musk is willing to spend on A(G)I research. simons funds CS + math heavily. there are a few real cases. but rare.

yeah but those are all short-term useful research fields

@Slereah understand what youre saying, but not exactly... that is, they are not exactly all short term/ applied focus. eg A(G)I is what google calls a "moon shot"...

People love funding AI research and quantum computing

It's all the rage

whomp

Classical GR hasn't been cool since the 80's tho

bombed that interview...7am is not a good time for me lol...so many brain farts >.>

It's why I've been mostly applying for string theory stuff

It's the closest to something well funded in what I like

@Slereah good to hear (theres something nearby) gotta make compromises. as long as you dont read Hossenfelder youre gold :P (am ¼ thru her book... got inspired to start by her recent tumultuous blog lol)

Oh trust me the compromises haven't started yet

That's when I'll start applying to experimental stuff

there's tons of "compute cross sections for experiments" thesis

Analyze data

All that noise

compute cross sections is still theory ain't it

analyze data to find signals is experiment tho

Depends how you do it

if you find the cross section via experiment, I guess I wouldn't call that "compute"

I'd call that "verifying the cross section via experiment"

you can do numerical simulations for it

@Slereah btw how about a homegrown bell test? :) its been on my mind lately... always looking for collaborators on that... only about 2 decades :P

That's still theory tho

These talks worry me. How do physicists make money at all?

like I did numerical simulations for my PhD, it was theory :D

Not big on numerical stuff

@Nick Surprise!

@Nick no money

I kinda hate coding?

And here I am

Doing it for a living

OoO

Goodness, look what's on the HNQ

@Slereah maybe go into string theory, that's one of the places that still does a lot of by hand calculations afaik

@enumaris That's what I'm mostly aiming for, yeah

@Nick quite a few (high qualified!) regulars in here had to go from physics → software/ coding

Don't do it, it's terrible

@vzn i imagine they're now simulating their life's work on maTLAB

just die instead

I went from science to network management, and that was really good fun ...

pays the bills™ dont quit your day job™ :o

@JohnRennie Well the world don't move to the beat of just one drum

and what might be right for you

May not be right for some

@Semiclassical Written by the same guy who asked this

I guess the rest of the lyrics don't really apply here

@Slereah and in the future, you go back to science and win the nobel prize for... network theory of strings maybe..

"work to live vs live to work" o_O

The point is there are lots of other IT jobs that offer a challenging and interesting career. You don't have to cut code for a living.

@Slereah another point in support for time travel

@Slereah Didn't you get a degree in CS?

I do feel you on the hating coding part but doing it for a living though

there's very few IT-jobs I'd consider interesting and they're probably harder to break into than physics really

@SirCumference Sold my soul to the silicone devil

@danielunderwood Sometimes I feel like the only one who likes programming :/

I mean you know

Programming can be fun

But as a job you're not really doing anything interesting

It's always the same shit

@SirCumference I like parts of it. I just feel that most of what I do doesn't have any technical challenge, which just isn't fun to me

especially when there a visuals

people keep suggesting data science to me

and AI

@danielunderwood I see

The challenges in programming as a job are mostly "Why does that library not work"

Not "how do I make this algorithm"

and at the level of "working with data" I can dig that

@Slereah That's a shame :/

but at the level of "everyone wants to work with big data!!!" I find it tiresome

@Semiclassical it has a large scientiific component and is a great mix between data + science + ... (capitalism)

I wouldn't mind doing like

AI work

Might be fun

There are occasional challenges, but that's rare. There are jobs out there that are more focused on the challenging parts though

Like agent programming

Cloud computing, high Performance computing, Big Data... everyone wants to have a pie chart and present it too :/

That kind of stuff

But it's even less likely than a thesis

There ought to be jobs out there for the more challenging/cool parts

The most challenging part of my job algorithm-wise is just "extract data from that database according to some dates"

Most money in academia comes out of milking students. Let's be honest about that.

@Slereah SQL could do that easy :/

Well not directly

But yeah it's boring and dumb

@Slereah 1 possibility is getting a research engr position, something like coding in a scientiific lab

@SirCumference that's what I'd like to counter. That's not always the case considering what most governments are investing in defense as opposed to education.

Sure, but I don't have a degree in AI

So unlikely to get hired for it

public sector was a thought myself

@Slereah It saddens me that so few people care about theory. I mean that's where all the cool stuff is

@Slereah No one does. You do a PGDM for that.

Not that AI's hard, really

It's basically linear algebra and calculus

and statistics

but most of the jobs tend to have minimum qualifications along the lines of "must know something about govt process already"

which...how tf would I know that?

I had a job doing football match prediction

It was basically all Poisson distributions

@Slereah You confuse AI for ML and DL which are its supersets.

Football is all Poisson

@Nick DL?

I think it depends on your location. Certain places are hiring just about anyone with a math background for AI/ML, but I don't know how the situation is in France

@Nick That's basically all they are math-wise

Deep Learning. Subset of Machine Learning.

@Slereah "AI" is not hard. "AGI" is one of the hardest (not even known to be possible)...

Oh I'm not saying it's not hard to get results

@vzn What's AGI? Is it the theoretic super-AI?

But the basic idea behind it isn't hard

too many acronyms

@Nick artificial general intelligence, ie near human level, its what musk et al (deep mind etc) are chasing

Yeah. Not happening without digital humans happening first.

@Nick there are very recent impressive advances into curiosity...

With all the predictive analytics data scientists are pushing into stock prediction, I'm surprised no one has modeled a brain that can predict my next thought. (should be landmark achievement of this or the next century)

@vzn the mars rover??

@Nick it has a lot of algorithms, some nearly ML maybe... havent heard that it uses much ML. it has to be predictable so maybe they avoid that stuff. am sure they are using lots of ML on the ground to analyze the data.

@Chair Screenshot of ASCII is not real ASCII.

@vzn Fooled by Randomness by N. Taleb made it seem everything was Monte Carlo engines and crazy math. Whatever he did in his trading days, I'm sure someone can do more easily on google colab today.

@Slereah I think a lot about algorithms and never about libraries in my job.

@ACuriousMind you're a lucky man

@ACuriousMind you're no true scotsman programmer, then

Really I think a lot of my job would be easier if I coded everything from the ground up

@Nick great writer. Mr Black Swan, did you read that one? ML has advanced heavily in very few years. it does have a downside eg in finance. algorithms shouldnt exactly be blamed a lot for 2008 etc, the bigger problem might have been what could be called something like "junk finance"

But then that wouldn't be nice for whoever has to maintain it

Just make the whole website in C!

That will teach 'em

I hope you like sockets

@Slereah websites are quite the minefield with their libraries. I've tried making them from the ground up and they always end up coming out terribly

@danielunderwood Oh I'm not saying it wouldn't be horrible

But it would work

At least until I pass it off to some poor schmuck

Good luck!

@EmilioPisanty Oh, I can rant about undocumented and cryptic legacy code with the best of them :P

fortunately my next contract is at least not web development

COBOL undocumented and cryptic legacy code? I'm shocked

@Slereah It's mentioned on the cover, yeah, no haven't yet. Algos will get better hopefully.

But data stuff

@danielunderwood The mistake there is coding a website :P

@ACuriousMind We should all do it like people did in the 90's

Make a word document

Save it as an HTML file

And put it on the FTP

Omg, that iS an option!

QED

@Nick a terrible option

@ACuriousMind I didn't choose to write a website...it chose me!

but you can

@EmilioPisanty Trust me I've tried hard to get less boring coding jobs

And @ACuriousMind what exactly do you do again? Internal ABAP stuff?

@Slereah better than Jekkyl to say the least.

There was one cool looking

@Nick I've found Jekyll to be mighty friendly

Making math libraries for some military company

Had the perfect profile for it really

But I didn't get picked

@EmilioPisanty that's because you can spell it right.

episanty.github.io/publications <- auto-generated from the whatever-it's-called json-like data format

I feel like Jekyll could be quite useful for the right purpose. I've tried to use it with tex and things come out less than ideal

@danielunderwood with tex?

why?

yyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyy ?????????????????

Well mathjax

(generally curious)

Not like preprocessing a whole tex file or anything

@danielunderwood oh

@danielunderwood I do static analysis of ABAP code (both the specific analysis as well as the framework for it)

like, say, this? episanty.github.io/publications/arxiv.1808.05193

@danielunderwood Some templates for LateX on web should help.

that worked just fine

That looks quite good...I have no idea what my issue was

you just double the slashes and everything works fine github.com/episanty/episanty.github.io/blob/master/…

Did you have to manually add mathjax?

Also I like how your publications are all accompanied by a figure and they're pretty much all better than any figure I've ever made

The fundamental polarization singularities of monochromatic light are normally associated with invariance under coordinated rotations: symmetry operations that rotate the spatial dependence of an electromagnetic field by an angle $\\theta$ and its polarization by a multiple $\\gamma\\theta$ of that angle. These symmetries are generated by mixed angular momenta of the form $J_\\gamma = L + \\gamma S$ ...

@danielunderwood ::grin::

TBF four of those vignettes are figures prepared by co-authors

but also TBF there's plenty of other figures in the papers themselves

Well, I'm going to hang out here a lot from now on. That's for sure.

@danielunderwood you must master the art of latex art

But you didn't have to add a library or manually insert the scripts for mathjax? I think I jumped through some hoops trying to get that to work

@Slereah oh, hell no

@danielunderwood ah, yes, obviously

why is the fact that the eigenfunctions of hermitian operators span the entire space is important in quantum mechanics?

ahh I'll have to look at your repo and revisit it...though I have been thinking about just trying to generate html from jupyter notebooks

@Yashas because the space needs to be defined as a functional space.

@Yashas you can form a basis with it

and decompose wavefunctions into those eigenstates

@danielunderwood yeah, it's all in the repo

exactly

@Yashas Because it means that every state can be expressed as a linear combination of eigenfunctions, with the square of the coefficient (if the state is normalized) being the probability to measure the corresponding eigenvalue. Isn't that neat?

@Slereah I've made exactly one figure in tikz and that's because it was boring enough that it was feasible

I've made whole diagrams in Tikz

terrible idea really

@EmilioPisanty lol, I have made none. How do you manage such a feat?

is it like a necessity or just a good thing?

You kinda have to do it line by line

@Yashas I mean it's a mathematical theorem

btw, I have actualy taught TeX/LaTeX several times. Last in a college hall, during SEP.

@EmilioPisanty you could have your own art gallery of figures! Especially combined with some of your answers I've seen on the mathematica SE

All operators that are [mumble mumble] give a basis of eigenvectors

if we had at least one operator who's eigenfunctions would form a basis, would it suffice?

@Nick github.com/episanty/PhD-Thesis/blob/master/8-Spin-HHG/Figures/…

@EmilioPisanty oh my

the question is actually why is it important not whether it's true or not

@danielunderwood that's kind of the plan with the github-pages website

well it is practical for calculations

@Yashas Suffice for what?

if at some point I have the time to build it

@danielunderwood you have inspired me to get back on deviantart with this.

@ACuriousMind would QM break if no operator's eigenfunction would form a basis?

it's true that all hermitian operators' eigenfunction span the space but is this a very critical requirement of QM?

@Yashas I don't understand the question. There's Hermitian operators on every Hilbert space, and it's a theorem that they have eigenbases.

@Yashas Kind of a hard question to answer

If you don't have Hermitian operators, you don't have a Hilbert space.

That's like saying "would physics break down if 1+1 wasn't 2"

I.e. you're not doing standard QM.

Whatever you're doing might be equivalent to it, or it might not be.

Well you can have non-hermitian operators, but then the eigenvalues aren't real

@ACuriousMind that's only if you stick to definition alone.

So they're not measurable

@Slereah um, my statement exactly.

@Slereah they don't have to be real, at least

Wait Hermitian operators are required for a Hilbert space?

you can have non-hermitian operators with real spectra. But they're weird as heck

begin by saying no, follow inductive method

I think there was some attempt to do QM without hermitian operators?

Like just using symmetric operators

But it never went anywhere

@danielunderwood "required" is a weird word. If you have a Hilbert space, there are Hermitian operators on it. There's no "requirement", it's just a consequence.

@Slereah There's some stuff about PT-symmetric Hamiltonians, which is a strict weaker property than being Hermitian

bye guys, I have to hit a textbook. It's been a while.

yeah

the pt-symmetric property is what I had in mind, to be clear

@ACuriousMind ahh I think I see. I misinterpreted your wording

the context was: proved that eigenfunctions of hermitian operators span the entire space and then a statement followed "this is a very important result for quantum mechanics"

the question was why it was important

@ACuriousMind does anybody actually care about PT-symmetric QM?

that theory btw is part of Streater's big list of LOST CAUSES IN PHYSICS

I think "Because it means that every state can be expressed as a linear combination of eigenfunctions, with the square of the coefficient (if the state is normalized) being the probability to measure the corresponding eigenvalue. Isn't that neat?" answers it well

I mean, PT-symmetric QM simulations on photonic systems are cool as heck

PT-symmetric operators are cool. PT-symmetric QM is not

@Yashas The most important reason for a physicist is that you can just do $\psi = \sum a_n \psi_n$ and do your computations on this and pretend it's a real proof

it makes most proofs a whole lot easier

@EmilioPisanty I guess the people publishing about it do? :P

To expound on the above: I think that PT-symmetric systems, when they appear, are interesting

But I don't see any compelling reason to think of them as telling you about QM

Now there is a line claiming "this property is essential to the internal consistency of quantum mechanics"

Isn't every part of a theory essential for its internal consistency? :P

@Yashas I think quantum collapse might not work out if you couldn't express it like that, too?

I think someone's just trying to sound profound without having a precise technical meaning in mind :P

I hate that kind of question

So this is pretty much the first I've heard of PT-Symmetric QM, but does that mean you'd have to choose a position basis and lose the abstract algebraic formalism of normal QM?

googled this last friday and found this Introduction to PT-Symmetric Quantum Theory/ Bender arxiv.org/abs/quant-ph/0501052

I mean imagine this scenario

You have two observables $A$ and $B$

> In the past few years it has been recognized that the requirement of Hermiticity, which is often stated as an axiom of quantum mechanics, may be replaced by the less mathematical and more physical requirement of space-time reflection symmetry (PT symmetry) without losing any of the essential physical features of quantum mechanics.

@danielunderwood no

deep breath

You measure $A$

The state collapse into some eigenstate of $A$

The simplest case I know is to consider a hamiltonian like $H=p^2+e^{i x}+e^{-2ix}$

If you measure $B$ but you cannot build that state with the span of $B$, what happens

That's a complex potential and therefore $H$ is non-hermitian

however, if you do a parity operation and then complex conjugate, then $e^{inx}\to e^{-i n x}\to e^{i n x}$

Like say you have the 2-dimensional Hilbert space

$|+\rangle$ and $|- \rangle$

Then some state $$\psi = \frac{1}{\sqrt{2}} | + \rangle + \frac{1}{\sqrt{2}} | - \rangle$$

You measure $A$ and it collapses to $\psi = |+ \rangle$

But this isn't an eigenvalue of $B$

Then the measurement of $B$ will be $B |+ \rangle = 0$

Unitarity is lost

All is doomed

that's the most QM-specific reason I can think of

Oh yeah just defining the Hamiltonian like that makes sense. I was kind of thinking of an equivalent to $H = H^\dagger$ abstractly...but most QM I've encountered we're defining the position Hamiltonian anyway. Oops

And that reminds me of a question I had before I was in this chat and didn't know if it was appropriate for the main site: are there known conditions under which you can "factor" the Hamiltonian like the normal process with the harmonic oscillator?

In the case of interest for me, that Hamiltonian was arising in a convoluted way from a classical stat mech problem

how do I go about showing that eigenvalues of a Hermitian operator which has a continuous spectrum are real?

It did not have a direct interpretation as the Hamiltonian of some QM system

Isn't QM on the level where stat mech should be irrelevant?

Like I've always thought of stat mech as macroscopic quantities

eh. The mapping was a mathematical one

thermo is more the macro

stat mech is the bridge between QM and thermo

You took the stat mech system, expressed it as a statistical field theory, and then wick rotated to get a quantum system

but in stat mech you make the assumption that all micro-states are equally likely so you've pretty much precluded having to calculated wave functions and the like there lol

And looked at the resulting wave equation

The energy levels of said wave equation translated back into the energy levels in your grand partition function

So for instance the classical minimum of the potential told you about the ideal gas contribution to the system pressure

As I said, it was a mathematical mapping rather than one with a neat physical interpretation

So questions about the physical relation between QM and stat mech aren’t really pertinent here

Is that the type of thing people typically look at in grad school or was it special to what you studied?

> Now it is time to grow up. We set $k_B = 1$. This means that we measure temperature in units of energy.

I don't know why I'd expect anything else

Temperature is a bogus unit anyway

There's literally no reason to not measure temperature in Joules

I remember a recurring thing in our thermo class was to ask what temperature really was

Different ways to define it

I don't think we even covered it as a Lagrange multiplier in there

Though usually it's like... $$T = \frac{\partial S}{\partial E}$$

I think

Or the inverse, i forget

Yeah I think that was one of the main ones. Also something about average kinetic energy I believe. I think there may have been another definition, but I'm not sure what it was

Also what's the deal in thermo with writing expressions like $\left( \frac{\partial F}{\partial x} \right)_{y, z}$? Isn't that implied by a partial derivative?

@danielunderwood Can be tricky if you're not sure everything is independant

since in thermodynamics, $P$, $V$ and $T$ are not independant

nice/ brief overview of PT Symmetry advances by Hook edge.org/response-detail/27165

Ahh we never covered anything where that wasn't the case, so it always seemed a bit redundant to me

@danielunderwood it was a research project, so no it wasn’t standard

Also if my daily mail is anything to go by, you could fund yourself with a student loan...with just enough interest that you'll never pay it off!

Yeah I'm not funding 3 years of PhD on my lonesome

I don't really have 80.000€ handy

I mean I could sell the house but that's a bit drastic

Yeah I don't think it would be my choice either

womp, took a look at SAS text analytics

not bad, not bad

they have a python wrapper for their packages so the api should be relatively straight forward

I don't even have a house.

Suckeeeer

check out that equity

@Slereah I thought it's a flat, not a house? :P

SAS text analytics sounds like something that would cost an insane amount of money

@ACuriousMind It is a flat, yes

w/e

2 and a half hours of meetings coming up... woohoo?

not the salient part

Is a flat like an apartment you own?

Well a flat is an appartment

And I do own it

@danielunderwood well my company is buying the whole SAS package, the text analytics would be an add on to that package...

they would have to do a server implementation of it so...yeah, buttload of money going to SAS lol

I once went abroad to attend a conference, which provided free housing, and I stayed there over one month, because my dorm was broken and I hated it.

It is neat that they have a python lib though...I thought all SAS stuff was in their own language

Also SAS really seems to love their science PhDs from the job listings I've seen

They wrote a python wrapper around their packages

I think it's normal since python is so popular nowadays

@danielunderwood Scandinavian Airlines? ;-)

it'd be a pain for me if they didn't write that wrapper for example, cus the rest of the pipeline I built is in python lol

@Loong need lots of PhD researchers to optimize those flight paths!

I'd imagine that airlines may have pretty strong operations research departments though. Lots of stuff to be optimized

there's a lot of flight data collected

but I dunno how much of that data is handled by the airline or if most of that data is for the aerospace engineers building those aircraft or what

I'm sure they have a bunch of passenger, cargo, and ticket price data as well

@danielunderwood Ah, that's why my gate always changes three times when I have to change planes.

If working in the IT industry taught me anything it's that airline companies just hire an endless stream of shitty companies to write their softwares with underpaid interns

Underpaid interns is a feature!

Write bad software so they can get another contract to write new software!

@Slereah ipht.fr/Phocea-SPhT/ast_visu_spht.php?id_ast=109 18-19 version of ipht.fr/en/Phocea-SPhT/ast_visu_spht.php?id_ast=267

It's a feature, not a bug

@bolbteppa Thanks!

Might send one to that CFT guy

and the QCD too I suppose

send something to the GR guy too

he'll be sad otherwise

why French-speaking universities often don't have English-version websites?

zey are french

GR guy?

I don't see any GR one

@Slereah good opportunity to find one then :D

don't worry, I already send applications to every random dude who works in GR

Would be nice if the Bruxelles university posted more theory thesis

So far nothing new!

Does Bruxelles university post any thesis?

yes

One interested me but so far no answers from the guy

@Slereah so cynical! there is some of that but there are very big/ deep optimization problems in airline business that have big heavy hitters behind them...

Sure, but everytime I applied to some shit company

They always gave me the same list of clients

Including a whole bunch of airlines