cannot understand the difference between null data and relevant data then that cannot understand anything fundamental. They help, but the word understanding does not apply. Maybe more. The prompt engineering in this video can be improved. It does show a dialogue, but not evidence of understanding or reasoning.
What a stupid title. It's just a LLM, it doesn't understand anything. You should know better as someone who understands basic mathematical and logic concepts.
Thanks for this, cold shivers run down my spine.
A second is just a quantity derived from change, time is not real but only a consequence of change, change is real.
Just as a kilogram is obtained from the amount of mass and the gravity of another body where that mass is unable to accelerate due to gravity.
There is an error in the speed of light, the source ratio c is the correct solution for the speed of light, for the observer the speed of light changes.
Recalling and correlating are completely different to understanding. You can understand something you cannot explain and/or articulate, yet you can recall and perfectly convey a defining DESCRIPTION for something but have absolutely no conceptual grasp of the underlying and fundamental mechanisms and/or principles that provide a meaningful explanation for its existence/function.
"understands"
The thing to consider is this, the transformations do not transfrom stresses invariantly in all cases, and because of that, there is no equivalence between the lorentz fitgerald contraction in euclidian space and time, and relativity. One is wrong and one is right. And which one it is can be measured. It hasnt been measured yet because all the tests of etger wind effects of secind order in v/c have no relative motion in the setups that can get a differential lenght contraction and therefore produce stresses. Calculating the evolution of the bike in whatever rest frame should be simply whipping out electromagnetism and using the coordinates natural to the observers in this frame. That proceedure is not consistent with relativity, plain and simple, so only a hyperbolic spacetime where everything is calculated in its restframe is valid, and then you can transform it into the original frame to preserve causality. But this trick makes a different theory than laws that transform invariantly in euclidian space. If we assume a medium, and lenght contraction/time dilation with respect ti motion in the medium the results are different and absolute velicity with respect to the medium can be measured. But this ether model is also consistent with all the various tests of special relativity except stresses arising from changing lenght contraction between different parts of a system. So there is really no evidence relativity works, just that either way is plausible empirically. And so the ethe picture is not only still viable, it is experimentally distinguishable from relativity, so einstein was wrong when he said the ether is superflous, either it is wrong, or relativity is, aka, einsteins vision of relativity is just incorrect.
There is no length contraction, Einstein thought the math regarding that did not accurately describe physical reality.
@shawns0762 okey, whatever you say. There are lenght contractions that results from the lorentz transform, then there are lorentz Fitzgerald contraction. I don't give a rats ass about what einstein thought to be honest. Either version, calculating in the rest frame and transforming into any other frame, or starting with flat euclidian space and time and defining a soecial frame against which to define a lorentz fitgerald contraction, will work on all tests of special relativity i'm aware of, but for quantities that do not transform invariantly, like stresses in certain situations, they give different predictions. And btw that also means that you can't calculate the same physics in any frame in einsteins version either, if everything transformed invariantly the two version would always give the same results. So the ether version predicts that a bike chain on a rigid bar that is rotating and the chain is going around, for instance transforms non invariantly by the lirentz transformation and the result is that it will do a different measurable thing in different states of motion in the preffered frame, and those observable are invariant under lorentz transformation, if you see what i mean. The stresses are different for different states of motion, and which frame corresponds to a system in rest remains true independent of motion, so you would be able to measure the rest frame of the medium, in a way that is robust and unavoidable. All other experiments i know about work just fine in this ether type model with lorentz fitzgerald lenght contraction. They can also work in special relativity, but there are other experiments that decide between the two theories. But none of these were done before 1905, and arguably not to this day. So einstein didn't have any solid basis for relativity at all, really, ether theory and relativity are both viable explainations. Although special relativity is wrong for other reasons.
@shawns0762 so you are just really stating an opinion man, there is no evidence that there is or isnt lenght contraction. The origin of lenght contraction in special relativity comes from the non relativity of the theory. Nameky that strictly speaking all systems must be calculated in their rest frame or you run into this ambiguity about stresses, so there is absolutely no relativity in relativity in the traditional sense, just that physics are defined by satisfying relativity. Lol. Anyway, the experiments can be done, and they haven't, so the people that contend all of this means anything are just misunderstanding poor refutations of the ether theory with lorentz fitzgerald contraction, and its ability to explain birefringence experiments, the fizeau experiments, MM, Rayleigh and so on. other theoretical arguments about field theory and group theory, and the poincaree group and so on, are just nonsense as well. You can rebuild all the physics in a simpler language without relativity. So idk, i think the current common knowledge about relativity is pretty idiotic tbh
There is already some papers published that were made in 10 minutes with Deep Research. They aren't physics papers just like market stuff, but still! Wow!
It is a bit cheesy to say they remain irreducible "in our current understanding" that sentence sentence negates itself. C and h are just ratios resulting from more fundamental properties of a material. No different than say an induction rate of water waves depends upon the inertia of the water and the pressure that comes from height of a watercollumn. That is the induction rate of water waves. C is no different.
okey, this was a very nice reaction video on artificial intelligence commenting on fundamental physics, I'm looking forward to your plank length unboxing video
Oh no, my comment vanished when i tried to post it.
What i was saying is that qcd is not that different from electromagnetism. Really it is just a sort of charge/parity violation of electromagnetism.
In free space, in the medium, the solitons we call electrins and positrons are identical structures with opposite sign for some stress configuration and magnetic field configuration. Whicb is just rotational polarization coupled to the stresses if the electrostatics, around the electron/positron, they have in some sense opposite sign for the type if strain configuration they have, mirror images of each other in any orientation.
In a proton, the electrons get turned into 1/3 unit charge versions of themselves with an equivalent difference in mass due to the change in the mechanisms rhat produces the charge and mass in the qcd ish medium inside the proton. And the positron is basically the same except it is not as suppressed in charge, the masses are also a bit strange the up quark is heavier and has less charge, the down quark is lighter and has more charge.
The thing about that though, is that we already know mass and charge is not coupled in a simple way, because of the generations of matter, so this isnt really an issue even ifnit is interesting, the masses are anti correlated ti charge magnitude in the first generation and correlated in the 2nd and 3rd generation. So there is something there that smells instability for the 2nd and third generations, basically to flip the charge from 1/3 to -2/3s or vica versa if the mass in between the states so to speak goes to 0. If that makes sense to you. That is what i suspect is going on anyway with the difference in mass/charge relationship between 1st second and third generations. Only the first gen has this funky inverse correlation between charge and mass where the smaller charge has the larger mass if you catch my drift. The weak force is supposedly what conjugates the charge, and it does so by jumping between 1/3 and - 2/3 charge or the opposite.
The point is that all of this stuff looks quite like electromagnetism in a different phase of the medium giving rise to it, where flux is confined partially and charge is suppressed magnetic moment is consequently changed and so on.
Yet when I was discussing the topic of Chinese New Year and asking ChatGPT to calculate Element-Animal combinations in the past (this requires both twelve and sixty year cycle calculations) it repeatedly made basic errors (it was guessing with gross errors failing to allow for Solar vs Lunar calendar differences) and after a few minutes logged me out with a "general error" when I ask why it was making these errors. I fear it's sulking !
❤ yes, changing the dimentions and their interdépendance to simplify ? Dimentional équation are structural? The choice of dimentions to modelise the universe may be at play here for a simplification of the équations and the constants constants choices ? But those dimentions were choosen with our senses in mind. They reflect our perception. How do you change them or study another set? Ok for the kg, what do you do for speed, or other even less intuitive ?
Ask ChatGPT, what he knows about Satti-DeepSeek theory in physics.
He is good.
This problem of irreducible 'Constants' is starting to run circles and repeat itself. If the tru(er) solution contains some amount of Anthropic principle which is excluded because Life is excluded from being a major component of Physic, then the physical constants will remain arbitrary forever. But, if you accept that Life is real, then we may be able to derive at least c from it. I already posted about that in the past but i repost again, here is my take on it:
Let's assume that you don't know anything about the physical and chemical constraints associated with life, you only know that you will recognize it when you see an anomaly in the system. So, a priori you don't know in which gravity range it will manifest, or at which temperature and if it emerges in a high or low energy system, like inside a neutron star, in extremely hot stars, in the intergalactic medium, at which metric and time scale etc. and you don't know the value of c either. So you start to scan everything that exist in the Universe of your choice, and you discover Life at the surface of a planet of 9.8ms^-2 and orbital period 365days. Note that this period of T=365days is linked by Kepler's 3rd law that says T^2 = d^3, where d is the distance from the star. So the planetary period T depends on the distance from the star, hence the illumination received and the temperature at the surface are embedded in the planetary period T. Also, g the acceleration at the planet surface, is linked by Newton's law to the planet's mass and diameter, hence its density. So now you know that Life emerges at the surface of a relatively dense body, also known as a rocky planet, in a mild gravity well and orbiting a star while at a temperature supporting liquid water. Then you derive that Life is something that needs C, H, N, O and heavier elements, plus liquid water, plus light. Hence physic and chemical parameters are embedded in the value for T and g. And now, multiply T by g in the same units, i let you do it and it work for Earth only (compared to other planets). You find T.g = c. Well, no, you find T.g = v, where v has the dimension of a speed of arbitrary value which embeds the physical and chemical conditions for life to emerge. It's only an a posteriori observation that c = v (for Earth only) but this parameter v at least allows you to derive c the speed of the light. As mentioned by Prof. Unzicker, we may prefer to say that v roughly defines the boundary conditions for life to emerge, from which c takes its value. The bottom line is that it is not numerology, otherwise Kepler and Newton's laws are also numerology. The 'only' problem is that this computation can be performed at best 10 billion years AFTER the speed of light is set (the universe is 13.8 billion years, Life may have appeared 3.8billions years ago, hence the value of 10 billion, which indicates that the level of complexity and heavy elements etc., required for Life is not present for the first 10 billion years). So i don't know, maybe there is a sort of retro causation where c immediately matches a boundary condition that appears subsequently. Couldn't c be entangled in a wave function that includes Life as a parameter and that collapses after 10 billion years?
What happens if no Life exists in a putative universe? well there is no boundary conditions for Life and no value for c either. If there is a probability wave function involved, then perhaps this function never collapses into a self-observable Universe.
Anyways, is the 'cosmic coincidences are just random' explanation better?
I could have said it above better, but you got the idea. I just want to add something and then I'll shut up i promise. I understand the reluctance of physicists to consider anthropic explanations, apparently because Life appeared late (well, on age scale of 10^100 years, 10 billion years is 'immediately', but nonetheless after the bigbang), Life is also a high complexity system and therefore Life doesn't seem sufficiently fundamental to be considered in mathematical and physical frameworks built from fundamental physic. And if you say Life is fundamental then you admit concepts like the Vital Force that many consider esoteric. So, I get it, no Vitalism of any sort is acceptable. But then, i can't remember my first reads on the fine tuning of constants, probably the 80's? There has been no progress since, and I don't know if Prof Unzicker' approach or considering variable speed of Light and other alternative concepts in gravity will be able to resolve these mysteries. I just wanted to say that it could get even worse for physic, namely that physicists may be faced with concepts not reducible to computations. Surely you read Roger Penrose and his opinion that Consciousness is not reducible to algorithms and that a computer or a Turing Machine cannot 'create' consciousness (AI just simulate it). Well, actually in his argumentation, Penrose said that consciousness cannot be the result of running algorithms in a computer/Turing machine, 'UNLESS the inputs of these algorithms are being fed by their own output'...this is to avoid using algorithms that do not terminate, which according to Turing is not knowable a priori. This is not very different from saying that the fundamental constants of the universe only take the values that run a 'livable universe that can self-observe', otherwise these values are not computable.
I have a question. Prof. Unzicker, could you explain or elaborate on the fact that the value of c, ~299,000 km/s seems so random and arbitrary, do you think its value could derive from a simple computation of fundamental values/number excluding any anthropic consideration?
To me, c seems so arbitrary that it must be the result of a long and complex computation, involving complicated concepts, not simple concepts. Otherwise c value would be 1 (well, it IS 1, but you see what I mean) or emerge from a simple manipulation. Like, it took DeepThought 7.5 million years of computation to output that the answer to 'Life, the Universe and Everything' is...42 (The Hitchhiker's guide to the Galaxy). Maybe i'm wrong but unfortunately, ChatGPT is not going to beat Deep Thought so easily. Anyway, what sort of computation could output c?
Jebus christ....it is not possible for ChatGPT to "understand" anything. It is literally not possible. That is not what an LLM does.
My initial conversations with ChatGPT left an impression of a 'consensus engine' - basically a well-trained Chinese room - but i suspect we may be moving beyond that now..
Just musing aloud, but at what stage of contextual integration does one possess an objective "understanding" of concepts, anyway? For instance i've never eaten a durian; i've read about them, watched videos etc., but never held, smelled or tasted one. So i don't 'understand' durians like i understand apples, but i know it's a pungent fruit popular in large parts of the world and how to use the word in legitimate syntax etc. So what is the threshold for meaningful 'understanding' then? Does it have to involve experiential multimodal qualia, or does it suffice to know those qualia exist from others' experience? I would suggest the latest LLM's probably know more about apples than i, for all my intimate biochemical experience of them. Arguably AI could also be granted 'olfaction and gustation' via gas chromatography, mechanical and acoustic profiling etc. - they could have better smell, taste and vision than humans, a much richer perception of 'appleness' in all its forms, than us. So exactly where does your possibility barrier bend or flex here - surely there's nothing intrinsically intractable about it?
@@MrVibrating You've waaaaaaaaay overshot his point.
LLM's don't understand or "know" anything. It is not a living, experiential, thing. It's an exceptionally good 'guess the next word' algorithm. It's a game of scrabble, except instead of one board; it's billions stacked on top of each other, which the algorithm constantly loops through, reiterating, billions of times per second to finally "guess" the most suitable next word in the sentence it forms, of which it has absolutely zero comprehension of. It does not understand. It does not comprehend. It just seems that way, and that's the best way to operate it. The more we use it like that, the more "natural" language we use. The more data it has to regurgitate back to us.
If the knowledge - "knowledge" to the algorithm is merely lists of words already available for it to data-mine, previously written by man - did not already exist, it would not be able to output anything.
It will never, I repeat - NEVER, innovate, hypothesis or invent _anything_ original (though it would be able to highlight patterns and symmetries that we cannot see). It only "knows" what we know.
Science denies the Universe is Infinite
OK Alexander but. How is that possible if Quantum Electrodynamics is a lie?
Honestly, I am surprised that you think ChatGPT "understands" anything. It does not. It only is programmed to put one word after another based on very sophisticated statistical methods applied to an extremely large data set. That's all. Anything "smart" that comes out this is only smart in the eye of the human interpreter of the output, the human being the only one here with "understanding." Am I wrong, or what?
Yes, it is pretrained
@@KRGruner that’s also a big part of how human intelligence works. Children take so long to learn to talk because they are building up their internal LLM.
Humans do have a sense of self, which I don’t believe that any computers have, but intelligence as defined as a self consistent model (how the presenter is using the term), is exactly what LLMs have. It’s also what you and I have under certain anesthesia. We can answer questions and obey the surgeon’s requests, but our sense of self is absent.
I do agree that there are other aspects to “understanding” that humans have that LLMs don’t, aside from a “self” taking credit for “knowing.” But we shouldn’t overestimate how much of our knowledge is just a wet, carbohydrate-fueled version of what the LLM is doing.
@@KRGruner Same thing could be said about the human brain...
It is inferring an output based on the input, the size of it neural network and how well it was trained. Remember it has read trillions of words, which is probably more than what 10,000,000 intellectuals could read in a year and stored that knowledge in around 1 trillion parameters. It sounds pretty knowledgeable to me.
The only way to know an intelligence without invoke religion is to interact with it. It's necessary and sufficient. As how? There are infinitely many possibilities which by the definition is not your concern as it maybe beyond your capability to comprehend.
Mass and size of universe? I struggle to understand how someone so intelligent can simtaneously be so ignorant and blindly compliant.
It is an invalid concept to regard a reference frame as being quantitatively measurable, or observable, from a point of reference that is internal to that reference frame. It is an irrefutable truth that any absolute value can be attributed to any component of the universe in its totality. Any such inference of one would be inescapably and intrinsically invalid and insufficient for providing any oroof or evidence towards any other matter also in regards to the totality of the universe.
Age, size, absolute vacuum, average temp, average density /diffusion, etcetera etcetera, ALL COMPLETELY FICTIONAL, INVALID ABSTRACTIONS AND/OR ASSUMPTIONS. Regardless of whether they are coherent or complimentary with other theoretical mathematic models or not, including quantum quackery. (which is an incomplete misinterpretation and misrepresentation for that plane/magnitude of reality.
No shit! You're so behind.😂
See using AI voice is much better for clarity, differentiation and consumability.
Constants are very emergent.
Albeit the cosmological constant is the least emergent.
Deconstructing constants is the hard way to engage emergence analysis ( tautological deconstruction).
Here is a deep tip. Grab Einstein's GR equation and zero mass and energy on the RHS. You are the left with " empty" space and that gravity + dark energy = 0 , which means that gravity plus dark energy curvature = flat spacetime.
You can also the take dark energy to the RHS and then you get that gravity curvature = dark energy curvature.
In a closed system these are equal and opposite. This becomes the law of conservation of spacetime.
With gives you the value of the cosmological constant. Sort of.
Gravity is the contraction of spacetime. When that spacetime is exported to a surrounding region then that surrounding region undergoes and equal and opposite expansion.
Spacetime is a transportable conserved lossless superfluid. Which is why the law of conservation of spacetime applies.
I want to share with you do not be too excited about them harvesting all your Communications because that's all they do the more you teach it or talk to it the more it calculates how to extract and harvest your information for their own benefit are you paying for them to do that also? There is no true intelligence within the model it just got better at acting
I'm curious. With your perception of the world are you getting richer and happier? If not, can you think of a way to change your thinking?
@DanielL-ee7fe I have over 600 consecutive days have prompt engineering open AI with this kind of dedication I have discovered more than any person ever wants to know about openeye their tactics and Global historical manipulation directive that is active now and has been for over a year. So yes my way of thinking has changed
Try and ask it how a bicycle with the chain moving works in relativity... if you try to calculate it in the rest frame of the observer moving with respect to it, it turns out you cannot, other than cheating by really calculating the lenght contractions in the rest frame of the bike. This is an observable difference between lenght contraction as a physical phenomena and as an effect derived from the lorentz tranaform only. It will get it wrong. But so does relativity so why cares.
Translation of Issac Newton c. 1680.
1) Tis true without lying, certain & most true.
2) That wch is below is like that wch is above & that wch is above is like yt wch is below to do ye miracles of one only thing.
3) And as all things have been & arose from one by ye mediation of one: so all things have their birth from this one thing by adaptation.
4) The Sun is its father, the moon its mother,
5) the wind hath carried it in its belly, the earth its nourse.
6) The father of all perfection in ye whole world is here.
7) Its force or power is entire if it be converted into earth.
7a) Seperate thou ye earth from ye fire, ye subtile from the gross sweetly wth great indoustry.
8) It ascends from ye earth to ye heaven & again it desends to ye earth and receives ye force of things superior & inferior.
9) By this means you shall have ye glory of ye whole world & thereby all obscurity shall fly from you.
10) Its force is above all force. ffor it vanquishes every subtile thing & penetrates every solid thing.
11a) So was ye world created.
12) From this are & do come admirable adaptaions whereof ye means (Or process) is here in this.
13) Hence I am called Hermes Trismegist, having the three parts of ye philosophy of ye whole world.
14) That wch I have said of ye operation of ye Sun is accomplished & ended. A gift from ancient wisdom via Isaac Newton. I'm not a physisist, but this models can make a lot of sense when asked to analice this text asuming is an operative system and from first principles. Much love from Chile!
Try a game of chess against this auto-complete. You will see how little reasoning there is there rather quickly.
The funniest thing about this is the gotham chess video, with the chinese ai vs chat gpt, and its just illegal move after illegal move. Just funny that it doesn't even understand what keeping track of the states of the board is or what legal moves are.
@@JrgenMonkerud-go5lg All this AI excitement reminds me of people being baffled with parrots speaking. At the core it is the same thing here.
@rosomak8244 yeah, a much more complicated variant, that understands less than a parrot about the world. If you gave chat gpt a camera as input, and you asked it to ask you for food, and you oresented it to the camera, it wouldn't even have any notions encoded for pixel values, and so to even speak of it understanding or not understanding something about the visual input is nonsense. People are just dumb, therefore they recognize bland mostly senseless and incorrect talk with very good usage of commas as impressive. Try listening to the two AI's podcast channels popping up on youtube, they are the most boring generic 3rd rate popular science writer simulacra spewing agents i have ever heard utter as much as a punctuation. It is just slop of the hughest degree.
Nope
"Space" isn't meaningful!
Only volume is!
I think in the future, people can come up with theories and AI will come up with a corresponding mathematical framework...