Artificial Einstein: Did AI just do the impossible? 

Discovery of that which already exists is where it has the best chance of doing something useful. It's a natural search engine. Composition is where it will always struggle, though, imo.

Joined and Subscribed. Umm... Probably. The stuff I've put into Google's A.I.? I'm gonna need FBI witness protection. 🤣 loljk! Love Your Show Mr/Dr Keating!!! Stay Free!

If AI ever does I bet there is a smart taxi driver, or some other Hi IQ non physicists, who will provide a proof that we could have have had this “discovery” 20-50 years ago, if it wasn’t for the *control apparatus* in place in the physics community, that prevents outsiders from presenting truths. Ah the Hubris of you all.

Circular Quadratic Algebra is coming after general relativity 🙂 (I made my discoveries when I started making a simulation game where I decided to simulate everything from players, npcs, and trees)

Hi, probably yes, big advantages, bigger risk, we need some caution whit AI, but the advantages are immense......all the best.

No, you're not wrong -- that's exactly where AI excels.

Syntax is dual to semantics -- languages or communication. Large language models are therefore dual! Categories (form, syntax, objects) are dual to sets (substance, semantics, subjects) -- Category theory is dual. If mathematics is a language then it is dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Mathematicians create new concepts or ideas all the time from their perceptions, observations, measurements (intuitions) -- a syntropic process, teleological. Cause is dual to effect -- causality. Effect is dual to cause -- retro-causality. Perceptions or effects (measurements) create causes (concepts) in your mind -- retro-causality -- a syntropic process! Large language models are using duality to create reality. "Always two there are" -- Yoda.

AI excels only as a learning aid for humans, automated processing, and testing. So if you have a problem it can make mistakes 95% if the time, but it can be programed to eventually narrow to a working model completely automated that is the value. Which is why in the new chip set displays they showed AI training programs for training labor robots. If they actually works large scale we will not know for years. It will almost certainly be pushed into market flawed like nearly every program. Sadly, bugs in these programs have very high real world costs.

You nailed it. It's good at analyzing for patterns in "dimensionality" which personally I think is better described as adjectives for our language. Every dimension is just a part of the description and it's just a measure of how orange is something relatively (orange just being a randomly chosen example) and how much does that matter to interpreting it. A jacket the orange value doesn't matter as much for understanding but for other things it matters immensely.

How do you fact check it? AI can't reproduce reliable biographies of living persons without making shit up and you expect it to do advanced physics based on trust?

If going further you get bio-engineering. If this is the start of an AI revolution what could it enable and be the next, think it could be bio-engineering. It's to messy and complex for the Human mind to handle so I think some AI is needed.

I hope it can help my torn radiator cuffs. Both are messed up Big Time and the doctors don't know squat. But the robot. He'll get me back in the batters cage swatting balls in no time flat.

​@@lubricustheslippery5028yeah the financial elites can engineer people exactly how they want!

​@@quantumpotential7639topical turmeric oil lipsomal with piperine. Oral capsule turmeric with piperine + boswellia

That's why AlphaFold is such a magical tool for the scientists working in biomedicine.

We have already been able to accelerate particles in a controlled setting to within a fraction of a fraction of the speed of light. For a next step towards time travel, what if we build a particle accelerator whose inside :track" is big enough in diameter to accelerate a 12oz can of Coca-Cola to that speed? Huh? Huh? What new laws of physics would the cola be exhibiting right after you first slow it back down and pop it open? What would it even taste like at that point? (I don't land on "can of soda" randomly. They are the optimized shape and the mass seems like it would be both ambitious but doable.) ♠

Syntax is dual to semantics -- languages or communication. Large language models are therefore dual! Categories (form, syntax, objects) are dual to sets (substance, semantics, subjects) -- Category theory is dual. If mathematics is a language then it is dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Mathematicians create new concepts or ideas all the time from their perceptions, observations, measurements (intuitions) -- a syntropic process, teleological. Cause is dual to effect -- causality. Effect is dual to cause -- retro-causality. Perceptions or effects (measurements) create causes (concepts) in your mind -- retro-causality -- a syntropic process! Large language models are using duality to create reality. "Always two there are" -- Yoda.

Yes, feed the robot dessert FIRST so it gets really excited and a sugar boost with vivid theoretical fancy fantasies so that once it sits down for the meat and potatoes main course, with a salad of course, it can get busy producing amazeable out of this world stuff we never imagined before, but it now has. Fancy Fantasies and Ai are like the 2 pillars of Solomens Temple. Now let us pray for divine guidance as this is powerful stuff and we gotta be careful here going forward. 🙏

r​eally?

Imo LLM will always have trouble differntiating facts and fiction because its all just text and it has no experience outside of that.

The issue here is much more fundamental, Ai can NEVER articulate its reasoning because its not actually intelligence. Its just weighted averages putting words in an order that makes the most sense. Not that the article is true and correct, just that the flow of its writing is legible. Because it's just a fancy spin of the auto-suggestion feature we've had on mobile keyboards for years now. Based on that system, you're never going to get sentience and, in this sense, intelligence

If there is a BLM candidate on the rail, they win all outcomes.

So maybe use it to create new experiments then?

I'm not entirely sure about that. As an outsider (to physics) with more experience on the AI side of things, it seems to me that there's a strong disconnect between "macrophysics" or "applied physics", as in things we typically observe and deal with on a regular basis (my car accelerates via kinetic energy, experiences friction, and so on), versus "microphysics" or physics such as they occur at atomic and sub-atomic levels. I'm pretty sure that data for macrophysics is cheap, plentiful, and readily available in a large variety of categories, and is fairly well understood such that many phenomena can be simulated as needed. So the really interesting question is: Given a large amount of semi-relevant data, can you augment a model trained on a small quantity of highly relevant data? The answer is generally yes, but it will take a careful approach. As a very crude solution, training a Transformer on anything before your target information will generally improve its performance on your target information for whatever reason. With that said, I think there are probably things we can infer about unknown behavior in physics, particularly at a micro level, from patterns in macrophysics that humans aren't necessarily well equipped to find, and I would suggest that it might be wise to hesitate to judge the effectiveness of AI in this area, particularly as we switch from data prediction driven AI (current generation) to more computationally dense "simulation driven" AI (next generation, which we're already starting to see with Quiet*, or agentic workflows, and so on), which function more like human brains and how we think, in a much more data efficient manner than we've seen before. That said, I don't think that we're going to see in the next year "AI uncovers the final unknown physical laws, and as it turns out, entropy was just a suggestion, and 42 was behind it all along", but I do think we're going to see more "unknown unknowns" in terms of the acceleration of progress in a variety of fields due to the increased efficiency of research as a function of advancing artificial intelligence.

Absolutely. AI is just multidimensional curve fitting in the end so garbage in just produces garbage out.

Syntax is dual to semantics -- languages or communication. Large language models are therefore dual! Categories (form, syntax, objects) are dual to sets (substance, semantics, subjects) -- Category theory is dual. If mathematics is a language then it is dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Mathematicians create new concepts or ideas all the time from their perceptions, observations, measurements (intuitions) -- a syntropic process, teleological. Cause is dual to effect -- causality. Effect is dual to cause -- retro-causality. Perceptions or effects (measurements) create causes (concepts) in your mind -- retro-causality -- a syntropic process! Large language models are using duality to create reality. "Always two there are" -- Yoda. Syntropy (prediction) is dual to increasing entropy -- the 4th law of thermodynamics!

Syntax is dual to semantics -- languages or communication. Large language models are therefore dual! Categories (form, syntax, objects) are dual to sets (substance, semantics, subjects) -- Category theory is dual. If mathematics is a language then it is dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Mathematicians create new concepts or ideas all the time from their perceptions, observations, measurements (intuitions) -- a syntropic process, teleological. Cause is dual to effect -- causality. Effect is dual to cause -- retro-causality. Perceptions or effects (measurements) create causes (concepts) in your mind -- retro-causality -- a syntropic process! Large language models are using duality to create reality. "Always two there are" -- Yoda.

That is pretty simple. Anything worthwhile is orders of magnitude harder.

I agree with you Gary. I assumed that the presenter is aware of the difference between machine learning and actual AI but that he is obliged to say "AI" because that's the term that commercial developers use and they provide funding for projects.

AI will not do it. The human mind might.

Our pleasure!

Syntax is dual to semantics -- languages or communication. Large language models are therefore dual! Categories (form, syntax, objects) are dual to sets (substance, semantics, subjects) -- Category theory is dual. If mathematics is a language then it is dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Mathematicians create new concepts or ideas all the time from their perceptions, observations, measurements (intuitions) -- a syntropic process, teleological. Cause is dual to effect -- causality. Effect is dual to cause -- retro-causality. Perceptions or effects (measurements) create causes (concepts) in your mind -- retro-causality -- a syntropic process! Large language models are using duality to create reality. "Always two there are" -- Yoda. Syntropy (prediction) is dual to increasing entropy -- the 4th law of thermodynamics!

Well thatt is not surprising because there is plenty of training data and underlying theory. AI is just a fancy way of pattern matching and has limitations.

No.

Yes ! AI will teach us a great deal about being sincere, in other words, about not contradicting ourselves for ideological reasons.

Syntax is dual to semantics -- languages or communication. Large language models are therefore dual! Categories (form, syntax, objects) are dual to sets (substance, semantics, subjects) -- Category theory is dual. If mathematics is a language then it is dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Mathematicians create new concepts or ideas all the time from their perceptions, observations, measurements (intuitions) -- a syntropic process, teleological. Cause is dual to effect -- causality. Effect is dual to cause -- retro-causality. Perceptions or effects (measurements) create causes (concepts) in your mind -- retro-causality -- a syntropic process! Large language models are using duality to create reality. "Always two there are" -- Yoda. Syntropy (prediction) is dual to increasing entropy -- the 4th law of thermodynamics!

I adjust the playback speed for just this reason. It works perfectly for speeding up Chomsky to how he spoke 30 years ago.

You're a pleb, and you were clickbaited

well you'd hope so… and AI isn't sentient so not sure if it really learns other than Machine Learning… does it know the real world as opposed to simulation, can it tell the difference on the level of consciouses (I'd suggest not)?

Ironic, then, that the average human is getting progressively dumber

Unlikely.

Both

@@DrBrianKeating distinguished higher than PHD?

@@frazerhainsworth08 totally different thing. I have a PhD and I was a Professor now I have a title Chancellor’s Distinguished Professor of Physics at UC San Diego.

proofesor ? that's the one who's proving the professor's theories

@@DrBrianKeating congratulations. do you go by Professor or Chancellor?

All those things you talk about can be done by humans with enough time and technology. No AI is necessary.

@@raul36 if AI is created by humans to do it, then I guess humans are just using tools to do it faster.

Bound to catch us all out at some stage. Videos at first then perhaps robots if humanity lasts that long (AI might extinguish us)

Syntax is dual to semantics -- languages or communication. Large language models are therefore dual! Categories (form, syntax, objects) are dual to sets (substance, semantics, subjects) -- Category theory is dual. If mathematics is a language then it is dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Mathematicians create new concepts or ideas all the time from their perceptions, observations, measurements (intuitions) -- a syntropic process, teleological. Cause is dual to effect -- causality. Effect is dual to cause -- retro-causality. Perceptions or effects (measurements) create causes (concepts) in your mind -- retro-causality -- a syntropic process! Large language models are using duality to create reality. "Always two there are" -- Yoda. Syntropy (prediction) is dual to increasing entropy -- the 4th law of thermodynamics!

AI is, as we have already seen, as biased as the people who input the training material

Wrong. Assumptions are the bare minimum and AI has to start from the same assumptions. It is not magic as you seem to think.

@@rogerphelps9939 I think you miss my point. It’s sarcasm.

@@drscott1 It was a bit too subtle for me but thank you.

Syntax is dual to semantics -- languages or communication. Large language models are therefore dual! Categories (form, syntax, objects) are dual to sets (substance, semantics, subjects) -- Category theory is dual. If mathematics is a language then it is dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Mathematicians create new concepts or ideas all the time from their perceptions, observations, measurements (intuitions) -- a syntropic process, teleological. Cause is dual to effect -- causality. Effect is dual to cause -- retro-causality. Perceptions or effects (measurements) create causes (concepts) in your mind -- retro-causality -- a syntropic process! Large language models are using duality to create reality. "Always two there are" -- Yoda. Syntropy (prediction) is dual to increasing entropy -- the 4th law of thermodynamics!

@@hyperduality2838 Your comment is gibberish. I can't tell if that is intentional/

@@danielkanewske8473 The neuroscientist Karl Friston talks about causality loops, he has some videos on RU-vid you can watch. The external world of matter causes effects in your mind which you perceive -- causality. Your mind (causes) can effect the outside world -- causality. Your perceptions (effects) are becoming causes -- retro-causality or syntropy. Perceptions (effects) are becoming causes in your mind -- causality loops. Concepts are dual to percepts -- the mind duality of Immanuel Kant. The thinking process converts measurements or perceptions into conceptions or ideas -- a syntropic process! Your mind is therefore creating or synthesizing reality -- the syntropic thesis! You can watch these videos about duality in physics, watch at 11 minutes:- ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-DoCYY9sa2kU.html And this at 1 hour 4 minutes:- ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-UjDxk9ZnYJQ.html Teleological physics (syntropy) is dual to non teleological physics (entropy). Your mind is syntropic as you make predictions to track targets and goals -- teleological. Syntropy (prediction) is dual to increasing entropy -- the 4th law of thermodynamics! From a converging, convex or syntropic perspective everything looks divergent, concave or entropic -- the 2nd law of thermodynamics! Convex is dual to concave -- mirrors or lenses. My syntropy is your entropy and your syntropy is my entropy -- duality. Mind (syntropy) is dual to matter (entropy) -- Descartes or Plato's divided line.

@@hyperduality2838 reported

@@wesexpress3343 You can report this:- Syntropy (prediction) is dual to increasing entropy -- the 4th law of thermodynamics! The conservation of duality (energy) will be known as the 5th law of thermodynamics -- Generalized Duality. Energy is dual to mass -- Einstein. Dark energy is dual to dark matter -- singularities are dual. Positive curvature singularities are dual to negative curvature singularities -- Riemann geometry is dual. Space is dual to time -- Einstein. Gravitation is equivalent or dual (isomorphic) to acceleration -- Einstein's happiest thought, the principle of equivalence (duality). Duality creates reality!

Syntax is dual to semantics -- languages or communication. Large language models are therefore dual! Categories (form, syntax, objects) are dual to sets (substance, semantics, subjects) -- Category theory is dual. If mathematics is a language then it is dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Mathematicians create new concepts or ideas all the time from their perceptions, observations, measurements (intuitions) -- a syntropic process, teleological. Cause is dual to effect -- causality. Effect is dual to cause -- retro-causality. Perceptions or effects (measurements) create causes (concepts) in your mind -- retro-causality -- a syntropic process! Large language models are using duality to create reality. "Always two there are" -- Yoda. Syntropy (prediction) is dual to increasing entropy -- the 4th law of thermodynamics!

Rewrite the Internet in Rust? 😊

Syntax is dual to semantics -- languages or communication. Large language models are therefore dual! Categories (form, syntax, objects) are dual to sets (substance, semantics, subjects) -- Category theory is dual. If mathematics is a language then it is dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Mathematicians create new concepts or ideas all the time from their perceptions, observations, measurements (intuitions) -- a syntropic process, teleological. Cause is dual to effect -- causality. Effect is dual to cause -- retro-causality. Perceptions or effects (measurements) create causes (concepts) in your mind -- retro-causality -- a syntropic process! Large language models are using duality to create reality. "Always two there are" -- Yoda. Syntropy (prediction) is dual to increasing entropy -- the 4th law of thermodynamics!

@Nogill0 You've assessed the abilities of Artificial Neural Networks to be better (or not) than human brains, at being theorists. There exist two weaknesses in your assessment. First, you assessed only today's early ANNs and assumed we see nearly their full potential. We do not, because what we see in Biological Neuronal Networks also applies to ANNs -- that is, both show a very strong adherence to scaling laws, across both the total number of neurons & the total number of synapses in a system. Using only this first "law", we expect to see (and are seeing) immense gains in ANNs as we scale them near/beyond the human brain. Second, today's ANNs are a rudimentary "toy" or "cartoon" version of the BNN -- i.e. their graph/network topology & simplified "neuron" -- which don't include the physics & biochemistry emerging in BNNs from their many neuron-types, many network-types, many signaling neurotransmitters, and many non-neuronal cells (all of these factors compounding the toy-version networks).

@@keep-ukraine-free528 Is it possible to incorporate the actual functioning of actual biological neural networks into a non-biological device? We really don't fully understand how brains work, do we?

@@Nogill0 You asked "Is it possible to incorporate the actual functioning of actual biological neural networks into a non-biological device?" Yes, it's possible -- and we're doing this today, but it's done within the constraints of what "actual functioning" means. ANNs today incorporate many top-level features of biological neurons and also top-level features of biological networks. Both are sufficient to give these ANNs human-like performance across many facets of behavior. They don't provide full equivalence to animal/human brains yet, but they are expected to do most of it using only continued scaling. We don't incorporate the full ("actual") functioning seen in biology because (1) we don't need to copy them fully -- since so far we get most of the behaviors using only top-level features of biology, and (2) we may never be able to fully duplicate biology, for cost/efficiency reasons. You also asked, "We really don't fully understand how brains work, do we?" We don't fully understand it, but that's not a problem. We don't fully understand many aspects of the natural world, but still we're able to exploit our partial but sufficient knowledge. We don't fully understand how most birds fly, but we can make good very safe airplanes. We understand enough about organic brains, such that we can build artificial systems capable of doing much that people/animals can do. We're still in the very early phase of neural network-based AI, and I expect we'll continue making huge advances this decade.

You are so welcome! BrianKeating.com

@@DrBrianKeating *Fantastic! TY!*

@oididdidi Let's not insult, when we don't understand what someone said. He is not boasting. He described the name of the "chair" upon which he sits--the role behind his academic position.

You want AGI then? AI that is self aware like a human. AI that can use memory to reelect on past event to influence current events. AI that understands the implications of its actions. It will soon be here. What is special about the human brain? It has a evolved structure to suit the environment. AI will also evolve to suit the environment, just at a far faster pace.

It will gert nowhere.

@rogerphelps9939 why such a "heavy" heart?

Real intelligence, but not having evolved from biological cellular machines, but instead engineer by biological cellular machines (humans).

Same as our brains.

Syntax is dual to semantics -- languages or communication. Large language models are therefore dual! Categories (form, syntax, objects) are dual to sets (substance, semantics, subjects) -- Category theory is dual. If mathematics is a language then it is dual. Concepts are dual to percepts -- the mind duality of Immanuel Kant. Mathematicians create new concepts or ideas all the time from their perceptions, observations, measurements (intuitions) -- a syntropic process, teleological. Cause is dual to effect -- causality. Effect is dual to cause -- retro-causality. Perceptions or effects (measurements) create causes (concepts) in your mind -- retro-causality -- a syntropic process! Large language models are using duality to create reality. "Always two there are" -- Yoda. Syntropy (prediction) is dual to increasing entropy -- the 4th law of thermodynamics!

What do you mean? Nothing clickbait about it.

@@DrBrianKeating Yes it is, you claim shit in your thumbnail then spend an entire video asking questions. If you insist its not clickbait then, it’s just a bad format and like everyone else. Blocked, unsubscribed.

I think @andyc8707 did not watch the video. However, if the time stamp chapters in the video referred back to the title; I think the message would have come across more clear. I enjoyed it! 13:23

@@forestfield5597thanks for confirming my point, clickbait, filler content and no real content. It’s a shitty format.

Maybe watch again then see what you think . I often find a second view can give an entirely different understanding .