Neural Networks and the 2024 Nobel Prize in Physics - Sixty Symbols

[@abigailcooling6604]

Maybe the real issue with the Nobel Prizes now is that the old categories/rules don't really reflect science in the modern day. There needs to be one for Maths and Computer Science (which obviously didn't exist when the prize was created), and they should be able to give it to groups or at least more than 3 ppl. Science is done across many disciplines and involves far more ppl now than it used to, so the prizes we give ought to reflect that.

[@lenno15697]

Math has the Fields Medal & Abel Prize. Computer Science has the Turing Award.

[@TheMap1997]

Doesn't mean Nobel can't create their own version. ​@@lenno15697

[@leonsteffens7015]

Wym “we”?

[@tkonan]

Algorithms are not science, but tools of science. Of course, this silly controversy arises out of the ludicrous narrowness of the Nobel categories. Computing wasn't a thing in 1895, but Maths was. Fit for purpose?

[@xthomprya]

I think the difference between the Physics prize and the Chemistry prize is that for the Chemisty prize, computer science was used to solve very difficult problems in (bio)chemistry. But for the Physics prize, it was old and well known physics that led to breakthroughs in computer science, so the arrow is pointing in the other direction.

[@ianstopher9111]

Your example of Chemistry and Physics illustrates the prevalence of CP invariance.

[@ASLUHLUHC3]

Yeah, machine learning is not yet tied to any breakthrough in physics.

[@iselgrey9332]

waiting for the Fast Fourier Transform to win the nobel prize in physics

[@peterfireflylund]

Gauss died a long time ago.

[@crowlsyong]

@@peterfireflylund Wild that Gauss figured it out (Fourier transform and fast Fourier transform) and didn’t even publish it.

[@iselgrey9332]

@@peterfireflylund it would have gone to cooley and tukey but they passed away (this century). Arguably FFT has been much more influential for physics than backpropagation and the whole boltzmann machine thing. And for computer science too. Yet they got no nobel prize for it. It is also weird that an argument for allowing this is that "things are multidisciplinary now, so we give this prize to other disciplines". It's still not a nobel prize in physics, call it a different thing, categorize things properly, this is science ffs. Do it right.

[@thequantumworld6960]

​@@iselgrey9332 I'm a huge fan of Fourier analysis, use the FFT almost on a daily basis, wrote a book about how Fourier transforms connect metal music and quantum physics (for one), and would heartily agree with you that a Nobel prize for the development of the FFT (an algorithm :-)) would have been in order, but... "It's still not a nobel prize in physics, call it a different thing, categorize things properly, this is science ffs. Do it right." There are exceptionally strong interconnections between statistical mechanics, information theory, thermodynamics, and the foundations of computing, to the extent that it's often difficult to tell where one starts and the other finishes. How do you categorize what's physics, what's computing, and what's mathematics in this case? Similarly, with quantum information and quantum computing -- is this physics or computer science or mathematics? Or all three (and more)? How do you categorize these "properly"? The *wrong* way to do science -- **absolutely the wrong way to do science** -- is to force arbitrary categorisations that limit scientists' (and mathematicians' and engineers') ability to cross disciplines. The true innovations happen when boundaries are crossed, not when everyone is forced into arbitrarily-defined categories. Philip (Moriarty, one of those speaking in the video)

[@RaysDad]

Physics and chemistry are essential to making a tasty soufflé, so I'm expecting a future Nobel Prize to go to a couple of chefs.

[@busybillyb33]

I'm still at a loss here. Physical concepts underpin almost everything in any field of study. Prof. Moriarty's analogy here could equally apply to other higher systems such as the collective behaviour birds in flight, or ant colonies, microorganism colonies and so on. Should people studying those be eligible for physics prizes?

[@thequantumworld6960]

Why not?! If statistical mechanics methods/models provide key insights into those systems -- as they did in the context of Hopfield's and Hinton's work on neural networks -- then why shouldn't that be in the domain of physics? See, for example, "Revisiting Lévy flight search patterns of wandering albatrosses, bumblebees and deer", Edwards et al, Nature 449 1044 (2007). Is that physics, biology, statistics, or mathematics? Or is it a really neat mix of all of those disciplines? I'm not a fan of prizes in any case, and even less of a fan of the disciplinary tribalism that they seem to foster (not least in this comments section)! Philip (speaking in video) P.S. There was a great paper in Nature Physics not so long ago on the fluid dynamics of ant colonies...

[@goldnarms435]

Probably so. Or expand the categories. But using your example, no work on ant colonies would be suited for a noble prize. 🙂

[@hive_indicator318]

If it was foundational for learning something about physics, yeah

[@goldnarms435]

@@hive_indicator318@hive_indicator318 Then, in that case, there is no issue because there is a category for physics.

[@youngphilosophy6178]

@@hive_indicator318You’re an idiot. The whole point was that other people have made ground breaking discoveries in condense matter physics and they didn’t win it.

[@halocemagnum8351]

Idk. Imo it really did feel like a reach to me from the perspective of an ML person, because they picked two arbitrary discoveries that really weren’t all that foundational in AI, and latched onto those, because one was made by a physicist and the other was based on statistical mechanics. Yet the merit of the citation was based on all the OTHER actual foundational stuff done in ML. Personally If we’re going full AI bandwagon, I would have dropped Hopfield, and done Hinton, Benjio and Lecun, since those are the people who did the most work laying the foundations for modern Deep Learning. If they wanted controversy they should have just given it to the g-2 guys, at least then the flame war wouldn’t have engulfed my field as well. But alas, Stockholm has spoken, ML is Physics!

[@LifeSizeTeddyBear]

5:10 Hah, "ZZ9 plural Z alpha". Nice Hitchhiker's Guide To The Galaxy reference

[@hive_indicator318]

Phil is my kind of nerd

[@berulan8463]

Thank you for pointing that out. The qoute tickeld some of my neurons, but they don't came to be fired.

[@Michaelonyoutub]

Learning statistical physics was truly eye opening. You could reduce extremely complex real systems into a bunch of simple systems, apply statistics, and get out surprisingly accurate and real results, that can apply extremely well to the real world when adjusted. When hearing about the problem of making AI that can think like a human, solving complex problems, and not like a normal computer solving simple problems we tell it to solve, I always wondered if you could not just break down our brains in a similar way as we break down the physics of particles with statistical physic. Why couldn't we have some kind of simple neuron model, it simply accepts input that either we feed it, or input from some other neurons, and using that input, determines its outputs to some other neurons. Then you just need some way to figure out how to determine how all the neurons react and respond, and make that do something. The rise of the concept of neural networks seemed so obvious when I was first started hearing about it, though I honestly would have never thought up all of the mathematical tricks people have come up with over the years to make them run so fast and efficiently, and make training them so easy, and that stuff is the true hard part.

[@ai_outline]

Using physics as an inspiration to have breakthroughs in computer science should not be worthy of a Nobel in physics, but a Nobel in computer science (Turing award for instance). There was no breakthrough in physics… none, zero, nada.

[@thequantumworld6960]

Information theory, stat mech, thermodynamics, and the foundations of computing are deeply, fundamentally linked, to the extent that an advance in one often drives an advance in the other! (See, Feynman's "Lectures on Computation", for example). Where do you, personally, draw the line between computer science, information theory, and physics? The key issue here, I guess, is that awarding the Nobel prizes in distinct disciplines just serves to entrench disciplinary boundaries that often make very little sense. Philip (Moriarty)

[@8beef4u]

Phil is flat wrong here. With his reasoning many discoveries are suddenly “physics”

[@thetommantom]

I like how he says he gets to make a video with Mike lol 2 great communicators

[@MrLewooz]

Computer science should have a separate Nobel price... It's (compared to the other) a "new" field....

[@123FireSnake]

ohh yeah but the nobel peace prize exists, gimme a break at least computer science is actual science and worth awarding

[@1daft_]

i've wondered if their trust/endowment grows enough that they could opt to add one. Heck even if it was prizeless many people would be proud to have it.

[@oasntet]

The Nobel committee has invented new prizes. They're technically not Nobel Prizes, but that doesn't stop every economist that has won one from claiming they're a Nobel Laureate...

[@MrLewooz]

@@123FireSnake I give you a break.

[@lenno15697]

Turing Award

[@stephendemone]

The opinions in this video are not surprising given how much these men rely on academia to feed their families and (possibly more importantly) their egos.

[@jacksonstarky8288]

As someone who graduated with an interdisciplinary degree in cognitive science almost 25 years ago... and then never pursued graduate studies with it... this is very interesting to me, both for the interdisciplinary aspect and for its connection to machine learning and AI. Science has become so specialised over the last century or two, simply because it has grown at such a rapid pace that it has become impossible for someone to be a science generalist and know any meaningful level of detail about it all; there's just too much data, information, and whatever other term you might use. I have tried, with what time I have outside of my "day" job (I work nights keeping shelves stocked at my grocery store), to keep up with major developments everywhere, which is why I love Brady's channels and those of some of his former guests who have gone on to become KZbin hosts in their own right.

[@ParadoxProblems]

"The core of the prize is absolutely physics." Because he has a "physicist mindset". Thats just means nothing.

[@hesgrant]

Love the othello board for the analogy :) great video!

[@jjpp1993]

If that’s so… why not biology?? I mean…. neural networks

[@hedgehog51]

oooh been waiting for this one to drop

[@duncanhill4434]

I can’t wait for ChatGPT to win the Nobel prize for literature

[@steveyd78]

Love Phil's Nemesis t-shirt!

[@andhag]

Came into this video thinking that it's not physics, and after watching I'm now convinced it's not. Great job.

[@ASLUHLUHC3]

@@thequantumworld6960 If ML were used to solve a long-standing physics problem, uncover a new law of nature, or revolutionize how experiments are conducted, it might qualify for a Nobel Prize in physics. You'd expect the recognition to emphasize the fundamental discovery enabled by ML rather than just the algorithmic or computational advancement itself.

[@aL3891_]

i mean its an amazing discovery, and these guys are mega big brains.. but i definitely see the point of people saying "its not physics" because the actual resulting thing is a computer program, an algorithm.. i guess you can argue that everything is physics in the end and it feels a bit like the pro-argument is arguing a bit :) i say this as someone who has been writing code for a long long time The real problem is that there isn't a nobel price in computer science imo.

[@jonascarlsson5025]

I'm quite interested what you think about the Nobel prize in chemistry going to AlphaFold, if that is your stance. That is quite clearly a computer program, and one that is based on machine learning at that.

[@rachel_rexxx]

Ohh its literally the same video. Well... Thank you for confirming that Computer Science ⊆ Physics ⊆ Mathematics.

[@Luco129]

Wow what a good narrative. Simply explained and very insightful.

[@QuantumHistorian]

And with this prize, the Nobel in physics becomes as meaningless as when the Peace one was given to everyone in the EU, and the literature one was given to Bob Dylan.

[@D1ndo]

So, why bring up the Ising model and then not explain anything?

[@YuanLiuTheDoc]

I mean, CT (computer tomography) received the Nobel Physics, right? Don't know about the general public back then, but physicists were excited for that win. Maybe the problem is social media:-)

[@ASLUHLUHC3]

That was the Nobel Prize in Physiology or Medicine

[@davidrenton]

i don't think as someone who is a a Computer Scientist myself, that this should have won the Physics Prize, there should be a seperate catergory for Computing

[@xwize]

But how else would the physicists be asked to drone on about topics outside of their direct area of study

[@Craznar]

The video explained exactly how the research was directly related to physics. Computing doesn't have any function stand alone - it always serves another discipline.

[@xwize]

Then science has no function standalone. It only serves to guide the engineering. Ridiculous comment.

[@Craznar]

@@xwize That may be true, but the Nobel prize is awarding science prizes - even if science has no function stand alone.

[@BlueLightningSky]

@@xwize Has it ever entered your mind that science could be done for science's sake?

[@Stephen_Lafferty]

Thank you, Brady and team, for knitting together these disparate channels for this synthesised Nobel Prize award! There is no Nobel Prize for Mathematics, famously, and Nobel did not anticipate the advancement of mechanical and physical computation to the point at which it supports/underpins many separate disciplines. Does this series of awards mark the watershed between Human/AI-assisted achievement? What happens when an AI 'discovers' a Nobel-winning event without human oversight/direction? It makes me think of Jocelyn Bell Burnell and Quazars, when Anthony Hewish received the prize instead of Bell Burnell.

[@crowlsyong]

AI generated thumbnail? Presumably cuz the content relates to AI?

[@seandzanda2038]

Back propagation seems like the final, logical thing to do. Is there some other method theoretically?

[@greggashgarian8360]

At U of Nottingham what is the precise value of a gadzillion?

[@cwstevens71]

Was Juan Garrahan the voice of Steven Hawkings computer? Sound just like it.

[@schm00b0]

Why would there be hidden layers?

[@bierrollerful]

Thanks -Oslo- Stockholm, for giving us the University of Nottingham crossover episode we've all been waiting for.

[@viltvalt1057]

[@krissp8712]

Isn't it Stockholm that awards the Nobels? 😅

[@bierrollerful]

​@@krissp8712 You're right. I only knew of Oslo because that's where recipient of the peace prize is selected and awarded - and I assumed the other prizes were awarded there as well.

[@theodoornap9283]

everything is physics. Chemistry is just physics with extra steps, biology is just chemistry with extra steps, etc etc

[@theodoornap9283]

and physics, of course, is just applied math

[@lucidmoses]

If they wanted to fix this, All the Nobel Prize people need to do is hire people that know about Physics to be the judges. :p

[@godofmath1039]

Did AI make the thumbnail? 😅

[@RuhrRedArmy]

"The Alfred" "The Nobel"

[@N0Xa880iUL]

All this is hinting me is that the Standard Model of physics is the best we can do experimentally.

[@FergalByrne]

AI also won the Chemistry prize for protein folding.

[@bobcabot]

ja since there is no science without computers at all anymore possible the whole shabang is more about an old generation still in charge but not willing to retire...

[@Cossieuk]

Given how much cross over there us with all fields of study is this type of thing going to become the norm.

[@mattwillis3219]

Nice to see even the nobel prize isnt immune to big data, whoops we sent the world broke on ai better give it a prize

[@michaelrubinstein4088]

Am I the only one that noticed the "cosmic bs" book?

[@Danyel615]

It doesn't sound like a reasonable discussion if you describe people disagreeing with you as "whinning". And now all of the responses from Prof. Moriarty are gone (?). Ok!

[@syjwg]

Alfred Nobel's will indicated that the awards should be granted in the fields of Physics, Chemistry, Physiology or Medicine, and Pease. A sixth prize for Economic Sciences was added in 1969. The advancements in neural networks (combined with computer science) are perhaps more related to a new category. Biology?

[@Exaskryz]

KZbin showed 60 views... on Sixty Symbols! My day is made

[@pierQRzt180]

I want to see the reaction on the physics one, because the prize practically went to a CEO (hands on CEO but still)

[@patentpendulum]

I am from physics, and I support this.

[@seanoconnor1984]

There are chains of assumption in neural network research that go back unbroken to the start. I don't like the lack of scientific methodology. Persistence research can yield practical results though, such as LED lighting.

[@kingplunger6033]

I think the problem is that most people judge it through the lens of today and what the current situation is. The Nobel Price is for the foundation decades ago, though and back then it was basically physics. Does it fit perfectly ? No. But it kind of does and that's enough for me. I would have liked if someone else got it, but its still a valid choice.

[@appa609]

It's a defensible choice but it's motivated mostly by the Nobel Commitee's desire to keep the prize socially relevant in a way a lot of "pure" physics hasn't been. They see this as their second Gandhi opportunity

[@michaelblakey7794]

I find this quite unfortunate, there are plenty of pure physics projects to give this to. We have the Turing award for these sorts of things. My background is PhD in computer science and yes I love the recognition to our field, but this is the wrong way to do it

[@davidrenton]

@@thequantumworld6960 have a seperate Prize for Computer/AI related discoveries , it seems a stretch to lump machine learning under Physics and deprives Physics of a winner and relegates Computing to a sub class. I've studied CompSci, worked in IT all my life, never considered it to be part of the Physics field because it's not. If we wish to be pedantic, then nearly anything can be classed as Physics, likewise Chemistry. I understand they wish to recognise AI, and that maybe there isn't the same competitive field within Computing that merits it's own prize, but this seems off

[@michaelblakey7794]

@@thequantumworld6960 looking back at previous years, perhaps better wording is domain specific. I’m no expert on what should or shouldn’t be considered.. but entangled photons vs general AI algorithms, I know which one sits better with me! Personal opinion of course

[@TheIgnoramus]

I think you’re over focusing on rewards given from institutions. It’s just recognition and a payout. It matters little to the world at large, and even less towards the work done, and the value the individuals get out of it. Also, 4:30.

[@thequantumworld6960]

​@@michaelblakey7794 But those AI algorithms stemmed directly from statistical mechanics. (For one, they're not called Boltzmann machines for nothing!) It's a beautiful example of a physics mindset and methods being brought to bear on a problem outside the traditional boundaries of the subject. The true innovations in science tend to happen at the boundaries. Parisi was one of the winners of the Physics Nobel prize in 2021 for his work on spin glasses, which is exactly the topic that underpinned Hopfield's work in developing the associative memory network that was highlighted by the Nobel committee. (See the blog post linked in the video information for Hopfield's own view of the role of spin glass physics/stat mech in his work.) So if it's OK for Parisi to win the prize, why is it somehow verboten for Hopfield? Is it the application of the physics outside the traditional "bounds" of physics to which you're objecting? Philip

[@michaelblakey7794]

@@TheIgnoramusI do agree, my problem is that there are other places for this, and by extension those who might of got the recognition if that had happened. For instance, David Patterson could well be argued for changing the computing landscape, was this a Nobel prize in physics.. no, Turing award.

[@TheHarmonicOscillator]

Excellent!

[@alapandas6398]

The irony is that most of those critiques are string theorists whose own physical ground is questionable.

[@MoodyMarco-vj3oe]

Did Juan do the voice for Stephen Hawkings's wheelchait?

[@kapoioBCS]

Very disappointing nobel prize this years. They kinda feel sponsored/lobbied by tech companies.

[@JonBrogaard]

They might as well start awarding the nobel prize in physics to people in economics then

[@appa609]

or quantitative finance

[@MelindaGreen]

By similar logic we should also give them Olympic gold metals. The importance of the tools they developed are not the point.

[@danielmichael3610]

the word "similar" is doing some real heavy lifting in your argument there lol

[@MelindaGreen]

@@danielmichael3610 I'm not hearing a refutation

[@hive_indicator318]

What? Those are given for competitions of physical tests. How are they similar?

[@MelindaGreen]

@@hive_indicator318 How is an algorithm physics?

[@MelindaGreen]

@@danielmichael3610 No, I would prefer if you express your thoughts as clearly as I did and not expect others to read your mind.

[@ysakhno]

Double-upload... Yeah, I believe KZbin will _llllllove_ you for that!

[@zadrik1337]

Lot of copium abuse

[@Anar10n]

bruh

[@Ian.Murray]

No thanks.

[@dylangergutierrez]

Just popping in to say 👎 to the AI-generated thumbnail

[@michaelnovak9412]

This is absolutely not Physics! Enough with this bullshit!

[@GeorgePlaten]

Ah really Phil, nobody cares about the controversy of trolls, just talk about the science!

[@alexandru5316]

do not use ai for your thumbnails, please. ai is trained with people's works without permission and its only purpose as you are using it is to not pay a person to generate assets.

[@juliusEST]

Stockholm syndrome moment

[@harriehausenman8623]

i still rather prefer the interviews with english speakers. Moriarty seriously needs subtitles.

[@WyrdieBeardie]

Is this really physics? ABSOLUTELY! I'd recommend checking out E.T. Jaynes. He did thermodynamics before moving into Information theory and machine learning. Largely ignored in his time, he fully embraced Baysean statistics. He was ahead of his time, and he knew it. 😆

[@pianoerk9827]

first

[@candyland195]

Delete this i definitely won

[@pianoerk9827]

@@candyland195 nope, sort the comments

[@ibrahiymmuhammad4773]

Hahaha

[@Just_a_user3]

second

[@naveenjayakody2047]

third

[@Madzarzour]

The Nobel prize in chemistry and physics were really computer science prizes. Both used AI.

[@barefootinbriefs]

0:34 Is it me or does this guy strike you as Sir Stephen Hawking? 😳😳

[@candyland195]

First