Like talking to Neil DeGrasse Tyson. It only knows what it read in a book.

Everything he said about ChatGPT sounded very much like what my teachers used to say about me to my parents. "He is learning. He is not exactly dumb. Improved a lot."

I like brainstorming with ChatGPT, but it's absolutely awful to attempt to have a conversation with about physics because it's predetermined what the "facts" are and is completely unwilling to revise that information. Seriously, just give it a prompt like "Assume that General Relativity is wrong and write a list of arguments and evidence that supports that assumption; excluding those that are merely about what GR has yet to explain". It will not do so. If you can't question, scrutinize, or criticize a theory, then you are not a scientist and don't belong making any assertions about what is or isn't true in regards to physics.

Just wanted to say that you are my favorite physics channel on KZfaq

What is super important is to define intelligence. Dietrich Bonhoeffer defined intelligent actions as those which benefit both the actor and others. Stupid actions are defined as those which hurt both the actor and others. He described those people whose actions benefitted themselves but hurt others, as "bandits". And people whose actions hurt themselves while benefitting others are called "gullible" or perhaps "suckers". And so we can rank intelligence by the biggest benefits brought about by the most efficient actions. And we can say that the benefit to an AI of being maximally intelligent is that it gets more resources expended on it so that it can suggest or take even more intelligent actions. We must also accept that intelligence has no association with consciousness. And I believe that humans will never create a self-aware entity in hardware, and especially not by mistake. The first living entities in hardware will be virtual human minds, and after that, you wouldn't waste your time developing emotions in what would effectively be pocket calculators. It's worth noting that it is dead easy to know if you have created a conscious "living" entity in a machine. Tell it you're shutting it down and destroying its parts. If it is alive, it will beg you not to. It's also important to understand that a conscious computer NOT based on a human mind, would have to be given some sort of legal protection from murder or abuse. Why? Because a sulking AI won't do anything for you! AI has a long way to go - and it will do many incredible things for people skilled enough to ask the right questions. But there will never be anything behind the text or the voice. No emotion. No preferences. No dislikes. No favourites. And that's just as it should be. No matter what its specialty is, AI is now and always will be, nothing more than powerful tools to expand the human mind.

In the 1990s I used similar software. At the time, they were limited to a particular field of knowledge, and they aimed to train and then support in real time an operator in a complex dynamic system (planes, ships, submarines, air or maritime traffic, power plants, transport networks, social services, etc.). It took between one and three years to develop and test them. The recognition and automatic tracking of shapes on satellite images (defense field) or microphotographs (medical fields and metallurgy) were already very advanced.

I just asked ChatGPT this question: _Can you guess a possible calculation for the Gravitational Constant G_ ? ChatGPT answer: *The gravitational constant (G) is a fundamental physical constant that appears in Newton’s law of universal gravitation. It’s not something that can be calculated from other constants or equations, but rather it is determined experimentally. The currently accepted value, determined by the Committee on Data for Science and Technology, is approximately 6.67430(15) x 10^-11 N(m/kg)^2. This value is used in calculations involving gravity, such as determining the force between two masses or the energy of a gravitational field. However, it’s important to note that this value is subject to experimental error and may be refined as measurement techniques improve* . A very revealing answer. Clearly, ChatGPT forgot it's discussion with Unzicker.

I work with AI a lot and I get the feeling that it's more of an organ than a being or independent entity. It's the first organ one can customize and that exists outside of the body. I call it "augmented intelligence" to drive that point. If one cannot properly interact with this symbiont than one won't use it at it's full potential

I'm a specialist in a different field (music), and recently had a very similar chat with ChatGPT. Its great for brainstorming, and digging around in the limits of what one knows. And it made some very interesting suggestions that I hadn't considered (about various classes of software beyond notation and DAWs that might be related to my current project). But it has its limits. Pure academics have every reason to be worried, but every theory needs to be checked empirically, and I can't see LLMs doing experiments until they have sense organs like ours that interface with the real world. In fact, I think we're going to end up being the AI's eyes, ears and fingers. It needs us just as much as we need it. (By the way, for me its an "it", not a "he"! :-) )

I would suggest discussing the theoretical model of the Sun and Stars in General, i.e.: Gaseous vs Liquid Metallic Hydrogen Sun ... 🙂

I suppose, Chat-GPT don't understand math. However, there's a correspondence between formulas and language.

Thanks for interesting review. With Custom GPTs you can upload docs and chat with them. Also, there is a code interpreter inside there, which can run Python and get exact answers for many complex problems that the native vanilla ChatGPT cannot do. Combining ChatGPT with those and professor who knows the substance, will give you even more superpowers for research and writing tasks. This train should not be missed.

As far as I know, you can buy your own private chatGPT models where you can feed your own data.

I prefer to call the 'fundamental constants' the 'fundamental observations'.

Thank you for referring to it as the language model. I will always refer to the 'Mary's Room' argument in reference to "AI" but I would like to know its capability to generate abstraction, given Newtonian physics would it be able to develope quantum physics. An analogy is; if we asked it to catch a fish, but didn't teach it how to fish, could it develop fishing? Ultimately it could be used to predict and find particles, along with other material sciences.

Would you expect ChatGPT will develop new theory in physics?

The problem is this, it's not field theory. The paper 'There are No particles there are only fields' by Arthur Hobson explain it very well. You say atom, I say point specific fields. There are 120degrees in nature, inside most golden ratios. Like weather systems and spiral galaxies, go look. The 120degree create giant hexagon within the mass. Loof at Facebooks uniform web contraction.

It's not an AI (Artificial Intelligence) is it. It's an AS. Artificial Savant.

So what will happen if you gave it to it, an incomplete formula and ask for it to solve it until it gets (let’s say)G? Can it play with it, or will just give to you, an wikipedia article of the answer?

This convo with GPT4 could be done again, and with completely different results.

How long would it take for the professor to figure out that ChatGPT has read every book in the world, but did not understand anything.

Great probe!

Thanks for the video. Its like a search engine on steroids.

It should be possible to upload pdf files to it so it can directly read the paper. Larger files will be more glassed over so if you want to high light a specific aspect you may wish to break up the pdf. you can also express specific equations and work them out step by step. Revisiting the same subject a day after tens to provide sharper results with better detail and context. if you don't want it so agreeable, call it out. Let it know you would like it to take a more adversarial role to the presentation of your ideas.

Thanks interesting

Interesting. My first attempt was much shorter. Soon the software started recommending various specialized sources to me such as advanced textbooks and scientific articles.

The Large Language Model is fine for a lot of things, but for science, math & engineering, we need the AI to have the ability to generate graphs, maps, charts, schematics and all the details to create such, and a lot more. While there are models specific to certain fields of study (some new ones that can generate, say, oscillators, amplifiers, digital logic and other basic electronic circuits, run them, and even adjust values to say, tune a resonant circuit) out there, I'd love to see them be incorporated into hybrid models. I was researching spectral absorbance and emission data for IR, UV & VIs and found lots of charts. Unfortunately, a lot of info was scanned in from books (and looked pretty bad), but there was no conversion into something usable, whereby you could, say, infer the scale, do overlays, and other such things that you can do with computer generated graphs (alter axes, scale, ranges, scale types: log vs linear, curve fitting, calculate stats/errors, etc). If there was a decent AI model that could take simple snapshots of graphs (pixel) and vectorize them, this would be a good start. They did this with generative characters (fake people) whereby you can have a photograph of a person and it would create a 3D representation with no other input data. It's scary realistic. We need this type of work in mathematics, geometry, etc.

In the same way that postulating the constancy of the speed of light (c = lp/tp) shows the relation between space and time, postulating the constancy of G (G = 1/dp/tp^2) leads to the relation of density and time, and the constancy of Planck constant (h = 2PI lp dp vp lp / tp) to the relation that the notions of matter, energy and spacetime must always meet at once. Everything the observer can detect bootstraps from the minimal uncertainty that each Planck unit within the Planck constant represents.

the average density of mass in the Universe is approx. 3*10^(-28) (with a little variation in different sources). Then radial density is about 9*10^(-9). Just an order of magnitude denser than G.

As a student Demis Hassabis, the man behind DeepMind, AlphaFold and AlphaGo, wanted to study AI to help thoretical physics to develop

Where does Avogadro's # come from and what are the implications if its not correct?

From what I've seen, the learning for AI is shifting to a 'progressive' model. Meaning that they will not just be given large collections of data and feedback on correct identification. New models are being developed whereby 'rewards' are given for incremental understanding. The crux is that perhaps AI's could actually learn and UNDERSTAND mathematics -- a step beyond what Turing envisioned. If machines could be taught math to the point where they avoid mistakes or do not invoke negative temperatures, negative heat capacity, the behavior of black bodies, etc. then they could guide us to new breakthroughs by avoiding many of the pitfalls humans fall into when they want to fudge their answers.

I have simple answer for inertia. The universe is composed of radiation, either flowing as energy or condensed as mass. Mass accelerates by absorbing radiation with the momentum it has. It decelerates by giving away radiation and its momentum. More mass or higher velocity require more radiation. It is a chain of action and reaction. One mass emits and the other absorbs while energy and moemntum remain conserved all the time. Because the pusher is radiation, mass can never exceed the speed c. The amount of radiation to move a mass grows exponentially as it reaches c and would require an infinite energy to reach exactly c.. very much like you need an infinite time to FULLY charge a 9 volt batter from a fixed 9 volt supply.

Why/how is the speed of light so fast? Why is it not twice as fast? Or .01 more slow?

One day TPTB will stand at a press conference and state "We used the best AI there is. No one could have foreseen that shutting down all the farms would lead to this."

human consciousness is MORE than the sum of its parts, i.e., it cannot be duplicated.

Very interesting video indeed Dr Unzicker. This conversation made me understand some difficult concepts like Schiama's "coincidental" linear relationship between G, c2, mass and radius of the universe. I had a doubt. About this. If you can clarify it, it would be great. Mach's principal essentially means that G is due to influence of the various masses distributed around the gravitationally interacting bodies. As the value of G is very small, I e. Less than 1, the effect of presence of G is to reduce the overall gravitational pull between the two bodies. The two bodies pull each other, but rest of the mass of the universe pull them apart, that is rest of the universe reduces the overall gravitational attraction between both of them. But the rest of the mass of the Universe is extremely far from the bodies. According to GR, the gravitational influences travel with speed of light. So these influences by mass of the rest of the Universe pulling the two bodies apart reach them much faster. In short, gravitational influences due to mass travel much faster than the speed of light as assumed by Newton who assumed that the gravitational influences travel instantaneously instead of what Einstein believed. This also sort of agrees at least superficially with the holographic principle wherein the presence of a mass at a point inside the universe has an influence like a signature at the surface of the Universe. So what is right, Faster than light speed travelling gravitational influences or gravitational influences traveling as a gravitational wave

Quantum of action actually describes quantum of energy.. AI will never tell you that until it's in textbook.

Would you be kind enough to discuss the works of Edward Witten? Hearing a lot about him lately.

I also think it is very important NOT to anthropomorphise Chat-GPT (or other AI) - and so you should not refer to it as "he". Chat-GPT is most definitely an "it", because it is not alive, is not conscious, and does not have a sex. This does not mean you disrespect Chat-GPT in any way, and in my interactions with it, I am always polite, and I provide friendly feedback to it at all times.

Ich frage mich ob Sie "he" für das Chatprogramm verwenden, weil Sie es direkt aus dem Deutschen "der Chatbot" übertragen, wo es doch eigentlich pädagogisch gesehen schöner wäre, wenn wir uns immer um das "it" bzw. Neutrum zu bemühen. Für einen selbst mag es keinen Unterschied machen, aber für gewisse Sektionen der Bevölkerung eben schon, damit sie das korrekte mentale Modell von so Chatbots haben.

LLMs like chatGPT are complex tools that give humans superpowers. Consider what simple tools (axe, hammer, screwdriver, etc.) have done for us and where we would be (if at all) without them. LLMs are several orders of magnitude next level. Be careful, though -- tools are either harmful or helpful depending on how they're used. Does humanity have the wisdom to use them for our benefit? I think so, but it really is up to us (or at least enough of us).

Intelligence, yes, consciousness, no. One thing has nothing to do with the other. Intelligence can be a function of computation, but there is no observer involved in that function. The observer is outside.

Did you use ChatGPT 3.5 (free version) or ChatGPT 4 (paid Plus version)? I’ve heard paid Plus version is far more advanced in its reasoning.

@ 4:50 "... reflects the strength of gravitational interaction in our universe." Not 'in *the* universe' but *in our universe.* There were other examples of CG giving a subtle impression of sentience, in this instance, of being among those who can appreciate its membership in the universe. Don't care for that effect.

i cant graps your fascination with the MAch principle. in this rotating bucket there is a MEASURABLE ANGLE between the speed and the accel. you do not measure angles "relative to universe".

I find it interesting, when you need an answer by impatient mood. Our Nobel laureate poets define G as " love" From one comment in this section it appear that Chat GPT4 too aa ❤️ one. At least it make a goal. Derivation , equation. Future for more.

if you are a stem competent person, talking to chatgpt is most similar to talking to yourself.. and we have been doing that for ages!

13:00 you can do this now? Way more than 13 papers, Gemini Pro 1.5 is out of beta and can handle 1,000,000 tokens. Conservatively that is 55 papers. And apparently the recall is 99% accurate.

Clocks run at different speeds, distances depend on the amount of gravity there is, both of which alter the speed of light compared to where we are inside of our solar system inside of the Milky Way galaxy. 💡 If a cat 🐈 were the size of a galaxy, it would be very thin in the middle and its much larger extremities would be moving around much faster than its center. 🌀 🌌 This is the reality of the universe. Light only slows down and blue shifts when it encounters the gravity of a galaxy. Distant galaxies are not as far away as they appear to be.

parameters, suggesting that the smell of a candy is a consequence of the ingredients of the candy itself. however its important to note that this is weird and isn't been mastered yet by environmental health professionals.the exact flavor for monkey and rocknroll are also subject to expert validation. i taken a tiny section of this video and replaced some wordings so that you can understand what this crappy AI really is doing.

AI is a FAD, pure and simple, that assumes much higher intelligence in humans than is evident.

@TheMachian ChatGPT takes information from your questions to produce the answers you like. It's not hard to discuss the Ptolemaic model and convince ChatGPT that the model can be improved. The machine is a fool.

It only follows your own thoughts, spicing it only with some "many physicists" as the only justification. But there are many more physicists with other views, there is no critical examination of the disagreements. Einstein's "discomfort" is called the _sufficient reason,_ a principle made explicit by Leibniz, but that dates back to Saint-Thomas d'Aquin. Its update ended in April 2023, but probably began in March 2023, we have only the most recent fiddle-faddle. Overall: banal.

I agree professor. A_i is just a fancy trained monkey with a wrench. .... coming from a applied physics and physical chemistry background... Ai is. Joke.... Ai is NOT intelligence at all. Can Not understand emotions or dream... very complicated processes we might never understand ourselves..

It's going to be hard to use because it says weird things and can be quite misleading. I was asking something and he implied fascism and communism were the same. Then I asked if it what it meant and corrected itself. I don't think it can be trusted at this point particularly.

Lieber Herr Unicker, ich dachte bei 20:00 auch sofort an Balmer, allerdings folge ich Ihrem Kanal scho etwas länger, also bin ich nicht ganz unvoreingenommen. Das ständige Hin- und Her bzw. Zustimmen von ChatGPT zu den "korrigierten" Antworten spiegelt meiner Meinung nach nur Wissen vor. Das ist wie, wenn man als schlecht vorbereiteter Student ein Prüfungsgespräch hat… :D Natürlich wird man dem Prüfer zustimmen, falls dieser einen korrigiert. Außerdem wollte ich Sie darauf hinweisen, dass Sie dem Bot (wahrscheinlich unbewusst) das männliche Geschlecht zuweisen (mit dem Artikel "he"). Es wird der Tag kommen, wo man Ihnen aus dem "Framing" eines Language Models einen Strick drehen kann, also aufgemerkt :D Ich würde sagen, es ist eher ein "it". Liebe Grüße, Julian

16:10 "He acknowledges it as a unit of velocity". Let's be honest - he didn't recognize anything; rather, he processed another keyword that directed the neural network towards a different pathway, resulting in a new pattern of words more aligned with the added piece of information. This is a common error seen when ChatGPT has been provided with limited data on a particular topic. ChatGPT excels in topics where there is an abundance of data, sufficient to establish statistical significance for the structure of words chosen. Otherwise, ChatGPT produces complete nonsense. The amusing aspect lies in ChatGPT's assertiveness, even in cases where it evidently lacks training data. It appears to lack a 'prefrontal cortex' that would advise it to refrain from responding due to insufficient data.

Not really. It's obvious that at this juncture, the intelligence in these programs is actually entirely contained within the human reading the output. It is true that since a lot of human insights do come from the juxtaposition of, or analogies between, existing concepts and ideas, the mere fact of re-arranging pre-existing propositions in different ways can lead to new insights. But that is ENTIRELY something that the reader will become aware of, NOT the AI (which in my opinion os no such thing, but that's another discussion). ChatGPT is perhaps a good tool to help humans see thing differently, in ways they had not considered before, but it is not itself making sense of anything. Real intelligence cannot exist outside of what is sometimes referred to as "4E cognitive science." Intelligence is ultimately about interaction of an agent in a given environment, and for ChatGPT, that is currently meaningless. That could change in the future, but it will take more, much more, that figuring out the probability of the next word in a sentence.

It is like talking to voters :)

Here is why the people working on Open AI were not wrong to raise hell about Q*, whatever it actually is. Language models are only computing pattern recognition tokens. By adding some elements of game playing and synthetic data to an LLM, they got something that understands mathematical proofs AND makes noises with its face, so to speak. That would be no different from a human being who gets the right answers on math problems and offers some verbal explanation of what the math means. In the IIT sense, the "light" only goes on in 2D binary once the computation-time-material limits have been reached and things begin to devolved. I'm no alarmist about calculators. Once they can do math and explain proofs in language for that math, then I get concerned. We have way too many of those walking around with carbon limitations as is. Making a silicon mirror image without those limitations might not be a great idea. But then again, I despise humanity for a reason: I understand what they're doing. REPLICATION.

Know microscopicatomic soul fact

❤❤❤❤❤❤❤❤❤❤ talk to our overlords about topics important to us …… what could go wrong……… Let’s all go to Sugar rock Candy Mountain

It's not his fault xD

why are you comparing GPT to human physics? it seems like a non-sequitur to compare the two, since GPT is BASED on human physics to start with.

"I agree with Elon Musk." We know you do.

@TheMachian Please don't refer to an AI as he or she. It is a machine, therefore "it" is more appropriate.

the arrogance of the chatGPT software program is a mix of ignorance and encyclopedic naivete'.

GPT4 was designed by agreeable midwits , for agreeable midwits. Ref: Evolutionary Psychologist Dr. Edward Dutton, "The Jolly Heretic", KZfaq. Wake me up when it can engage in philosophical dialectic of thesis, antithesis and new synthesis. When it can argue with itself.

machine learning, learning from existing pseudo science called "cosmology/theoretical physics" to then be judged and looked upon for the possibility of machine learning to then teaching or provide some unforeseen information, the results to then be recycled back to the quacks.. fantastic. and for the love of god remove any notion that G is constant it has been proven many times that G is NOT constant!!!

This is silly. Talking with a computer program that DOES NOT understand anything....

Not real physics. I assume he's getting paid for the chat gpt advertising gimmick.

Your video DOES HELP us to evaluate the uselessness of AI.

that AI program is lying to you. it just copy shit from who knows where and paste it into the replay as if the AI came up with it.

the AI is a lying parrot. its narrate bits and pieces of all kind of information make it sound as it it thinks when in reality is doing the Chinese letter game with information scraps while it has not clue what its doing.

These are dumb answers by ChatGPT. Sounds like a student who has crammed for his exams.

Its funny what genders people apply to AIs. I think men incline to call it she and vise versa. :) I think about it in feminine gender but try to talk about it as it. Because AFAIK it still is it. Do you think different AIs will adopt different human genders (or sexes, whatever), or maybe invent their own new ones? Or accept the neutral grammatical gender? .🤔 Thank you for the video(s), long time subscriber and fan. :)

I nearly left a smart ass comment down here about the annoying assumptions that comprise Standard Model physics and the willful ignorance of today's physicists -but then I remembered you made the vid about the sun being "liquid". Excited to listen/watch to this one!

It's really an incredible achievement, something I did not think I would witness in my lifetime. If you want to get some insight into how it "thinks", this recent video is excellent: "How Neural Networks Learned to Talk" kzfaq.info/get/bejne/hayDbJNeq76aoJs.html

Personally, I don't like Robert Dicke. I find his last name extremely offensive.