Stephen Wolfram - Where the Computational Paradigm Leads (in Physics, Tech, AI, Biology, Math, ...)
Вставка
- Опубліковано 21 лис 2024
- On Friday, October 18, 2024, Stephen Wolfram, Founder & CEO of Wolfram Research, gave a keynote talk titled "Where the Computational Paradigm Leads (in Physics, Tech, AI, Biology, Math, ...)" at the "Empowering Excellence: The Hertz Way" event, an evening for the Hertz community held at the American Academy of Arts and Sciences Cambridge, MA.
00:06 Stephen Wolfram's 50-year journey in the computational paradigm
02:08 Started using computers to understand science and complexity
06:03 Humanizing the power of computation
08:08 Understanding the machine code of physics and the structure of space
12:08 Hypergraph rewriting mimics Continuum SpaceTime behavior
14:14 Quantum mechanics is inevitable in these computational models.
18:10 Basic assumptions about observers lead to understanding physics structures
20:01 Systems encrypt initial conditions, leading to observer limitations in decoding randomness.
23:29 Simulation of all possible evolution paths and the impact of changing fitness criteria
25:20 Computational approach to biological evolution
28:51 Exploring the complexities of defining viruses and implementing immune responses
31:00 Computational irreducibility challenges scientific predictability
34:32 Neural nets achieve their purpose through complex, irreducible computation
36:14 Computational irreducibility in AI and natural world
39:46 Dark Matter may not be matter but a feature of space-time
41:30 Computational irreducibility in AI and its implications on designing molecules.
45:11 Difference between using LLMs and computation in generating outputs
47:00 Human insights influence AI capabilities
51:11 AI should have a clear specification for writing code.
Crafted by Merlin AI.
I can't decide what is more awesome and exciting, the things he says about the universe experimenting with every possible rule and the biological programming from nature and devising a general theory of medicine or the beautiful drum set sitting on the stage and awaiting to be played and cut the air with a multitude of raw, rhythmic, musical waves...
I am an absolute layman when it comes to understanding modern physics. If I understand Mr. Wolfram correctly then he is saying in his lecture at the beginning that space is discrete and not continuous. That if we get to a scale small enough, we would find that space is made up of quanta which somehow produce the illusion of continuous space. Now take the number line. One of the properties of a line is that for each segment of it, we have infinitely many points into which we can subdivide it. The set of real numbers is dense everywhere. What Mr. Wolfram is suggesting is that what does not hold for a one dimensional line, holds for three or higher dimensional space. I am not convinced by his brief arguments. However, he seems to be a talented individual. He is right that we can and should delegate routine computations to machines after one has actually understood how routine computations work.
Your comparison seems a bit off, as you are comparing a mathematical concept to characteristics of physical reality. If you were to concretize a number line in physical space, you would likely find it is not infinitely sub-divisible. Similarly, if you increase the dimensionality of a number line to a higher order in a non-physical context, you would find the the spatial measures to be infinitely divisible. This apples-to-oranges comparison may be the root of your skepticism.
It's fair that you weren't convinced by these arguments in this format tbh, he didn't get nearly enough time to explain it. If interested though, theres a few other podcasts with Wolfram out there, like on Lex Fridman or MLST that may intrigue your interest even more!
I'm into learning math and have delved into his work a lot, and I am constantly flabbergasted how he's managed to explain so many things with a single framework! Whether or not it "actually represents the universe" is secondary to the mathematical power, insight and understanding we've gained. With him, we can see that Quantum Mechanics doesn't have to be so weird, he shows if you think about it in a certain way it kinda makes sense. He's also potentially unified Einsteins work with Quantum mechanics which people have struggled with for a long time.
The best thing I learned from this lecture is that his theory now has a testable prediction. Once we get quantum computers he said we should be able to see a "floor" that is predicted by their models. It's so cool! Look into "cellular automata" if you haven't already, they show such complex behaviour! Wolfram is implying our universe could be a more complicated version of "the Game of Life" by John Conway
he is not dumb, but far from genius. he bullshits a lot.
It helps to know that the edges in the ruliad, insofar as we can even talk about their "length", are predicted to be dozens of orders of magnitude smaller than the planck length, so we'll probably never actually see them at our scale. You can think of it as being discrete underneath, but at such small scales that we literally can't see the units so it appears continuous. Compare that to macroscopic objects - they don't look like they'd be made of quintillions of molecules, they just look "continuous".
"What Mr. Wolfram is suggesting is that what does not hold for a one dimensional line, holds for three or higher dimensional space" Where does he suggest this?
I enjoy this so much. I've wanted to know more about his life, and how he started. The history, I particularly like. Fascinating. Thank You. Rob
Now that I'm looking at this a little closer, is it possible that the 8 rules that he's using can be further reduced into XOR, NOT, OR, and AND operations? (2 inputs 1 output with 4 rules) ? You got to love Wolfram! Please keep up the work! This guy is amazing.
We’ve met in Terry’s lab.
Even if irreducibility stymies prediction, it looks like it allows for sets of possible causal interactions and forecloses others. This is still a happy pragmatic result.
When does he play the drums?
he has invented a new way to play drums
😀
They should have put the drums away how disrespectful to Steve
@@irishtombyrne A New Kind of Drumming 🥁
the drum solo emerges from a very simple rule
For every lecture Stephen Wolfram gives without crediting Jonathan Gorard, another fairy dies.
The thieving little folk had it coming
Fairies stole my teeth! Not my diet of coke! To hell with them!
can you elaborate for someone who is unaware of the context
@@stretch8390Jonathan Gorard is a gem of a man who is behind much of the theoretical research Stephen speaks of in recent years. Stephen has a frankly nasty habit of assuming his discoveries are brilliant and failing to mention those who assist him and those whose shoulders he stands upon. He’s a brilliant man, but he has quite the ego.
@@stretch8390 Jonathan is the co-founder of the Wolfram Physics Project and was really the brains behind the progress they made, which has slowed notably since he reduced his involvement.
To use the man's own words, he said in response to Sebine Hossenfelder's recent video on the project "Spending the last 5 years watching Stephen take sole credit for ideas, insights, developments, and discoveries that were the products of our collaboration has been a uniquely exhausting experience".
Stephen Wolfram is obviously brilliant, and I very much enjoy listening to him talk, but he is renowned for having an inflated ego and for taking credit for the work of others.
In a demonstration of "reducible" computations, he clearly stated earlier that planetary motion has closed-form solutions in our solar system. I think he shouldn't gloss over that, he should present THAT to the world first (and maybe save me a couple thousand dollars on GPU costs for numerical PDE calculations).
Are closed form solutions the goal and why? It seems to me even polynomial forms aren't so hot.
If some of the current methods for partial differential equation solving (or solution approximation) could be replaced with formulas (where you just plug in the time and space variables), that would be an unfathomably massive revolution both in applied and pure mathematics. Current methods usually involve some discrete grid and discrete time steps (FEA, FDTD, ...) and the quality of approximation can drop dramatically if we simply try to cut down on memory or running time requirements, like using larger time steps or a more coarse grid. Plus the finite precision of the floating point number representation can become an insurmountable problem - even apart from numerical degradation, it turns out that in chaotic systems a lot of information about the future states is essentially "encoded" in the countless digits of the initial conditions, or predetermined by them - to use a more accurate word. This kind of borders on some philosophical questions, but it's also the simple reason why we usually cannot make predictions deep into the future when partial differential equations are involved.
But of course the point is that these processes that are all around us are just as great examples of computational irreducability, as that of certain CA's, and I find it unfortunate that I heard Stephen denouncing this similarity, though it is true that he only briefly mentioned it, and didn't really expand on it whenever I heard him say it.
Some things appear random when we don't have the algorithm that explains its structure. This is the beauty of a neural network; it simply approximates an algorithm without even having to understand its own function.
this is actually incorrect.
check 33:00.
Neural networks by themselves are a mere function, taking an input and generating an output, but this has no side-effects.
In effect, I can know all the outcomes, it's the range of all potential bit values of the registers it outputs to.
He is referring to something more substantial, which are systems that are turing complete.
Neural nets are not turing complete initially, and if you try and make them turing complete by making them call themselves with large intermediate values, noise will destroy your result or ability will be underwelming for computation expended (think a 10 gig model doing something a 10kb c++ program can do)
here's a simpler explanation:
if some things are irreducible to doing the computation (because it is highly dependant on initial conditions), then a neural network can't execute it, because any given network will finish in constant time (every run of the same network takes the same amount of time).
@@qwertyman1511 While I agree that they alone are not Turing complete, each combination of neurons can be shown to be gradient (16 bit float) approximation of XOR, NOT, OR, and AND functions. This is a new way of looking at computation because we're most familiar with the Boolean operations (0 or 1, not gradient). Check out the 3D plot of the XOR function in a standard NN. A guy can learn a lot from it. When the question is: what is the XOR value of (0.23, 0.8)? The question may seem absurd as a Boolean operator, however, the NN will give you an approximation anyway. This is where the NN really shines.
You nailed it
Just because we have been clever enough to build computers doesn't mean the world has a computational structure. I don't think it does, as evidenced by the failure of computers to simulate quite simple looking natural phenomena such as the flow of liquids.
Nature is analog holistic computation, we are mostly thinking in digital structures which are reduction
That last demi was wonderful!
There is a third way of building a wall. Reduce the land, on both sides of where you want the wall to be and it emerges.
I think what Your Theory are the Best!! I Follow lot of Years ago Your Ideas, can revolucionate like the Boltzman Brain, Theory, Bcos, well entrophy cab be another kind of "Order" in that kind of diagram, and well of course "Universe" Use the Minimal Action Principle. and that make what Energy Flow in the more short way. But well lle You can Know maybe one rect way are not the more short. :-) Hey Go Ahead!!
I was about to ask if for the sum you'd use a "posupposition presupuestal", a necessary redundancy able to start dictating the remaining' computational boundness
"successive simultaneous sequences of times' 24:58 is from now on my best quote ever.. How to verbalize disorder / randomness and non-descreteness in an orderly orderly fashion.
how to even start to choose some set of starting hypergraphs and rules from the ruliad to model some particle interactions ?
I think this is one of the main issues and most difficult problem in Steven's theory. The Ruliad has essentially infinite different rules and to pin down the correct one of our Universe seems almost impossible as of right now
So points and relations between them is the most fundamental thing.
So, what part of the nodes/hyperhraphs or their structure causes them to evolve? @Stephen Wolfram
He can't yet model an electron. Are you sure you understand what you are asking?
@@paulschrum4727 of course I'm sure, (that doesn't mean I'm correct in my assumption tho :D)
I was (hopefully) asking if the model can explain time/why computation happens; why anything happens. Because otherwise it's just a marketing trick of over-inflating the substance of the product and selling it as something it's not
@@simdimdim If you are saying that the presence of rules strictly entails a Rulemaker, I totally agree. But if you are asking him to demonstrate biological evolution, he's nowhere near that and his physics model never will be.
@@paulschrum4727 Ugh, no.. I was just asking if the driving force behind the evolution of the graphs is somehow encoded in the graphs themselves or they fail to account for it altogether (or something like that)
@@simdimdim Thanks. It looks like I was the one who "didn't understand what you were asking." Sorry about that.
I use photography to make circuts more complicated in plateing systems that each line of the circuit can be duplicated in a few minutes or hours depending upon the sizes of the plates under and overs layering a circuit so its a whole grouping of different contacting points is imparting electronics capacitors and chokes resistor capacitor transistors etc even more importantly each layered version separates the size to smaller than normal electronics ❤
5:35 No, in "exact science", you still have differential equations, where you can also not "jump ahead".
if you can obtain from the differential equations an expression for an arbitrary future state, you avoided computational irreducibility., and so you can "jump ahead"
Physics is interesting to laymen since they can see tangible things mathematics is interesting to physicists and mathematicians only.
Psychedelic fractals my guy
Looks like a linkage to a sanscrpts that has multiple levels of languages in a single verse ❤
Can any of these models consider mass vs energy?
I don't think the "formula of the universe" can be solved with mathematics in some sort of discrete, cellular automata sense. I think what we can do is to try to use deep learning to produce a model of physics at the lowest level, that can accurately predict interactions of matter, fundamental forces and so on. And for those who would be bothered about the fact that then we cound't truly "understand" the model, I'd say: Well, how is a mathematical formula any better?? It is just an abstraction of things. Just like we can't "see" at the lowest level, because light doesn't work there the same way, I don't think we can apply human semantics at the lowest level. I want to propose the question: What if a deep learning model of physics, based on mystical vector correlations, is the best we can have? Even if we can't truly "semantically understand" it.
I was convinced Stephen had already merged with the AI, and uploaded his body to the cloud.
Can you post the slides please?
Deep respect!
It is obvious, that a single Photon, which should have a monochromatic wavelength, doesn't even have a defined energy, because this E depends on the relativistic speed difference between emitter and sensor!
Does a photon even exist?
@@tricky778 have you ever done a Hong-Ou and Mandel experiment? If not you should try...
@@deecyrlysons3401 no, I haven't, are they nice girls?
@deecyrlysons3401 will it tell me whether photons exist? It looks like it's just a special case of Maxwell-Heaviside
@@tricky778if photons don't exist what causes light, or what is light made up of? Do you mean it's a wave and not a particle? Or what?
I don't understand how it can it could be both a particle and a wave, unless that just means it's a particle that flows in a wave.
12:45 is a great Tiffany lamp shade
Unlike Levin, no creatures were harmed in the creation of these 'ideas' and they are actual ideas.
That first joke was fire.
Only a probability. But Life is energy, and energy is a Result not probability.
44:21 There is a multi-dimensional pattern it is picking up, but the pattern it is seeing is the totality of all weights and relations of weights in the model. A more efficiently represented explanation would require greater fundamental understanding.
An impressive tour de force!
Locality and realism are not fundamental but emerge from interaction
Interaction of what?!!!.
Interaction of properties, waves, or particles
@@AliBenBrahim-s9x we had plotted particles on a Cartesian grid over millennia, presumably a curved grid of the past century. More likely there is no grid at all, only interactions from which a local "grid" becomes apparent to observers who are also composed of local interaction
See amplituhedron theory (Nima Arkani-Hamed), Causal Sets-Sorkin, and Jeremy England an American physicist who uses statistical physics arguments to explain the spontaneous emergence of life, and consequently, the modern synthesis of evolution. England terms this process "dissipation-driven adaptation".
Is the application of these ideas to explain reality testable?
I guess Wolfram teachers never told him that graphs must be identified and have units😅
He's basically created a very big Spirograph program.
May I see graphs of neutrino oscillations affected by relativistic effects, i.e., when the neutrino source moves away or towards a detector at different fractions of c?
nice drums
Be mice if he dropped Marko Rodin name. Basically acting like he discovered something Marko has had for like 20 years. Granted Marko gives the proper credit to spirit.
informative
Why does Wolfram's voice sound pitched up in this video
is this the scientific version of Captain hindsight? although claiming to be theoretic it feels very, best outcome biased. Fitting to existance kind of thing.
Does this mean that Wolfram solved all of science at the same time?
great
Very correctly done hypergraphing the parts of the universe itself is a literal expression of portaling vibrations and energy levels that are apparently a thinking state of the universe itself as densities converge and diverge in the random calculations to an orderly system to think about this whole universe forever reaching the understanding of what is really happening in the process of a black hole and joining an other state of a black hole is a vibrational structure that mass falls into the state of silence where light is a vibrational structure its becoming a none vibration a
I could listen Stephen talking for the rest of my life (probably will), but 50 bucks for a pdf?!!! Come on....
Sticking with the 1st law.
I had to doubletake the date because a lot of these concepts look like the intial graspings of spinors. I thought we figured this stuff out already
how so?
I love his ideas. But his understanding of black holes seems to be off (from what I understand). TI 13:31. He shows black holes as something with nothing in them. But my understanding is everything that constitutes the matter of the black hole is inside the event horizon, not on it, as his model appears to suggest. But ought not a discrete space model of the infinitesimals somehow have tremendously dense matter, and not the absence of matter? (Seeking people to agree or correct my lay-understanding.) Note, his diagram for black holes in A New Kind of Science (2002) is also wrong (it has shows space broken at the event horizon, which does not allow the gravity, charge, or spin signals to get out, but they do.)
The number 42 is, in The Hitchhiker's Guide to the Galaxy by Douglas Adams, the "Answer to the Ultimate Question of Life, the Universe, and Everything", calculated by an enormous supercomputer named Deep Thought over a period of 7.5 million years.
#POU238 pls
excellent presentation. then he begins taking questions 😂
Stephen is onto something
Be as gross as to take the "concept of nucleic acid/ chirality, (that's a trace back point ie.), for geometry we have the FOL and sacred geometry part of the pie, and we advanced a little into the devising, I have q project called OCT standing for oscillestereogra(m) for medical, virtual enhancing, dropping, applications,,, in which bigger conventionalized patterns in many aspects of our lives become more synced with the available states ( big 13 in an hour pendulum, falling 17 meters,13min sleep cycles, creating or unraveling templates more similar to, something setting the tempo
Sometimes I wonder if seeing QM and GR in his hypergraphs is just a nun-bun for eggheads.
Many words, no experimental evidence.
Evr since chatgpt update for 4o you no longer need wolfram for accurate math
You do if you want to control knowledge what what you're working on. Chatgpt is on someone else's equipment.
How does 4o do accurate math?
@@fburton8 Before it will do the algebra just fine but for some reason it randomly used different numbers. For example in differential equations it will just get with the algebra at times
I was really hoping he would do a YYZ set on the drums. disappointing.
Doing mathematics in search of physics.
Not a good way of spending time. 😂😂😂
this is why you should realize all of these people have never done anything real in the real world, and they don't know shit about anything.
Replying to one of the comments: I agree I am a layman as well, less than that even.
"What he is saying is that space is discrete not continuous"... [end quote]
Yeah, if you are coming from and believing in that, then you are already wrong and off track. - "Qul hu Allahu Ahad!" Meaning, "Say, Allah, He is (the) One". The Ever living, self subsisting. Overseer (all seeing, all observing), maintainer, provider/sustainer, of the heavens/universe(s). Sleep, weariness, tiredness, does not touch or affect him abaddan (at all in any way). His is the command and The Dominion (of the heavens and the earth). The trees, the planets, the stars all make sujood (prostrate and obey) unto Him, Glorious and All Mighty. He wraps the night into the day and the day into the night. Look up into the heavens, do you see any rifts therein? Then look again, and again, (and again,) and your gaze will return weary unto you. It is not for the sun to overtake the moon nor for the moon to overtake the sun but all are floating, each swimming, in an orbit (all the planets and galaxies are actually in orbits of their own while within the galaxies the celestial bodies move in orbits of their own at the same, as well). It is not possible to cut and divide what can't be cut and divided. Nor can the infinite be made finite. All powerful means there is no incongruity, and if any breaks were to occur that would break time itself. There can't be any lapses. But dividing into discreet parts etc. (And that like) so we can calculate, calculate time for example, is and canube beneficial for us hu ans and we can progress in our understanding and make good use of (our, not actually our's, we don't own it nor anything in acutality)) time.
I forget where I was going with this.
Peace be upon you. ❤💫🌼🌠🌌☀️.
Shame the freq /DNA from the universe are being blocked FROM STUPID GEOENGINEERING.
omg the questions are so dumb. They don't understand the scope of the topic.
guff
Why do people with British accents manage to make a potentially interesting topic as boring as possible?
You need space to create discrete space 😅😅😅😅hahaha
what a nonsense talk!
That's not what he said. You only need relationships between points. The points aren't "embedded" in space, the points only have relationships to other points. Even if this isn't "the real universe", it's still an incredibly impressive mathematical framework that allows emergent complexity beyond anything humanity has even seen before
No he’s saying space is emergent from relations between point which are being rewritten. The points are not in a regular lattice or any space themselves
@@HamCar1000 does that make points infinitesimals?
he mentions that light is discrete, when he means that EM radiation comes in discrete units of energy.
@
Each light particle is Fuzzy cluster of EM energy! With invisible connections to other clusters as wave! But this type of energy is only 10% and remaining energy is dark!!!! Our bodies are all empty spaces with tiny energy pockets hologramming from the edge of blackholes! After death we become one dimensional chemicals floating around aimlessly! 👽👻🤖🤣😂😅🥹