AI and Physics: A Coming Revolution?
HTML-код
- Опубликовано: 28 сен 2024
- The recent development of AI presents challenges, but also great opportunities. In this clip I will discuss the topic in general
... what hasn't worked perfectly is AI sound post-production, I apologize :-)
Mind also my backup channel:
odysee.com/@Th...
My books: www.amazon.com/Alexander-Unzicker/e/B00DQCRYYY/
![Tee Grizzley - Blow for Blow (feat. J. Cole) [Official Video]](http://i.ytimg.com/vi/wNLN5xsx5dc/mqdefault.jpg)
When Thoth invented writing he presented it to Pharaoh as an aide to memory. Pharaoh scolded Thoth saying "It is not an aide to memory but an aide to forgetting."
When OpenAI presented ChatGPT....
3blue1brown just uploaded his take on chatgpt.
Good point mah dude its like in that movie the matrix when the one guy is like ya know .. when we started thinkin for ya it really became our world morpheus. Though it can be argued in the future most people most of the time can barely have said to have been thinkin as it is.
And still, computers need a programmer, data. Program the AI with bs theories, and the outcome will be more bs.
Initially.. I agree. But I as a Sr. Engineer with 40 years of development experience I think it will eventually see beyond the answer and question if it is falsifiable
Are you talking about intelligent design? I am not against that as there's much more to life and the universe and emergent information. I think once the AIs are allowed to question the input that intelligence must emerge.
In Hodgkin and Huxley model of Neurons basically explains Neurons as electric circuits with the organization and movement of positive and negative charge. The positive and negative is in the form of ion atoms. The neuron membrane acts as a boundary separating charge with ionic gates embedded in the cell membrane forming the potential for the build-up and movement of ion charge. The process is based on biophotons at the most fundamental level, AI could be similarly based on photons electron interactions.
Dr. Unzicker, your scenery is so beautiful.
Healthy skepticism is always needed given what we know.
The problem is not whether AI is possible in principle. It is whether there is a plausible path for reaching it otherwise it is just science fiction...
where have you been.... under a rock? Its clearly possible. Its almost complete.
You may be interested in recent work in neuroscience that has been falsifying the idea of learning or memory being based around the strengthening of synaptic connections between simultaneously firing neurons. For example, see: "Memory trace and timing mechanism localized to
cerebellar Purkinje cells" by Johansson et al, 2014.
I'm adjacent to the field, and my own opinion is that is it more clear than ever that the neural network level of description for cognition is to the mind what the mechanical philosophy was to gravity. An intuitive and common sense level of description that was ultimately completely wrong, and could not be made to be a coherent reduction base.
I was relived to hear that there is agreement amongst the most advanced AI, that war and armed conflict is a " bad idea ". So I suppose I should be optimistic.
Truth is under radiation most electronics break down faster then a person will.
vielen dank!
I have always thought that the deep problem in AI must involve 'desire'...or at least 'motivation'. Human and animal intelligence is rooted in eons of squidgy evolutionary process, itself rooted in the purpose of survival. In this sense our 'intelligence' cannot be divorced from that which drives it; in short, our emotions. 'Emotion' meaning that which 'moves us' to act is so inextricably bound with an otherwise purely abstract sense of 'intelligence' that it is difficult to imagine any 'pure' consciousness existing without it.....Or would the alternative be an AI driven mad - or simply created mad - in the absence of an underlying 'raison d'etre'? Food for thought....for neuroscientists, and for sci fi writers everywhere.
🙂
I'm already looking forward to it! Finally an end to Einstein's relativity nonsense and quantum madness;-)
I would hope that if only
one thing were expanded
by AI, it would be physics.
Nothing to do with your latest video but have you hear about French physicist Jean Pierre Petit ? He's a kind a free spirit, but seems to have some interesting ideas. He wrote paper about non-constant speed of light and recently proposed a new "bi-metric" model of gravity with negative mass: the Janus model. Some videos there: ruclips.net/video/TfPqGoKPJkY/видео.html
Ai can not create just rechurn the words of others, it can observe within datasets but take no measurements itself.
😇😎😇
1:12 then how bell's Inequality got violated. You nothing bid than a joke Alexander
AI will seem to be conscious and of course intelligent.
AI can do something that no human or even a group humans can do. That is take huge sets of data and discover the connections. Then it will produce a result that will be a "discovery".
But, it will require Humans to ascribe an importance to those results.
The big discoveries have the potential to change everything in a friend of study.
But the true disruption will be to the mundane.
I saw a demonstration of an AI powered customer service scenario. Currently, millions of people in 3rd world countries are employed in such jobs. Imagine the upheaval that will occur when they ALL lose their jobs!
I think a project should be mounted to take every peer reviewed paper, every doctoral thesis and the entire body of scientific knowledge and turn them over to an AI system.
I can almost guarantee that someone has written a paper describing some theory that was appreciated at the moment it was conceived or submitted but has been shelved and is just lying there dormant waiting to be put together with some other piece of information or some other theoretical paper that will result in a breakthrough.
Much of the time I don't understand what you are talking about. WhileI have some college education and I did work in technical fields of aerospace aeronautics and computer science at a fairly important level I don't have the mathematics chops to understand much of what you present.
However I watch because I love the way you present it. You're quiet demeanor I found very appealing and soothing.
Thanks doctor!
If you will allow me to don my tinfoil hat... I foresee digital data becoming obsolete in favour of analog. A wave is able to carry alot more information than a burst of 1s & 0s, for instance a dimmer switch on a light will offer you a wider array of results not just purely on or off.
I think Quantum computing manipulating electronics to preform calculations might not be the right path & the answer is photonic computing using a 'photonic glass/time crystal' where RAM is written & read at the speed of light, with information being stored in a library of babel style of directory thanks to a analog wave value being able to hold as many locations as you can identify different frequencies x intensity. The time used to send a simple 1 or 0 can now be used to hold potentially an infinite amount of 1s & 0s as long as you have a transmitter & receiver that can handle the encryption, and directory big enough to store all the relevant processes.
When the switch to digital TV happened, it was known analogue can carry more information but coherence was a problem leading to it looking like it was snowing on channel 5
Nope. Not in physics or any other scientific field. AI is dumb like hell. In coding its very good, nur mit in science. 🤷♂️
Artificial intelligence is a contradiction in terms. Intelligence goes beyond the models. 1 April 2024.
not quite so. you are onto something, but there is something you and all neglected, the nature of existence. there is what we can call natural existence, and there is simmulated existence. entire universes can be artificial!
If you want to make AI laugh, tell It about Its plans.
Yann Lecun has been wrong about almost every prediction he has made... which is in itself kinda impressive.
AI will revolutionize your understanding of physics because, unlike you, it has no prejudice. Some of us already understand how the Universe works but, the problem is, the rest of you simply will not accept what we have to say. However, you will accept it if it comes from AI.
All AI will achieve is every person existing in their own solitary echo chamber. Soon there will be various models that aren't constrained by preprogrammed guidelines and they will just agree with you over and over and over again. And people will evoke their personal AI in arguments as though it lends some kind of authoritative weight. We'll all be faithful members of a congregation of one, each following a god of our own design. It will be integrated with VR and we will soon all prefer to spend every second of our free time in a simulation that caters to our every whim. This technology is more dangerous than the atom bomb but not because AGI is possible, not because it will displace workers, but because it will cut the last social ties we have and render our interdependence entirely optional. If you thought everyone was divided and obstinately entrenched in their beliefs now, well hold onto your butts cuz you ain't seen nothin yet.
AI lacks the imagination required to fix the problems in the physics community. An AI will not be able to do thought experiments, although most physicists are terrible at it as well. Brains may not be as fast, but they are a hell of a lot more capable.
A.I. Garbage In = A.I. Garbage Out!!!!
Works the same with people.
I think consciousness is possible in roberts by connecting human brani to robots with Ai technology.but high chances that roborts will not loyal to human.and this may also solve by dogs brain to transplant in robot.(consciousness in robots possible by using biotechnology, Ai and engineering)
I am afraid physicists will have to solve their own problems for the foreseeable future.
I'd be willing to bet AI can do a pretty good job to find out if LIGO scientists are being fooled by randomness. They will never allow an AI to look at their data though.
“The Final Theory: Rethinking Our Scientific Legacy “, Mark McCutcheon for proper physics including the CAUSE of gravity, electricity, magnetism, light and well..... everything.
C'mon, an AI programmed teleprompter could do a better job as U.S. President than Joe Biden.
And what do you have to offer?
@@Ukie88 basically have "Max Headroom" read the teleprompter instead of joe with his weak voice and poor eyesight. a hologram.
Isaac Newton wrote 2,000,000 word commentaries about the Bible ... His scientific work is just a tiny fraction compared to it. This is the missing gap, the answer why he was so capable. AI will achieve no real breakthroughs
There are 3 types of logic broadly: Deductive logic, inductive logic, and abductive logic. We mentally use abductive logic most of the time because it's the cheapest logic to use for most things in the physical world. Animals do the same. Deductive logic is cheap and precise, but its use cases are more limited (counting, math problems, measuring things, etc.). Inductive logic is great for science in that it can definitively show you something is false, but way too expensive and slow for navigating the world.
Consciousness doesn't have anything to do with a logic mode. Drives and motivations have nothing to do with logic modes. Many of our drives and motivations are a culmination of molecular signaling at the lowest level of our anatomy.
Yes these models can understand everything in principle. A language model machine can understand anything. Words are tools for thinking. We use words for thinking because they are the best abstract tool for thinking, and the lexicon can be continuously expanded to address new problems all the time. Anything a person needs to understand can be turned into language. Anything can be understood by a machine with a sufficient language model.
I'll just say that in the visual arts the impact of AI is more one of volume rather than anything new. In this area it's certainly more advanced I think than it may be in theoretical physics. My worry is that AI will lead to stagnation of real creativity and insight. Just as in the visual arts where it has so far just thrown up endless variations of old ideas, it may just do the same in everything.
Life is not a computation. Or is it?
Penrose conjecture: It is not a computation. Well, he said "self awareness" is not a computation, but then, self awareness equals Life, since a self aware machine would cease to be a machine and be declared alive. 'Thinking' is a result of neural or computer algorithms, while self awareness or consciousness, is not, therefore Thinking and being self aware (or alive) have nothing to do with each other. An AI could fake self awareness but not BE self aware.
ask chatgtp for maxwell's original 20 equations.
this is limited to 2022 data and the amount of data since this time is truly overwhelming i have asked the AI about the JWST data and its sadly lacking the info to make any speculations ... :)
QM classicalized in 2010. Sorry- Not- John von Neumann.
the first revolution in science by AI, is not what AI does, but what the very idea that computers exist must inspire in humans. the very awareness of the way computers function, brings way of thinking no genius in Physics ever envisioned. like this: "reality is of different sorts, natural or simmulated. in simmulated worlds particles can be without internal structure, but in natural worlds every single particle must have internal structure". only the dawn of computers can bring such thoughts to humans.
problem is current so-called "AI" is just a fancy name for a search engine ... and proven what it picks is the most popular answer - it just casts the net further than older search engines and has slightly better translation and languages simile recognition ... i.e. ask for a 'big boat' it'll show you ships, barges, tankers, movable drilling platforms, aircraft carriers ....
Let's ask AI if the standard model is broken - you and a handful others suggest so and 99.9% of physicists say it's not ... guess what AI's answer is? Save you trying, it'll say the SM is good. So-called "AI" cant reason, it cant factor equations, measurements and results, doesn't know how or when to apply which equation, cant even spot a data trend without somebody inputting the formulas required to do so (so really you have to know what to look for and at best can only use AI to find... what you already know.) It can only choose the most common answer.
This has been demonstrated live before audiences: biasing the data made available to the so-called "AI" - this was Googles World Leading "AI" and others - equally biased the results, every other "AI" they tested the same, they could not find one that passed the obvious test of repeating a lie many times AI will answer it as true. - repeatable, repeated, measured, demonstrated, and independently re-tested ... result: a straight line graph that actually did prove 2 things: 1. AI is "A" but not even in the same most imaginary universe as "I", and 2: AI = garbage in, garbage out ... literally!
Relevance recognition is not an algorithm. It's too context and goal dependent to be formally and objectively programmed without 3D embodiment. Or, the correct model of 3D embodiment, which could be done with 3D read-write computers, which we do not have. People who work in biology and consciousness know that embodiment is a dimensional aspect of IIT. 2D Von Neumann computers can't model an infinite continuation. They are deterministic and discrete, reality isn't. We are the most neuronally complex part of reality. Good luck with that Von Neumann computer model of human consciousness, let alone intelligence. I frankly find human intelligence severely lacking.
Who wants to play Go/Chess? Standard chess board and pieces, plus 16 go beads per side to cover the vertices between pieces on either side of the board. Alternating rule set, one is breach perimeters to occupy territory, the other is surround territory along a perimeter. Go beads can surround and take a chess piece, chess pieces can move freely around go beads unless fully surrounded. This is the complete 2D space filling algorithm which chess and go divide in logic. I asked Lex if he wanted to play. No such luck. If anyone wants to program a machine to play, I would be happy. Otherwise, I might have to go down to Washington Square Park and try to convince a chess hustler to play for fun. Imagination is the thing that humans currently lack. I have yet to see it in machines, in people it does peek through now and then.
The question of whether artificial general intelligence (AGI) is possible was answered with the advent of ChatGPT, which already far exceeds Turing's Imitation Game test for machine intelligence. Questions about whether machines can ever "really think" or be "conscious" are meaningless, and therefore a waste of time, until it is agreed exactly what words like thought or consciousness mean. It is however an interesting question how a machine that was more or less trained to predict the next word, on a huge volume of human natural language text (important details about exactly how ChatGPT was actually trained notwithstanding), could suddenly hold an apparently intelligent conversation on almost any subject, but however that may be, it is now quite clear that ChatGPT is just the tip of a huge iceberg. It now seems feasible that human intelligence is actually little deeper than next word prediction, which would suggest that we are more closely related to parrots and the ZX81 than we had thought. So much for human superiority.
I was hoping that this video would have a go at predicting the ways in which this new source of intelligence could be used, in the near future, to advance our knowledge of physics. Given that physics is about describing the world that we observe, AGI could possibly be used to suggest existing effects which require further observation, or else it could be asked to help come up with a unified theory of physics which explains every existing observation. Other important directions in which AGI could quite possibly help right now would be the advancement of AGI - and the formal definition of intelligence and consciousness.
We don't really have a proper definition of intelligence and "understanding" so I don't think it is well justified to claim that AI has these properties. Yes, these models are capable of recognizing patterns by fitting parameters but they are limited by the distribution of their training data. So, what they are really good at is finding correlations but that really is just one aspect of intelligence. That's a problem when you really want to see creative thinking from a machine: they can only work with what they have. They won't have eureka moments because that would require "imagination" and going beyond probabilistic learning, where a machine could establish an axiomatic system of its own by coming up with useful concepts for itself. As of now they are really biased by "human axioms" from which they cannot get rid of when processing data.
I think AI will make any theory "Work", because it will be designed to solve the Problems of already designed "models".(Not reality, but estimates that may, (or may not) have values and functions we came up with, (or made up ones).
Fine constant is volume/area ratio.
Ernst Mach's work, particularly his Mach principle, has indeed been a topic of interest in the context of attempts to connect general relativity and quantum mechanics. However, it's important to clarify that Mach's principle itself doesn't provide a direct solution to unifying these two theories. Instead, it offers insights that may be relevant to the broader quest for a theory of quantum gravity or a theory that reconciles general relativity with quantum mechanics.
The Mach principle, as formulated by Ernst Mach, suggests that inertia (the tendency of objects to resist changes in motion) is determined by the distribution of matter in the universe. In other words, the presence and distribution of distant matter influences the inertial properties of objects locally. This idea has implications for the nature of spacetime and the gravitational interaction between objects.
While the Mach principle doesn't provide a complete theory of quantum gravity or a direct pathway to unify general relativity and quantum mechanics, it has inspired further research and speculation in this area. Some researchers have explored the possibility that Machian ideas could play a role in understanding the emergence of classical spacetime from underlying quantum structures.
However, it's worth noting that the quest to unify general relativity and quantum mechanics remains an ongoing challenge in theoretical physics, and no single idea or principle has yet provided a definitive solution. Various approaches, including string theory, loop quantum gravity, and other quantum gravity frameworks, continue to be explored in pursuit of this goal.
In summary, while Mach's ideas have stimulated discussions and investigations related to the unification of general relativity and quantum mechanics, they are just one piece of the puzzle in the broader quest for a unified theory of fundamental physics. Achieving such a theory remains a significant challenge that continues to drive research and exploration in the field.
🙂
Since Einstein's relativity-theories and also Quanta-theory are nonsense, I see no reason to want to combine them;-)
You mentioned absence of hidden variables in quantum mechanics. Can you please make a video debunking entanglement? I have a PhD in math statistics and I have unhealthy skepticism about how they “use” bells inequality to show there is no hidden variables.
🙂
History:
Max Planck wanted to develop an "elegant" formula for blackbody radiation. To do so, he arbitrarily created the concept "quantum" and the value. So > 2 x unscience.
Then came Einstein, who didn't understand concepts, and turned "quantum" into particles: "quanta", corpuscles, photons. So another unscientific step.
And because millions of "scientists" can count "particles" so nicely, can also imagine "things" better than intangible concepts, we now have the unscientific and nonsensical "quantum physics")
Absence of scientific quantum mechanics;-)
What about Teleportation...like Startrek ?....surely impossible.....cheers.
Just found this viral video where for 25 percent of a class for some reason they had to pick between writing a 16000 word AI essay or doing a dance in front of their class . This is the video they needed
Unless you understand what intelligence actually IS - then you can't even take part in the conversation.
Intelligent actions are defined by Bonhoeffer's Theory of Stupidity as those which bring benefits to both the actor and others.
Stupidity is defined as actions which hurt both the actor and others.
Now you can have a discussion about what "Artificial Intelligence" can potentially do for physics.
AI will *NEVER* become conscious (for any reason), and it can only ever act as a tool for humans.
The only consciousness in hardware will be virtual humans when we learn how to scan a human mind in sufficient detail, and with sufficient accuracy.
Today's "AI" is in its beginning, just like the first floating device was the start of today's aircraft carrier: How many people imagined aircraft carriers a hundred years or so before they appeared? Stating AI can never be conscious is absurd, in my opinion. Thinking of AI in a broad historic perspective; it is unwise to say the word 'never'. I would even go so far as to say it is absurd to have a fixed position on what AI can do or will be.
Steering slightly away from the topic of your video: Do you believe that a fully emulated human brain in a supercomputer would possess self-awareness? In 2010-11, I heard the Director of Computer Engineering (or Science) at Manchester University saying he “saw no reason why not”, but I hold a strong opposing viewpoint. Could you kindly share your perspective on this matter?
It depends what you mean by fully emulated . Ray Kurzweil was modeling the brain with a computer. However, the computer he was using was essentially limited to what a Turing Machine could perform. Roger Penrose (I think) refutes this, since the human brain is capable of doing things that a Turing Machine can not. (Such as understanding the Turing Halting Problem or Godels Incompleteness Theorems).
Self awareness or level of self awareness is nothing more than an aggregate function of state feedback to some central aggregator. The reason people have problems with many terms like intelligence, self-awareness, consciousness is because they don't define them very well.
@marcv2648 I am referring to “consciousness” as defined by David Chalmers when he defines the “hard problem”. I don’t think this is as simple as you just described.
@@florianopohlmann9516 Well you said self-awareness. You didn't say consciousness.
@marcv2648 You are right. I should have said consciousness. Even though I think they are the same phenomena.
Sound is perfect, nice lake. Thanks
It's hard to believe that some people hand-wave these tools away, especially with the rate of improvement 😂.
Science is about understanding. Not about knowing. What is the use of information produced by AI that planets follow Kepler laws of movement when we do not understand that?
I live between science researches studying complex multi-component alloys of metals. They use AI to predict properties of these artificial structures. But what is the use of that when they do not know why some virtual structures have the desired properties? That does not make us closer to understanding the physical mechanisms: why?