Renato Renner: Quantum Mechanics Contains Its Own Contradictions

I remember the doubt before launching this podcast.

What if no one listens, what if I'm wasting my time?

If you've ever felt that way about starting a business, Shopify is the partner that

turns uncertainty into momentum.

They power millions of businesses and 10% of all USE commerce from all birds to gym sharks

to brands just getting started.

No straggler left behind.

Shopify's AI tool writes your product descriptions for you, it enhances your photography, it builds

you a stunning store from hundreds of templates.

Forget about the dormitive haze of bouncing between separate platforms.

Shopify puts inventory payments and analytics under one roof with the propriety of the

true commerce expert.

Their award-winning 24-7 support means you're never alone and that iconic purple shop pay

button, it's the backbone of their checkout, the best converting on the planet, turning

abandoned carts into actual sales.

It's time to turn those what ifs into with Shopify today.

Sign up for your $1 per month trial at Shopify.com slash toe, that's Shopify.com slash T-O-E.

It really troubles me that there is this problem, this contradiction.

I had some students who said for psychological reasons not to be able to buy her conserved

checks, this would disturb them too much.

Most physicists apply quantum theory and never turn it on themselves.

It turns out there are scenarios where quantum mechanics leads you to contradictions when

applied to quantum observers.

Your absolute is certain it's heads, I see its tails.

That's Professor Renato Renner of ETH Zurich telling me that it turns out one of these things

must break.

Number one, quantum theory applies to everything, including observers, that seems natural.

Number two, that measurements give single outcomes.

And number three, that reasoning stays consistent.

All of these are already assumed to be true.

We think that the large world is made up of the small and thus quantum theory applies

to everything else.

Yet, we don't know which one of these is wrong and getting rid of any is just as unsettling.

On this channel, I Kurt Gimungle interview researchers regarding their theories of reality

with rigor and technical depth.

Today, we discuss why this drove the professor from many worlds through to what he calls No Man's

Land.

We connect it back to the black hole information paradox, to the limits of probability, and

to why some of his students refuse to continue because these projects are about the fundamental

question of what you are.

Professor, you've proved that quantum theory can't consistently describe itself.

What does that actually mean?

So maybe I should first explain what type of question we are really trying to answer

here.

So what we have learned from quantum theory, or what at least I think I've learned from

quantum theory, is that there is a constraint on what we can know about the world.

So this kind of starts already with the Heisenberg uncertainty principle, which tells us

we cannot really know the position and momentum at the same time, but this is really a constraint

on the knowledge.

I don't see it as a constraint on what we can technically measure because we can individually

measure position very precisely and momentum very precisely, but somehow we cannot have the

knowledge of both at the same time.

So quantum theory somehow very substantially restricts what we can know about the world

in a sense.

And now, this raises the question that when we cannot know everything about the world,

can we actually still do physics in the way we saw it, we do physics?

Or in other words, we can say that usually when we did classical mechanics or electrodynamics

and so on, we kind of assumed that in principle we can know everything about the physical system.

Of course, we knew we will not know all the positions of the planets and every atom in

the world to an arbitrary position, but we always thought in principle it can be known.

So the would in principle be a very powerful physicist who would know all the positions

in momentum, all the particles in the world and then we can make predictions and so on.

And now when quantum theory tells us that apparently there is a constraint on knowledge, we can

turn this into the question, is there a constraint on doing physics in a sense?

And so what does this now mean for me?

It means for me that if I want to, or how can I answer the question whether we can do physics in our world?

I could now say a physicist is itself a physical system.

So I could, for example, try to describe you doing physics.

So you could, for example, let's say, interact with an experimentalist.

I'm not sure you are not doing experiments yourself, I guess.

No, except this right now.

So let's suppose you're interacting with an experimentalist, your experimentalist friend.

And so this experimentalist friend tells you about an experiment and maybe you go to the lab and

do some checks there.

And then you do calculations, let's say, and make a prediction for what he's going to measure.

Now, what I could do is to kind of analyze how you are doing that.

So I kind of take you as a physical system that interacts with the experimentalist,

who is also a physical system.

So I see all you're doing and all your experimentalist friend is doing as a huge physical system.

And then once I do that, I can see how information flows.

For example, I can see, and I can describe that now using quantum theory.

And when the experimental friend of yours tells you that the experimental setup looks like this,

then some information flows to you, to your brain, and you're picking that up.

And you know what is the experiment.

And then you will do a calculation and then maybe you tell him a prediction.

So some information will flow back from you to the friend.

And now, as I said, quantum theory tells us there are constraints on how much information can

flow or how much information we can have.

So I can now ask the question, is there a constraint of how well you can kind of predict

the experiment of your friend just by the fact that everything has to be modeled within,

let's say, again, within quantum theory in that case.

So in other words, I don't just consider how you apply quantum theory,

but I consider you as a physical system that kind of is part of the world.

And therefore, also described by quantum theory.

And I check whether this is now in a sense consistent with how you see the world.

So we have now two perspectives because you would describe,

you would, from your perspective, describe how your friend doesn't experiment.

You would make predictions.

That's your perspective.

And I take the perspective of describing how you describe the, the threat,

how he's doing experiments.

So I'm kind of on a second level in a sense.

I'm describing how you are doing physics.

Now, if you assume that quantum theory is a universal theory,

then clearly it should be able to describe not only the experiment itself,

but also the experimenter and you, how you interact with the experiment.

How you make predictions about what your experimentalist friend does.

So in a sense, I'm now saying that if quantum theory is universally,

it should also be able to describe how we are doing physics.

So the very process of doing physics, and by this, I mean,

to get information about an experiment,

make calculations, solve the Schrodinger equation,

if you like, calculate probabilities and so on,

that's itself a process.

And I describe this process now also within quantum theory.

And surprisingly, this usually hasn't been done.

So in a sense, I would say this is a very natural question.

If you have a theory of the world,

then because we are part of the world and we are physicists,

the theory should also be able to describe how we are actually using the theory

and doing physics.

So it's in itself in some way a recursion.

So we are kind of asking the theory to be able to describe

how users of the theory do calculations and so on.

But that's necessary for universal theory.

So I think in principle, this would already

be necessary in a sense for, let's say, termodynamics.

Because it's a termodynamics applies to everything.

It also applies, it should also apply to us as physicists.

So I could actually take an outside perspective in a sense and say,

I now describe using termodynamics,

how you are using termodynamics to describe your heating or so.

Your condition, whatever.

So you are applying termodynamics,

but I'm not applying termodynamics to this slide

how you are applying termodynamics.

And I model you as a termodynamics system.

And if you assume that termodynamics is universal,

this should work.

And this type of fashion has very rarely been asked.

And could say the thing that comes closest to that

is actually Maxwell's demon.

Because there we kind of describe

and not being operating on a thermal system.

And now we return to quantum theory.

I can ask this very same question with quantum theory.

But I could, in principle, ask this question with any theory.

So if you develop any new theory of the world,

I would always say this is a good consistency check

where the theory is good.

If you claim that the theory applies to everything in the world,

then in particular, it should apply

to the physicists who are using the theory.

So it's a new type of consistency check.

And so what we did in this work that you were mentioning

is to somehow apply this consistency check to quantum theory.

So we were asking is quantum theory

able to describe a quantum physicist applying this,

who is applying the theory to some physical system?

So that's the question.

Now what we found is that maybe I could say it in two stages.

What we first found is that there's

something that is kind of undefined

if you just apply quantum theory.

Because quantum theory is usually just applied

from the perspective of one physicist.

And so we have to ask the question,

what are even the consistency conditions

that we want to impose?

So for example, let me come back to the example I made

before you are describing the experiment of your friend.

Now I'm describing you and how you are

describing the experiment of the friend.

Now one would expect intuitively

that you are the description of the experiment of the friend.

This could be a prediction.

Let's say the friend measures a spin particle

and you predict the outcome of that measurement

will be spin up.

That could be your prediction.

Now I'm analyzing how you get to that prediction.

And let's suppose I come to the conclusion

that you actually indeed predict

that this spin is up.

And I also predict that the system

that you're talking about is in spin up.

Then I would say this looks consistent

because I've now kind of described what you are

describing and everything is kind of consistent.

But now, what if it happens that in my analysis of you,

I find that you think the spin is up,

but actually when I analyze the system myself

of this experimental friend of yours,

I find the spin will actually be down.

That could be because it just applies

quantum theory to a large system.

And now it turns out that for certain,

very involved setups, which are far from standard,

the so-called Wigner's friend setups,

that can maybe explain what they mean,

but they're very artificial setups.

But in these artificial setups, we get contradictions.

So in other words, if I analyze how you are doing physics,

I find that you get a result,

which is not consistent with what I get

if I directly analyze the system that you are analyzing.

So I kind of do two analysis.

I first analyze the system of interest myself.

This could be this experiment that your friend is doing.

I just analyze it myself.

And then I analyze how you are analyzing it.

And I now check whether these two things are consistent.

So one is kind of like a direct analysis.

And the other is kind of via you.

I just predict what you are going to predict.

And now the prediction of what you are going to predict

about the system should hopefully match

what I'm directly predicting about the system.

And that's a kind of consistency condition.

And we first need to ask ourselves

what type of consistency conditions

do we even require?

Because we could just say, maybe we don't require anything.

We already learned that quantum series kind of strange.

So let's just not even assume or let's not hope

that there is any consistency there.

And then there is no problem.

So what we did in this work is to kind of have

a kind of minimal consistency condition.

And the minimal consistency condition

is kind of that if you are very, so if I

come to the conclusion using my analysis of your process

of doing physics, I analyze how you do the calculation and so on.

And I come to conclusion that your calculation

will have the result that with absolute certainty,

the spin measurement will show that the spin is up.

In that case, if I now directly analyze

what the spin measurement should do,

I should also come to the conclusion that it's up.

Or I should at least not to the conclusion

that with certainty it will be spin down.

So that's kind of a minimal consistency.

So it should never be the case that you are certain that

the spin is up.

And I know that you are certain it's up, whereas I'm certain

that the spin is actually down.

That would be a contradiction.

So what you're doing is to say, let's ask for this minimal

consistency.

Let's just say this should not be violated,

this consistency condition.

Just a moment, I want to make it clear,

because people who are listening may think,

well, if you flip a coin, and you take a look at that coin,

the coin says heads, that's in your head.

Okay, so the head is in your head.

And then I'm thinking, okay, what is Professor Renner

thinking, what's in his head?

And obviously I have incomplete information.

So I could say it's 50% heads, 50% tails.

Obviously I can make a prediction and just say,

I'm going to say it's tails.

I was incorrect.

But there's no inconsistency in the laws of physics there.

So you're thinking about something that's extremely ideal

and taking quantum theory,

quote unquote, seriously, people like to use this term

and seriously think, well, whatever it means.

So there's something that's ultimate about it.

That's final, there's no wiggle room here about

just predictors and knowledge and so forth.

I need you to spell that out about the difference

between just making an arbitrary prediction

versus the ideal reason or with ideal rationality

or what have you making a prediction.

Yes, I think this is a very important point

to kind of make the distinction between uncertainty

or incomplete knowledge or contradictory knowledge.

So indeed, if I hear a flip a coin and I could do that,

I have on here, so I flip it, I have the outcome.

And now I could say, I look at the outcome.

So I have certain knowledge about this outcome

and I could now ask you, what's your knowledge

about the outcome?

And I guess you would say it's probably the 50%

or 0.5 probability up or heads and with 0.5 tails.

Now, this knowledge is not the same knowledge that I have

because I know for sure it's tails, I just saw it here.

So in that case, I would say you had incomplete knowledge

and this knowledge is not the same knowledge that I have.

And that's, however, not a contradiction.

Obviously, I mean, that would be very strange.

So why is it not a contradiction?

If I later tell you that it's actually tails,

you would say this was perfectly consistent

if you were just not knowing whether it was heads or tails.

But let's suppose you were actually, for some reason,

because you probably analyzed the way

that this coin was flipped.

You come to the conclusion that it's heads.

You are absolutely certain it's heads

and you correctly apply this theory and everything.

But I see its tails.

Then you would say there is somewhere a contradiction

and this is the type of contradiction

we found in this sort experiment that we were analyzing.

So it's really not just incomplete knowledge,

it's contradictory knowledge.

And contradictory knowledge, I would define it as

there is no kind of additional information

that I could give you that after the update

would make you have the same knowledge.

Because before in this sort experiment,

I could have done that, I could just, or I did it,

I told you its tails.

And then you do an knowledge update,

which is like you hear me telling its tails

and you say, okay, then you believe its tails.

So now you're also certain its tails.

So you just updated your knowledge

and got to the same conclusion as I came.

But if you were certain its heads

and I tell you know its tails,

then you cannot do an update.

You will just run into a contradiction

when you try to apply base rule,

you will basically divide by zero

and get nothing sensible out of that.

Okay, I like this classical analog.

I think it makes it much more clear.

So suppose the person listening

can model the coin 100 with 100% accuracy.

We're in classical mechanics

so we have no Heisenberg uncertainty.

And then they calculate that the coin landed tails.

And then you see it as landing heads

and you report to them know it actually landed heads.

They would think I must have modeled it incorrectly.

But you're saying, no, no, no,

let's assume you modeled it 100% correctly.

Then, and we're gonna get to your three criteria soon.

But then you could say, well,

either there's something wrong with me,

which we've removed by your data.

You're saying like, no, no, no,

there's nothing wrong with you.

You've used math correctly in classical mechanics correctly.

Or there's something wrong with classical mechanics.

Or there's something wrong with what's so and so.

There are a couple escape routes that you can take,

which we'll get to.

But this is the classical case.

It's just to build some intuition for the quantum case.

Yes, yes.

I would completely agree with what you said.

That's a very good summary of the type of contradiction, yes.

So indeed, I mean, you mentioned like we assume

it's very precisely that the experiment

and there is no error.

And because we are,

and I say it's talking about the sort of experiment

we can just do this idealization.

We can say, let's suppose all the agents that are involved.

So I call these agents to different physicists.

They just have a purely perfect information about the system.

So let's say before the experiment actually starts,

before I flip the coin,

everyone gets the very same data about how the world is,

what's the state.

And then we start the experiment

and I flip the coin and don't show you the outcome,

but you calculate it and so on.

But I can make these idealized assumptions

in a sort of experiment and then ask, does it come out well?

Because at the end, it's a statement about the theory,

not about the actual, let's say, experiment with flows.

Just, if I feed into the theory,

all the information I have, what does it give me?

And so by construction of the ZOT experiment,

I give you the full information

and the correct information about the world

when the experiment starts.

Okay, would it be accurate to say

that the classical analog would be something like,

suppose Laplace's demon exists,

then what if Laplace's demon is wrong in its prediction?

Then you would say, well, how does that even make sense?

If you're supposing Laplace's demon exists

almost by the criteria of it existing

of what Laplace's demon is, it has to be right.

So would it be that you're doing something similar

to that before the quantum case?

Yes, I would say this is another way to put it away.

You put it, but the question is,

how would I define Laplace's demon?

I would, I mean, if I define it as it would really make

the correct predictions, then this probably wouldn't make sense

because then by definition it's correct.

But if I define it to be a demon that applies the theory

that we are trying to analyze, absolutely correctly,

then I think the analogy works.

And that's what we are doing.

So we are not assuming quantum series correct,

we're just assuming all these agents

are applying the theory perfectly.

So it would be like a Laplace demon demon

who is perfectly applying whatever theory is.

It is that we want to test, let's say classical mechanics

or classical theory, but I'm not demanding

that this is actually the correct description of the world

because that's what we want to test,

kind of want to test whether the theory is a reasonable theory.

So it's kind of a perfect demon in a sense

with respect to the theory that we are testing.

When I'm wrestling with a guest's argument about, say,

the hard problem of consciousness or quantum foundations,

I refuse to let even a centilla of confusion remain unexamined.

Claude is my thinking partner here.

Actually, they just released something major,

which is Claude Opus 4.6, a state of the art model.

Claude is the AI for minds that don't stop at good enough.

It's the collaborator that actually understands

your entire workflow thinks with you, not for you,

whether you're debugging code at midnight

or strategizing your next business move,

Claude extends your thinking to tackle problems

that matter to you.

I use Claude actually live right here

during this interview with Eva Miranda.

That's actually a feature called artifacts

and none of the other LLM providers have something

that even comes close to rivaling it.

Claude handles interalia, technical philosophy,

mathematical rigor and deep research synthesis,

all without producing slovenly reasoning.

The responses are decorous, precise, well structured,

never sick of fantic, unlike some other models.

And it doesn't just hand me the answers,

the way that I've prompted it is that it helps me

think through problems.

Ready to tackle bigger problems?

Get started with Claude today at Claude.ai-slash-theories

of everything.

That's Claude.ai-slash-theories of everything

and check out Claude Pro, which includes access

to all of the features mentioned in today's episode.

Spell out the difference between being inconsistent

at a local level, so from your own point of view,

versus at a global level.

And which one does your theorem apply to?

So maybe one should understand the sort experiment

that I'm kind of proposing as really just the test purely

of the theory, not of, let's say, the experiment.

It's a sort experiment, which is based

on the assumption that the theory is correct.

And I'm not comparing it to the actual world.

I'm kind of asking the question, is it

the possible description of the world?

So the sort experiment is not really making, in a sense,

the assumption that the world is correctly described

by quantum theory.

It's just making the assumption that quantum theory

is a possible consistent description of the world.

And if I'm doing that, then I cannot really compare

to the experiment.

So I cannot have a direct contradiction

to an experimental outcome, because I'm just not

testing the actual world, in a sense.

I'm just testing the consistency of the theory.

So how can I then check the consistency of the theory?

I can say, generally, a theory is consistent

if different ways of reasoning within the theory

lead to contradictory outcomes.

So this is very general, I would say.

So I could have just a theory that is not even physics.

Let's say some mathematical theory.

And I want to, let's say, prove a statement.

And maybe the theory would be contradictory

if there is one way to prove that the statement is actually

true, and there is another kind of proof

that I find within my set of axioms

that shows the statement is actually false.

Then I've done two arguments, which

lead to contradictory outcomes.

And this is the type of contradiction we get that.

So we have quantum theory.

Quantum theory can now be used in different ways.

And in this case, it's kind of one way

to apply this in the example I had before.

I applied directly to an experimentalist

and make a prediction about the experimentalist.

That's one way of reasoning.

And the other way of reasoning is I applied to you

and started how you make a prediction about the experimentalist.

That's more indirect way of reasoning,

but it should also be an allowed way of reasoning.

And if these two ways of reasoning contradict each other,

then there is this, I would call such a,

let's say, global contradiction in a sense,

because I cannot even tell which one was now right.

I just have two different ways of coming to a conclusion.

And the conclusions are opposite.

This tells me something is wrong with the theory.

But I cannot see a contradiction to an experiment

because I didn't do an experiment.

I'm just theoretically analyzing the consistency

of the theory in a sense.

So I think it's really important to make

this distinction between a sort experiment

where I just think about the possible description of nature.

And I accept the fact that for the moment, I cannot test it.

And then the only thing I can do is kind of do consistency checks,

assuming that this is a possible description of the world,

but I will not test it against actual experiments in the world.

But I can still test where the different uses of the theory

lead to consistent results.

And so that's, in a sense, what you're doing

with quantum theory now in this sort experiment.

And so necessarily the contradiction is kind of a contradiction

between two different ways of using the theory.

But if both ways are allowed ways to use the theory,

then of course we are in trouble because then we have a theory

that somehow has too many, offers too many ways

to come to a conclusion.

And we cannot decide which one is the right one.

Let me ask you something funny.

How do you see quantum mechanics that's different

than your colleagues?

OK, yes.

So that's a very good question in this context,

because I could say, if I analyze how you are analyzing a system,

how could I ever come to a contradiction with me directly

analyzing the system?

That's actually the core point of the whole question,

because you could say, this shouldn't happen.

And this brings in, let's say, another question

of what do we actually do when we use quantum theory?

And what we are doing is we always

cut out the part of the world, which we are actually

describing with the theory.

We are never describing the entire world.

And this is necessarily so, because if we did describe

the whole world, we would necessarily

have to describe ourselves how we are applying the theory.

This would lead to a cycle.

If I describe the whole world, I necessarily describe myself.

And this may somehow be possible, but at least within quantum

theory, you don't have the tools to do that.

Quantum theory doesn't tell us how to describe a system

that is myself, in a sense.

Because I would have to automatically

describe the reasoning process.

And so this leads to a terrible recursion

that is just not dealt with in quantum theory.

So I would say the minimal or the maximal part of the world

you can describe is kind of everything without me.

I cannot describe anything larger than that using quantum

theory.

Now, this is actually a statement.

Some physicists would disagree with,

because if you are a manoeuvre or a Bohmian,

you would say, we can actually take an outside viewpoint

and describe everything.

And in a sense, they are correct, because they

would somehow assume they are outside.

Because by saying we take an outside viewpoint,

they kind of suggest there is another super observer

or something outside who does all the calculations

and everything, like a God-like being.

And if you say, OK, physics can just be applied

from kind of this God's perspective, that would be OK.

But my, let's say, the very basic ingredient

in the, let's say, approach I took is that at the end,

we want to have the minimal assumption

that we, as inhabitants of this world, can do physics.

And that not only God can do physics, which, for me,

is a very natural assumption, but it's very disputed.

So I have this assumption that I want to be able to do physics,

but I myself part of the world.

And if I make this assumption that physics should be done by,

it should be possible for us to do physics,

and not only for an external to the universe being,

and it's not even to what that means,

because if you're a being, then you are, again,

part of the universe, that leads to troubles.

So the question I'm asking is really, can I do physics?

And if I'm asking this question, I have to be kind of modest

and say, I can describe everything except myself using

the current theory.

Maybe there is some future theory that allows us to do recursive

reasoning, but current quantum theory doesn't even

talk about that.

It just assumes we are, like, the description itself,

the physicist is kind of external.

That's an implicit assumption, in a sense,

because otherwise, you couldn't even write down

the state or anything of the system.

Because the state will constantly change

as you write it down and so on.

So this now means that, so if I describe everything

except me in the world, and you also take a maximal

perspective and describe everything except you,

then we still have different perspectives,

because you didn't describe yourself,

and I didn't describe myself, but I described you

as a physical system, and you described me.

And this is the type of setup where we find the contradictions.

So we find the contradictions in scenarios

where there are agents who are not describing

the same part of the world, and they cannot

describe the same part, because they

are themselves part of the experiment.

So one important ingredient to the experiment

is that, at some point, there is a measurement

done on one of the physicists who are taking part

in the experiment.

That's what I meant when I said before

that this is a very special situation,

a very non-daily life situation, that one of the physicists

is kind of going to be measured by another physicist

in a very strange basis, not just in the, let's say,

I mean, a measurement would be, I just

ask you a question, that's also a measurement.

But I could do a very complicated measurement

like a shrewding and cut type measurement.

And if someone did such a measurement on you,

then he has to describe you as a quantum system.

But if you are also a physicist taking part in the experiment,

you will not be able to yourself describe that part

of the experiment, in a sense.

And so by using such building blocks in this sort of experiment,

so measurements of agents, I make it impossible for the agents

to describe everything.

So each agent has kind of a restricted view.

And the view is kind of just the view that we usually

have in daily life.

Like if you talk about the experiment of your friend,

you will not describe the whole world.

You just describe that one.

And so you have a different perspective from, for example,

from me, who decided to, let's say,

to describe you as a physical system.

And if you're now asked to have the same perspective as me,

you would answer, no, this is not possible.

I cannot describe myself.

So in a sense, you're constructing a situation

where it's inherently impossible for the agents

to all have the same view, because they

are somehow themselves part of this whole experimental setup

operations, autonomous, and so on.

When I'm wrestling with a guest's argument about, say,

the hard problem of consciousness or quantum foundations,

I refuse to let even a centilla of confusion remain unexamined.

Claude is my thinking partner here.

Actually, they just released something major, which

is Claude Opus 4.6, a state of the art model.

Claude is the AI for minds that don't stop at good enough.

It's the collaborator that actually understands

your entire workflow, thinks with you, not for you,

whether you're debugging code at midnight

or strategizing your next business move,

Claude extends your thinking to tackle problems that matter

to you.

I use Claude, actually live right here

during this interview with Eva Miranda.

That's actually a feature called Artifacts,

and none of the other LLM providers

have something that even comes close to rivaling it.

Claude handles interalia, technical philosophy,

mathematical rigor, and deep research synthesis,

all without producing slovenly reasoning.

The responses are deckerous, precise, well-structured,

never-sick of fantic, unlike some other models,

and it doesn't just hand me the answers.

The way that I prompted it is that it

helps me think through problems.

Ready to tackle bigger problems?

Get started with Claude today at Claude.ai-slash-theories

of everything.

That's Claude.ai-slash-theories of everything

and check out Claude Pro, which includes access

to all of the features mentioned in today's episode.

What is the minimum amount of observers

required for this no-go theorem or whatever you want to call it?

That's a very good question, and I'm

40, I don't have an answer.

I would say certainly three are needed,

but in the experiment we found we need actually four,

and I'm not sure I can reduce it to three.

OK, so is it important that my modeling of you

has your modeling of me in it, or is it

enough that my modeling of you has my modeling of you monitoring

person three, monitoring person four,

but doesn't have to come back to me?

Yeah, so the way we try, so this is also important.

So it's, in the sense, we try to avoid exactly such recursions

because then we would again be back in that situation.

So the experiment is set up in a way

that the individual, let's say, uses of quantum theory

are users that you could do without having

to describe yourself.

And it's not obvious that this works.

So we had to play a long time with arranging these

actions in such a way that it works.

Perfect.

OK.

And yeah, at the end it's, so because there's

often a criticism saying that, oh, at the end,

I actually describe myself indirectly

by describing the other agent.

But actually, to sort out this criticism,

we actually wrote the software to get

with one of my collaborators, Lydia Del Rio and Núria

Núric Alieva, we once kind of decided

to write the software that actually

implements the experiment.

And if we would have such a recursion,

we couldn't actually program it.

So the software kind of proves we can describe it

by not running into a loop.

And it's actually even publicly available.

We put this at some point on the archive.

So if anyone is interested to play with the software.

But this is a question that always came up.

And so we try to kind of convince people

that it works by having this kind of automatic reasoning,

if you like.

So the software allows you to specify what the agents need

to describe.

And then you see that the description

is not running into a loop.

OK.

There is, however, some loop going on,

which we may talk about when you ask me about how I think

we can resolve it.

But it's not the kind of description.

So it's not that, so just to be clear,

if you in the experiment you would describe me,

you don't need to describe me how I'm describing you.

You could just describe me on the only parts that are not,

or not only the parts of me that are not

involved in describing you, for example.

OK.

Now, is it important that the state in your experiments

that the other person measures is a definite state?

Or can I give my prediction of what

you see as a probability distribution or amplitudes,

or just a wave function that hasn't been localized?

Yes.

So we tried in this experiment to actually avoid talking

about states when we phrase the final statement

or the final claim.

Of course, we talk about states in the analysis.

And the reason is that there are very different views

on what it means to say that we have this state.

So there are people who think of the state, of course,

as knowledge and others who think of them as something real.

And so the entire experiment is a bit like,

you can maybe compare it to belts, serum,

where one could say at the end of the bell in a quality

it doesn't talk about states.

It just talks about correlations

that you find in the expectation values.

And this is a bit similar in this sort of experiment.

So in the analysis, of course, we use quantum theory

and I have to involve, I mean, that involves talking

about states or you can also take the Heisenberg picture

and then you talk about operators, if you like.

But at the end, the objects the agents are talking about

are predictions about outcomes.

So you could, the only, let's say, saying we need this

to realize the experiment, we have some initial state

that is known to everyone.

But because that's a state that is kind of known to everyone,

the question of whether it's a systemic or ontic

and all these questions don't arise

because it's then just assumed to be common knowledge

and then everything is the same.

So it's as if we would say, let's start

if it's been pointing in the up direction.

Let me kind of know what that means operationally.

So that's on problematic.

But as soon as I'm going to use state as knowledge,

I'm entering kind of a terror that is a bit ambiguous

because people may have different views on what that is.

So what we try to do is to say,

everything can at the end be phrased

in terms of the predictions.

So the agents have this initial state, they know,

and then they just calculate the prediction,

which is of the type, I have this probability

that this outcome will occur.

And this is not phrased as states.

So we try to kind of not give importance

to the meaning of states and therefore talk

as little as possible about quantum states in this experiment.

People should know that I'll be placing the links

to your work and the full articles in the description

and there'll also be on screen as well.

So for the physicist who's not a quantum physicist,

who maybe is just a relativist

or maybe they're a materials engineer

or something like that, I want you to give an analogy

that I'm forcing on you right now,

which won't work because analogies tend to not work.

But it can give a flavor.

So chess is often used as a way of thinking of physics

and the pieces are like particles

and then they have rules and so forth.

And then you're saying that many physicists like to say,

okay, well let's just take a look at the board.

And you're saying you're taking the perspective

of the player, which is the God's eye view.

But in physics, everything is physics.

It's as if everything is the board.

So you can't jump out and take a player's view.

Okay, firstly that's your critique about God's eye views.

Then what you're saying is let's imagine you're a rock.

You are a rock.

If I may say something about that.

So I wouldn't even phrase it as a criticism.

I would say even if someone believes there is an outside view

and that this view makes sense.

Like the players view, the chess players view.

I wouldn't necessarily, I wouldn't object.

I would say, okay, that's, you can take this,

have this opinion that there is this outside view.

But there should at least also exist an inside view

from the, let's say, chess or whatever,

there are these different characters on the chess,

but from the king's perspective.

Because the king is one of the parts of the world.

But they say, so I'm kind of asking that this inside perspective

also exists.

And I'm not necessarily excluding the exterior perspective.

I think it's not necessary because we don't have

this exterior perspective.

But I think whatever your view is about whether this

gots view exists, gots view point exists or not.

At least the inside view point, our view also exists.

So I'm just against denying the existence of that.

And I want, so my requirement on a physical theory

is that it's usable by us in essence.

Okay, perfect.

And if it's also usable from the outside, that's fine.

So I'm not trying to exclude that view point.

I would only object if someone says the inside view point

is not a legitimate view point.

Then I would say, okay, why are you physicists?

You're part of the world.

So by denying that as a valid view point,

you basically say you're not a valid physicist.

Right.

I remember the doubt before launching this podcast.

What if no one listens?

What if I'm wasting my time?

If you've ever felt that way about starting a business,

Shopify is the partner that turns uncertainty into momentum.

They power millions of businesses

and 10% of all USE commerce from all birds to gym sharks

to brands just getting started.

No straggler left behind.

Shopify's AI tool writes your product descriptions for you.

It enhances your photography.

It builds you a stunning store from hundreds of templates.

Forget about the dormitive haze of bouncing

between separate platforms.

Shopify puts inventory payments and analytics

under one roof with the propriety of a true commerce expert.

Their award-winning 24.7 support means you're never alone.

And that iconic purple shop pay button,

it's the backbone of their checkout.

The best converting on the planet,

turning abandoned carts into actual sales.

It's time to turn those what-ifs into with Shopify today.

Sign up for your $1 per month trial at Shopify.com slash toe.

That's Shopify.com slash T-O-E.

Okay, so most physicists are physicalists

in that they think that the world is ultimately physics

and every other fact,

whether it's about consciousness or about the economy or biology,

it's entailed by the physics.

Okay, if you believe that then you who is speaking

as a physicist are a physical system,

we can even idealize you to something that makes no mistakes

in your paper.

You have a computer that does quantum mechanics perfectly.

Okay, so in this analogy going back to the chess board,

you're a piece on the board.

You're a mystery piece.

I was going to call you a rook before,

but as you mentioned,

you can't actually fully know yourself even in regular physics.

You have some recursion issues.

So you're a mystery piece on some part of the board.

And then you look out and you see ponds

and you look to your left,

you see a bishop and so forth

and you see other pieces.

Then you model, well, what is that other rook doing?

What is it seeing?

Okay, so from this analogy,

what is your inconsistency theorem?

I think one could call it a no-go theorem.

That's fine because we make some assumptions

and say they don't fit together.

They're not all at the same time valid.

Okay, so from this analogy,

what are you saying?

Are you saying that you make a model,

the bishop, the bishop,

the makes a model of this and that

and then you're saying that bishop,

you're seeing a king,

but the bishop actually seems a queen or spell it out.

Yes, I think that's one way to put it.

I could say, indeed, I see the bishop

and I'm modeling what the bishop is.

Or okay, I don't know what the characters were.

Let's say I'm a rook, I'm seeing a bishop

and I'm asking does the bishop see

what figures are there else?

I don't know.

Rocks, knights, kings, queens.

Oh, yes, a knight.

I know the German terms.

That's why I'm not so familiar with the English characters.

I actually played chess

but I never translated them to English.

Okay, so let's say the rook sees the bishop,

the bishop now claims there is a knight

in front of it, in, let's say, that's the...

And it's important that you do not directly see the knight.

Or is that okay?

Yes, or then I could say,

so I know the bishop sees a knight

and then I could also ask myself,

do I also see the knight there

where the bishop claims it is?

And I don't see it there.

And then that's a contradiction.

Ah, interesting.

Okay.

So in the sense I make on the one hand

the direct statement about what I see

and then I also make a statement about

what the bishop sees and the two statements don't match.

And they don't only match in this way

as we discussed with the coin

that one is more uncertain

because that could also be,

that would not be problematic.

I could say,

oh, the bishop doesn't know whether there is a knight

or not in front of him.

And I know there is one in front of the bishop,

but that's not a contradiction.

That's just the bishop has uncertain,

incomplete knowledge.

But if I know the bishop is certain

the knight is in front of him,

but I actually see the knight is not in front of him,

then that's a contradiction.

And so this is the type of contradiction we get.

Yeah, so I think in this case,

the analogy works very well

because it's just about like two type of statements

being part of the world.

So there are all these pieces on the check board.

So we are them.

So we are one of them.

And I can make either a statement directly of what I see

or I can make a statement about what someone else sees.

And of course, I kind of expect these

to be compatible with each other.

And that turns out not to be the case now

in very special situations.

Okay, and now the assumptions that go into this theorem,

many terms has assumptions.

So what are the three?

And which route do you take as being the escape one?

Yes.

Actually, there are three that we state.

And of course, the assumptions can always be grouped

or kind of subdivided.

So this number is not so important in a sense.

But yeah, let me, I think it's a useful group

or a useful kind of fine graining that we have.

So one assumption is that quantum theory,

the way we use it is correct.

Let's say the textbook quantum theory.

Some people now subdivided that into several assumptions.

And it indeed has several parts

because it has at least two aspects.

Because what you are saying is that

in a sense quantum theory is universal.

But universal can be understood in two different ways.

I could say quantum theory is universal

in the sense that I can apply to any system I like,

maybe except for to myself, that's fine.

So I don't want to require that, of course.

But that's one type of universality.

It can take whatever I like, my cup of tea,

my office chair, quantum theory applies to everything.

So that's the first part.

But then there's another type of universality,

namely that everyone is allowed to apply quantum theory

from his or her perspective.

And that's another universality.

Because in principle, it could be that there is only

the external hypothetical external observer

who is allowed to apply quantum theory.

And that would exclude, that would be precisely

the thing that I think would be very strange.

They say, OK, we, as inhabitants of the world,

are not allowed to apply quantum theory.

We have to be external to the universe.

And only then we are really allowed to apply the theory.

But that's kind of what is implicit to many worlds,

or booming mechanics, because they somehow say,

we need to have this external perspective.

So in a sense, this second type of universality

is actually something some people would deny.

And they would say, no, that's not true.

We are not eligible to apply quantum theory

as being part of the world.

So the assumption that going to the first assumption

that goes into the series is this universality assumption

in this twofold way that I just explain.

And that's maybe the most important assumption.

And I think this is also the assumption

that was mostly misunderstood, because maybe that's also

my fault.

In the original paper, I was not even

myself aware that these two sides of universality

are so important.

Use them, because for me, it was very natural,

but it wasn't really emphasized.

But I think this is a very important point,

in particular, the second type of universality.

We can apply quantum theory.

And if I talk to people, I often hear people opposing to that.

I know we are not legitimate observers,

because we are ourselves in superposition.

That makes us not valid users of quantum theory, which

I find very strange criticism, because we

actually want to apply the theory.

OK, so that's the first assumption.

The second important assumption is this consistency

assumption.

So we have to ask or make an assumption about what

do we expect to be true when you describe something,

and I describe the same thing in what sense should it match.

And here one could make very different assumption.

A very strong assumption would be that if you tell me

the state of a system is this particular wave function phi,

then I should also think it's the same wave function.

That would be a very strong assumption,

and everyone who thinks epistemically of the quantum state

would deny it.

So we make a much weaker assumption.

We say that if you make a statement of the form

that you are certain that this particular measurement

outcome will occur.

So like you are certain the spin will be up.

Or you are certain if you look in front of you,

there will be a knight.

So that's a kind of type of statement.

Then we are requiring that it shouldn't be the case

that if I directly check that, or if I make a direct calculation

or prediction that I come to the opposite conclusion.

So if I ask what's your conclusion,

it shouldn't be opposite to mine.

So if you say it's been up,

I shouldn't be certain that it's been down.

Okay, and that's what the whole earlier part

of the conversation was about was about this contradiction.

Right, yes, yes.

So that's kind of the new criterion.

That's what I think was never tested before.

So when people propose theories,

they made all types of consistency checks,

but this check whether two observers

who are using the theory are compatible in that sense

was not something that was usually tested in a sense.

So I think this is kind of, let's say a part,

maybe the no-go theorem is maybe interesting,

but I think the more relevant part of the proposal

is actually to propose that test of a theory.

So this test should be applied

to any proposal of a theory.

It does it lead to this type of contradiction when we apply it.

Can we consistently apply it in that sense?

Yes, okay.

Can I really make sure that someone else

who applies it comes to the same conclusion as I do,

or at least not to contradictory conclusions?

Okay, now do you also see that as a necessary component

as to what a scientific theory is

because science has intersubjectivity?

Yes, I think so.

So for me, communication, in a sense,

is a very important part of doing science.

Of course, one could in principle think of me being alone

in the world and doing science.

And you could say, what would be wrong about this?

But actually, there is even communication involved there

because I kind of, my younger me communicates to me now.

Like I did some experiments in the past,

or maybe some sorts and predictions,

and now I'm verifying them.

And this is also in this, like the consistency should even hold there.

So for example, I could just see my former self

as a different agent and then ask for this type of consistency

and ask myself, if I know at that time,

I did a correct calculation and thought it's like this,

but now I do the calculation directly now

and I come to the opposite conclusion.

That shouldn't happen.

So this consistency requirement, I think,

is relevant for any communication,

including the communication of our former selves to us.

I see that as communication.

I mean, it's a very powerful communication

because it just stores stuff in our brain

and communicates via that.

But it's a form of communication.

If you like, from an information's rating viewpoint,

I'm constantly sending messages into my future

and read them later.

And so in that sense, communication is,

if I take that as part of communication,

then communication is unavoidable to do science

because otherwise it would just be a point in space time

and could not explore the universe.

So I have to communicate between different experimental,

let's say, instances that I'm performing

and whether I'm performing them now

or whether it was my former self or some colleague

or some former physicist who lived on this planet,

shouldn't matter, it's all communication.

And so if I, for example, only read just a book

in quantum theory, that's communication

from early physicists to me.

So we are actually doing, I mean, communication

is so inherently in our scientific development

that it's very hard for me to think about what science

even means without communication, even to phrase a thought,

to have a theory, means I can write it down

and read it later.

And if that's impossible, then I wouldn't really call that side.

So for me, this is almost a defining part of what science should be

that it's possible to phrase things,

to communicate it to others or at least to myself in the future.

OK, universality, quantum theory applies everywhere,

meaning that even what we think of as large classical objects,

elephants and robots or whatever, they're also quantum systems,

yes.

OK, and then consistency, which is what we talked about

for 40 minutes or so, and then the third was what?

Yes, then there was a search, we didn't talk about yet.

This is actually one that is so, in a sense, natural

that it's often not even mentioned,

but it's that if I make a prediction and then,

and I make another prediction, there shouldn't be contradictory.

So in this other assumption, the consistency

assumption, I should maybe have been a bit more careful.

And so actually, maybe I should trace it in the following way.

The consistency assumption is somehow split into an assumption

of how to combine knowledge and into an assumption

that there should not be a contradiction.

So let me explain that.

So the first part is how to combine knowledge.

This is basically just to say that if I'm certain that you are

certain that the spin is up, then I can take this also

as a certainty for me that the spin is up.

So for example, if you tell me that you have made a calculation

and everyone is reliable, I know you're a perfect physicist

and you tell me that measurement will lead to spin up,

then I will also be sure the spin is going to be up.

So that's strictly speaking the assumption C.

So the assumption C doesn't talk about contradiction.

It just tells me that if you are certain about something,

I can also be certain.

And now the assert assumption is now saying

that this should not contradict what I calculated before.

So maybe I calculated already that the spin is going to be down.

And now I have two different predictions.

And the assert assumption, which we call kind of the single

outcome assumption, tells me that if I come to the conclusion

that the spin is certainly up, and I also

come to the conclusion that it's certainly down,

then that's a problem.

So it's kind of the definition of what we mean by contradiction

that we have two opposite predictions with certainty.

Is a fourth assumption that you're allowed to fully

hit the world into a now moment and then

have time steps forward that everyone agree on?

Yes, so there are actually, if you ask the question,

is this now the complete set of assumption?

Then the answer is no, there are more assumptions.

And for example, one assumption, I mean,

even before the one you mentioned is we

can apply logical reasoning.

And you could ask, why didn't I even

phrase this assumption in the no-go theorem?

For example, the logical reasoning means

if I know A holds and then O, B holds,

then also the conjunction A and B is true.

Like, yes, we need that.

But of course, we need that to even phrase,

let's say, the sort experiment.

We need language and so on.

So in a sense, that's an, it is an assumption,

but without this assumption, I couldn't even

start phrasing the experiment.

Yes, yes.

And if that assumption goes, well, then every opponent theory

goes as well.

Right, yes.

So that's just to say that there are certainly more assumptions.

I'm not claiming there are not more,

but there are assumptions that are so basic

that we are kind of making them just

before we even start doing science.

Now, the assumption about, let's say, the foliation in time

is kind of one can see it as part of the assumption

of standard quantum theory applies, but it is an assumption.

So I would just see it as part of this assumption

that we apply quantum theory, because quantum theory tells us

there is a unitary that brings us from time to time.

But it's correct that this is an assumption.

When I'm wrestling with a guest's argument

about, say, the hard problem with consciousness

or quantum foundations, I refuse to let

even a centilla of confusion remain unexamined.

Claude is my thinking partner here.

Actually, they just released something major, which

is Claude Opus 4.6, a state of the art model.

Claude is the AI for minds that don't stop at good enough.

It's the collaborator that actually understands

your entire workflow, thinks with you, not for you,

whether you're debugging code at midnight

or strategizing your next business move,

Claude extends your thinking to tackle problems that matter

to you.

I use Claude actually live right here

during this interview with Eva Miranda.

That's actually a feature called Artifacts

and none of the other LLM providers

have something that even comes close to rivaling it.

Claude handles Interalia, technical, philosophy,

mathematical rigor, and deep research synthesis,

all without producing slovenly reasoning.

The responses are decorous, precise, well-structured,

never-sick of fantic, unlike some other models.

And it doesn't just hand me the answers.

The way that I prompted it is that it

helps me think through problems.

Ready to tackle larger problems?

Sign up for Claude today and get 50% off Claude Pro

when you use my link Claude.ai slash theories of everything

all one word.

Why I'm super excited to speak with you

is that you take an interesting escape route.

You take that quantum theory is not universal.

And I want to know more about that.

And then my question, which I'm going to get to later,

so you can feel free to weave this in.

What the heck does this have to do with gravity at all?

Yes, so gravity wasn't now coming in,

but actually I should tell you that there

is a third way to escape the paradox, which we didn't talk about.

And this is actually now my favorite way to escape.

So far, to be honest, I'm actually changing my mind

over time.

I have to admit that I was even, to be

honest, the many world, some years ago,

that changed my mind to becoming more a patient type person.

And also about this thought experiment,

I mean, this thought experiment has actually

still an impact on my daily life in the sense,

not only that I get many emails every day,

but also that I constantly think about what's the solution.

So for me, to be honest, I have no satisfactory way

to deal with it.

It really troubles me that there is this problem, this contradiction.

And so that's why I'm thinking constantly about that

makes you change your mind, because at some point,

you have some insights, or you talk to people

and think this is a good argument.

And one of the more recent insights for me

was that I realized that maybe all these assumptions

that I mentioned, they are so reasonable we shouldn't

give them up.

I mean, of course, you can say, OK, maybe quantum theory,

we haven't tested it to any precision.

Maybe there's something wrong.

But we have really no indication that quantum theory

doesn't apply to certain systems.

And also, clearly, we should be able to apply it.

Also, this consistency, or the combining knowledge assumption,

is really a minimal thing that you can even talk.

Because otherwise, it doesn't make sense to talk to each other.

If I tell you things, and you just hear them,

but it cannot really make sense of the content,

that's what it would mean if we give up this assumption

about this consistent communication.

And then clearly, we don't want to have contradiction.

So all these assumptions I really want to keep.

But now, actually, our statement,

I would really call it a no-go theorem,

says the following.

It says, if we carry out this experiment

with several agents, so I didn't really describe the experiment

experiment, goes along like someone prepares the spins,

and sits to someone else, and then

reasons what is the outcome of a measurement, and so on.

So we describe an experiment and say,

if we perform this experiment, and then

analyze it using these assumptions,

we will run into a contradiction.

But now, who tells us that we can actually carry out

the experiment?

It could just be that the reason why this is no problematic,

or these assumptions are unproplicated,

is that we can not even start analyzing the experiment,

because it's not executable.

And so far, I was always assuming clearly,

we can principle execute this experiment.

It's very hard to do it, because we need to,

as I said before, we need to do something

that I call a Schrodinger cut-like measurement.

So it would be like Schrodinger cut being a superposition,

and I measure in a basis, in a measurement basis,

that directly tells me there is this superposition.

So the way I would have to do that is basically

undo the whole process.

So in the Schrodinger cut example,

the Schrodinger cut is kind of poisoned,

and this whole process of being poisoned

has to be reverted during the measurement process,

which is, of course, incredibly difficult to do.

Now, so far, I was always thinking,

this is just a technical difficulty,

and there's nothing fundamental about that.

And now, my more recent, let's say, suits,

which also involves gravity, let me now

suspect that this is actually wrong.

We cannot carry out such experiments.

And maybe that's, it's not obvious why I'm saying that,

but there is actually a kind of circle in the argument,

or a loop.

And the loop goes as follows.

So maybe I'm sure you know about the recent developments

in this research that is concerned

with quantum reference frames.

So people are asking the question,

if I'm, for example, measuring an orientation in space,

like a spin particle, what resources

do I need to do this measurement?

And clearly, I need some reference.

I need to know what is up and what is down.

And now, if I do a more, if I analyze a more complicated system,

not only a spin, I need a larger reference frame,

because intuitively, the reference needs

to be larger than the system that I'm measuring.

So if I only have a spin as my reference,

then my measurement of another spin particle

will be very uncertain.

So I need something larger than a spin to measure a spin.

Now, and the gravity considerations

and kind of tell me that not only spin requires a reference,

but everything that could in principle

be measured requires a reference.

Maybe we can later talk about why this is the case,

but let's just assume every degree of freedom

that we measure requires a reference.

So for example, I mean, it's obviously

if I measure like two spins, I need a large reference,

and if I measure only one spin, because the reference

gets kind of involved in the measurement process

and it's a bit more uncertain, because in order

to measure the spin, I have to let the reference

and the spin interact with each other,

and this will kind of slightly degrade the reference.

So let's now assume that whenever I measure a system

of a certain size, let's say, L, I need a reference

that is larger than L.

Now if you analyze the sort of experiment that we have,

then we have kind of a circle of, in a sense, measurements

or places where we need a reference.

So we have four, or let's say, just for simplicity,

we have only three parties or three players

or agents in this experiment.

We actually have four, but let's call them Alice Bob and Charlie.

Now let's suppose Alice needs to do an operation on bulk.

Then Alice needs a reference that is larger than Bob,

because otherwise she cannot really do a measurement

on Bob as a quantum measurement on Bob as a system.

Now let's suppose Bob needs to make a measurement on Charlie,

then he needs a reference that is larger than Charlie,

but if now Charlie needs to make an operation on Alice,

he needs a reference that is larger than Alice.

But if you are now in this kind of weakness trend type setup

where you do measurements on the entire agent,

you need to also measure or apply the measurement

to its reference frame.

So the reference frame is kind of part of the agent,

which is also natural, like if I see a spin up,

this is relative to me as a reference.

So if now Alice has to surface a reference for measuring Bob,

Bob has to surface a reference for measuring Charlie,

and Charlie has to surface a reference for measuring Alice,

then we have kind of a loop.

And if everyone has to be larger than the other,

then this somehow doesn't work.

So we don't need to a problem with kind of the size

of the reference.

Each one has to be larger than the other,

and this is impossible.

So this is just a suspicion I have now,

so I don't yet have a full argument

that really convincingly says why these references

have to be larger, so I'm still struggling with this idea.

But my current hope is that I can show

that the experiment is not executable

because of these requirements of the reference frames,

so that each part needs a reference frame

that is larger than the other.

And this will make it impossible to do it.

So just to give you an idea to kind of do a Schrodinger

cat experiment, to undo that measurement,

would require me to control all the 10 to the 23,

or even 25 atoms that the cat consists of.

And that would require that I need an enormously large reference

frame, maybe as big as the universe or so.

And so the only thing that is difficult,

but now if I do a second measurement on that universe,

that the measurement needs to be even bigger.

Yes.

And so at the end, if I have a loop,

this is clearly impossible.

So that's one suspicion that I have,

or that's currently the root I'm trying to see better.

I can prove that this will lead to an impossibility.

And the conclusion would then be,

actually, there is no contradiction

between the assumptions that I just mentioned,

but the problem is we simply cannot build these experiments

where the contradiction would arise.

So the contradiction doesn't arise

because the experiments where they would arise

cannot be built for fundamental reasons.

That's a new root.

The economist covers math, physics, philosophy, and AI

in a manner that shows how different countries

perceive developments and how they impact markets.

They recently published a piece on China's new neutrino detector.

They cover extending life via mitochondrial transplants,

creating an entirely new field of medicine.

But it's also not just science, they analyze culture,

they analyze finance, economics, business,

international affairs across every region.

I'm particularly liking their new insider feature.

It was just launched this month.

It gives you, it gives me, a front-row access

to the economist's internal editorial debates,

where senior editors argue through the news

with world leaders and policymakers

in twice-weekly long format shows,

basically an extremely high-quality podcast.

Something else you should know about

is that if you go to their app,

they not only have daily articles,

but they also have long-form podcasts

with their editors and writers.

This is also available online.

Whether it's scientific innovation

or shifting global politics,

the economist provides comprehensive coverage beyond headlines.

As a toll listener, you get a special discount.

Head over to economist.com slash T-O-E to subscribe.

That's economist.com slash T-O-E for your discount.

Now, large can mean large in many respects.

Do you mean large in mass, large in volume,

large that it persists longer in time?

And even if so, whatever I just said,

that still doesn't have to do with gravity.

If you're dealing with quantum mechanics,

you're going to have large systems of large masses,

large composite systems.

I don't see where gravity enters.

Yes, I didn't say it.

So this is also not obvious.

But so by large, I meant kind of the Hilbert space dimension.

So as an quantum information series,

I would say in terms of the number of cubits of a system,

which is just a different way of saying

the Hilbert space dimension up to a logarithm.

Now, it's not obvious how this fits to gravity.

But there is actually a connection,

as you know, between size in that sense.

And for example, size of a black hole.

So you can, you know, a black hole has a certain information

content, so to speak.

Of course, that now depends again on the view

one has on quantum gravity and there are different views.

But one can associate an entropy to a black hole.

And that kind of corresponds to how much information

is contained in there.

And so if I now talk about the size,

you can say this is ultimately like the size,

I mean, the size in the way I talked about,

is kind of related to the size of a black hole

that I could build by putting all that information

as close or by kind of collapsing that information

into a black hole.

Now, by this relevant or by the way,

I'm on to make this connection for reference frames.

So one kind of other word that I did

completely independently of the theory

of this quantum mechanic sort experiment

was an information theoretic analysis of this experiment

where you throw information into a black hole

and then try to retrieve it from the whole king radiation

that is emitted from the black hole.

So this is often discussed in the context

of the information paradox.

So this is this famous thing like black hole surface

emitter that Patrick Hayden and John Preskill once proposed

that if you throw a notebook into a black hole

and then you collect the radiation from the black hole,

you can kind of retrieve the content of the notebook.

But the radiation will also look kind of completely

scrambled.

And so one can now ask what, I mean,

there are contradictory views on what actually happens.

And there's no consensus about that.

So some people would say, indeed, one can retrieve the notebook.

And others would say, no, then this is just thermal radiation

coming out of the black hole.

You will have no chance to retrieve the notebook.

So what we found in this work that I did

with one of my former students, Ginzao Wang,

is that actually both views can kind of coexist.

And they're just the different conclusion

is come from the fact that people are implicitly

assuming different reference frames.

So you could think of those who say

there is just thermal radiation coming from the black hole.

They implicitly assume there is no reference

that with respect to which you can make sense

of the information emitted by the black hole.

So let me make a very simple example.

Let's suppose I have just a source.

Let's suppose there is nothing in the universe,

a priori, it's completely empty.

But for some reason, there is a source

that puts new stuff into the universe.

Let's say it's been particle.

Now, let's suppose the first spin particle arrives.

And one now asks the question, is the spin upward down?

Then you would say, OK, that's not even defined,

because there is nothing else in the universe

with respect to which I could define whether it's upward down.

But now, let's say this process continues

and it meets further spin particles.

Then, of course, we can compare them and say, now,

relative to the first spin particle,

the next one is up.

And so on.

And now, this is an analogy of how one

could think of the Hawking radiation of a black hole.

Let's suppose you just create one single black hole

and it radiates Hawking radiation.

Then this first radiation would be like the first spin

that was just created.

So the radiation is there, but we don't have a reference

with respect to which it makes sense.

Sorry, just now, right now you're talking about one black hole

that just emitted one bit of Hawking information.

Yeah, let's say it even emits more,

or you could say, let's say, even the first Hawking radiation

quantum that would be emitted.

It will basically have a direction.

Yeah, you could even imagine, let's say,

we built a black hole in this otherwise empty universe

and then the first spin particle comes out of the black hole.

We wouldn't even be able to say what that means

that it's been partly, it's up or down.

But let's now suppose we build, like,

we do an experiment where we in parallel

build identical black holes, let's say, thousands.

And now we could check whether, like,

we could kind of take 999 of them

and take the spin particle that comes out,

the first spin particle that comes out of each of them

and put them together into a large like stick,

let's say they're all spin particles,

they all have a small direction.

If I put them together, it's kind of a larger arrow, if you like.

And now I could say the thousands black hole,

the spin that comes out there,

could now be compared to that stick that I have,

that now serves as a reference.

So I kind of take 999, like,

the first ones just serve as creating the reference

and then I say, now let's check whether the direction

now makes sense.

Yes.

And so what we actually did calculations

and found that indeed, if you create many black holes

in parallel, we could kind of regard take, let's say,

if you have many, or a sufficiently large number,

we could use them as a reference.

And with respect to that reference,

the other black holes would emit very ordered

for radiation, it no longer looks random.

But if I just take one black hole

and don't have this reference,

then it looks completely random.

So in other words, a black hole kind of creates

new type of information, like the spin in a universe,

where there was never a spin before.

So let's say if I have a universe without any direction

and I create the first spin,

this just doesn't make sense, direction doesn't make sense.

But if I have many of them,

I can start to make sense of direction

because I now have many spins that were emitted

and I can use them as a reference for the other things.

So in other words, what we found is,

like the type of information or the stuff

that comes out of a black hole,

if I just had one single black hole in the universe

and nothing else would look completely random.

It's like just something new,

like a new spin that is created in a universe

where there was nothing that breaks symmetry.

But then if I kind of use many black holes,

many authors as a reference,

then a new black hole that I create

and let Amy or like a collector at the Asian,

that radiation will be for respect

to the authors have a lot of structure.

So in the sense, I can use some of the stuff

to as a reference for the author,

like in this spin example.

Actually, something that I was going to bring up,

I said it was a no-go theorem,

then I corrected myself instead of was something different,

it was almost like an inconsistency result.

And then you said, no, I see it as a no-go theorem.

Actually, I still see it not as a no-go per se.

I still see it as an example of an inconsistency theorem.

And then I was thinking,

the black hole information paradox

is another type of inconsistency theorem.

And then when you said that you found a way

out of your inconsistency theorem,

I was wondering if it could be applied

to other inconsistency theorems

like the black hole information paradox.

And it seems like you have already.

Yes, that's what we did,

but I would have called both of them no-go theorems.

But, okay, one could also call them inconsistency theorems

because, okay, technically what they are is just,

there's a set of assumptions.

Like in our case, quantum series universal,

we can communicate or we can make sense of other,

like if you know something,

I can adapt that knowledge.

And there's no contradiction between my predictions.

So these are three assumptions.

And I'd just say these three assumptions,

if applied to an experiment or not at the same time,

cannot at the same time devalue it.

So in that sense, it's an OGO theorem.

If there are assumptions that cannot simultaneously hold.

So to be compatible with what I said before,

I should say a force assumption is,

we can actually carry out the experiment.

But these four assumptions are then kind of incompatible

with each other.

And of course, I could kind of always give priority to one

and for example, say this assumption that tells me

there is no contradiction.

Like as you didn't at the same time say the spin is up

and the spin is down, I give it a different status,

I'd just say the other three assumptions imply

that this assumption, there's no contradiction assumptions

wrong. So in other words,

the other three assumptions imply there is a contradiction

and then it looks like an inconsistency theorem.

But then it's like a different ordering

of the assumptions in a sense.

So I would say that no, go see it was very neutral

in terms of priority.

It just says, there's no priority to one of the assumptions.

Just take them as a set of possible assumptions,

but they cannot all be true.

You have to give up at least one of them.

Like in Bale's theorem, you have like locality,

you have re-choice, and one of them must be wrong.

Now what in a physicist retort for instance,

as Sean Carroll may say something like,

look, if you're going to be so pedantic

that we can't even write down a black hole

emitting a particle with a spin

because we'll have to pre-think something extremely practical.

Like how do we measure what is up or down?

Then I, is Sean Carroll,

because Sean Carroll had to talk about

how God is not a good hypothesis.

And he said, look, I can write down a theory

where there's just a single electron in the universe.

Right here, I wrote it down.

I, is Sean Carroll, cannot write down that theory

according to you because I would need to have something

that measures, how do I know what the charge

of the electron is?

So if you're going to be so pedantic,

then that completely limits any thought experiment

or any ideal toy scenario.

Okay, I'm pedantic, sounds a bit negative, I would say.

But indeed, in that sense, I'm very pedantic.

So I, but I want to be operational.

I will call it being operational.

I want to say that many of the programs we have in,

let's say many of the discussions that arise in physics

are because their discussions about concepts

that are not well-defined in the sense

that we don't know what they would mean

if we actually dig the experiment.

So for, I mean, this, with the spin is now an example.

If I have a universe or like this black hole example,

this is an example telling us, look,

as long as we don't specify

how we measure what is the reference,

we actually get contradictory statements

because in whatever you do,

you will either implicitly assume we have a reference

or you don't.

And as long as you don't talk about it,

we don't know what assumptions you make.

So it's no surprise that we get different conclusions

because some people now implicitly assume

there was this reference on it.

It's not, so I want to be explicit and say,

look, if you do the experiment,

we have to explicitly make the system

a measure interactive to reference.

If you don't know the nature of the reference,

whether it even exists or whether it's not there,

then we will just come to different conclusions.

So I think many of the conflicts

can be resolved in that way.

So that was also the message of my paper of Chinsao Wang

that neither of the attitudes is right or wrong.

They just were not specific enough

to make the question even decidable.

And I want physics questions to be decidable

and decidable for me means in principle,

I could do an experiment that decides it.

But if I want to make that claim

that there is in principle experiment,

then I should try to model everything

that is potentially relevant.

And of course, the challenge is to find out what is relevant

and what is just, let's say, technology

that is not, doesn't have to be modeled.

So I'm not claiming that we should really model all the detail.

I mean, I'm a quantum information series

and quantum information series really love abstraction.

So we want to get rid of everything that is unimportant.

So in that sense, I don't want to be pedantic in the same look.

If I have a qubit, I don't want to be forced to say,

which atom is it that realizes the qubit?

This is a superconducting qubit or is it a trap?

I mean, I want to be on that abstract level

and say, I don't tell you my statements

or anybody through independently of what the atom I'm using.

So I want to leave away unimportant stuff.

But I think in quantum information,

if there may be a victim of our own success,

I think quantum information has the enormously successful

with this abstraction, we could develop quantum information,

quantum computing on this very abstract level.

And everything worked.

But maybe we went now too far and said,

OK, we completely got rid of the notion of space and time

and reference frames and so on.

And maybe we have to rethink that

and reintroduce the concept where they may be relevant.

And so my own lesson, let's say, from working in quantum information

and now getting into gravity is that the abstraction went maybe too far.

And an abstraction where we just abstractly think about the spin

without actually talking about the reference

that we need to measure the spin

is not a good abstraction to answer certain questions.

Of course, it always depends on the question.

But if you run into troubles like with this sort of experiment,

we have to ask, are we on the right level of abstraction?

Did we abstract the way certain things that are really essential?

And so my feeling is now that abstracting away reference frames

was a mistake, at least for this type of experiments.

We need them.

And so it's indeed a matter of taste or of, yeah,

this is kind of a scientific, maybe approach,

which of the elements we consider as unimportant

that we can safely leave them away, now idealize models

and which things were actually important

and we were wrong to just leave them out.

And my feeling is really that reference frames

are so important that we cannot leave them out.

They play such a fundamental role in our understanding of space

time that dropping them will just not or a theory

that doesn't consider them will not give us the right answers.

And so I think in the case of the black hole information paradox,

I really see that as an example where there are two opinions

based because people make different implicit assumptions

about stuff that is not explicitly

modeled like the measurement of the hooking radiation.

They just talk very abstractly about the hooking radiations.

It's some quantum state.

But if I measure it, I need a reference.

And this reference has to come from somewhere.

I should also, I also need to know what's the unitary

that this creates the radiation.

People usually say, okay, there's some unitary going on.

But in principle, we would need to do quantum process tomography.

So we would need many black holes to do actually the tomography

because you cannot do tomography on one single system.

And that's another issue that if you create many black holes,

they will actually be correlated

because they are in the same space time.

And so this will lead to some warm holes and stuff.

So there are other difficulties that arise

that I think are very important to take into account.

So my, I think for me, this is very strong,

I have a very strong feeling that one needs to be patented

in certain aspects like including reference frames

and that we will otherwise miss something

or we have missed something otherwise.

Of course, I don't know what it is that is missing.

And so this is just a guess that it's reference frames.

So I'm not claiming this will be the right answer.

But at the moment, it's the approach I'm pursuing.

Most of my best ideas don't happen during interviews.

They come spontaneously, most of the time in the shower

actually or while I'm walking.

Until I had flawed, I would frequently lose them

because by the time I write down half of it, it's gone.

I tried voice capture before like Google Home

and it just cuts me off in the middle.

It's so frustrating.

Most of my ideas aren't these 10 seconds sound bites.

They're ponderous, they're long-winded.

And I wind around, they're discursive.

They're five minutes long.

Apple notes, even Google keep the transcription.

They're as horrible.

But plot lets me talk for as long as I want

and there's no interruptions.

It's accurate capture.

It organizes everything into clear summaries,

key takeaways, action items.

I can even come back later and say,

hey, what was that thread I was talking about

regarding consciousness and information?

In fact, this episode itself has a plot summary below

and I'm using it right now over here.

My personal workflow is that I have

their auto-flow feature enabled

so it sends me an email anytime I take a note.

Look, the fact that I can just press it

and it turns on instantly like right now

it's starting to record without a delay

is extremely underrated.

This by the way is the note pro

and then this is the note pin.

I have both.

Over 1.5 million people use plot around the world

if your work depends on conversations

or the ideas that come after them,

it's worth checking out.

That's plot.ai slash TOE.

Use code TOE for 10% off at checkout.

So wait, what is the specific claim?

Is it that a thought experiment is not meaningful

if it can't in principle be realized?

Or is it that the thought experiment,

what the thought experiment is about

doesn't exist or something

because something can be not meaningful

but also exist.

I mean, unless one wants to go anti-realist

and say that all that exists is just somehow operational.

I don't think it's actually operationalism,

that's a particular philosophical strain called operationalism

which I imagine you're not an operationalist.

Yes, so I have my personal list of requirements

to a good sort experiment.

And one very important entry on that

is that I want to be able to do the experiment

in principle.

So there should not be a fundamental reason

that excludes the thought experiment.

And now in order to answer the question,

whether it's really feasible in principle

or this is a question I can never answer

definitely unless the experiment is actually done.

And usually the thought experiments are of course

about stuff you cannot easily realize

but my hope in, let's say for the specific experiment

that we were talking about with the agents

who contradict themselves is that

this is something we could actually do

in the not so far future.

If for example, we have quantum computers

who would actually emulate the physicists

because instead of having a physicist doing the reasoning,

I could just have a computer doing the reasoning.

And it has to be a quantum computer

because it has to be a fully controlled system.

A quantum computer is a system

where we control all degrees of freedom completely.

So there I really have, like it's important to me

that I have this, like let's say a scheme in mind

that in principle I could build the experiment

and then ask myself what goes potentially wrong.

And then I would maybe find that, as I said,

maybe there is a missing or there's a problem

with the sizes of these quantum computers.

But if someone tells me a fundamental reason

why the experiment cannot be done,

then I would kind of say, yeah, in that case,

it's no longer a good sort of experiment.

When we learned something that I'd proposing it

and finding this fundamental reason,

but then it's kind of settled in a sense.

So for me as a good sort of experiment is something

where I know how in principle I would do it.

And otherwise, it's not well specified.

I mean, it's the same with, let's say, Maxwell, Demon

and all these other sort of experiments.

If, for me, they are interesting because in principle,

they can be built.

Or now we, I mean, there are claims

that they are built, we are not yet completely there.

So that's an important point for me.

I mean, I'm talking about salt experience in physics.

Of course, if I were mathematician,

I would have different requirements.

But I want to learn something about physics.

So it's just a salt experiment because we cannot do it

in practice, we don't have the technological means.

And as a theorist, that's for me a very important tool

to make progress.

Now, if something is only meaningful

if we can, in principle, realize it,

well, then thoughts about, well,

what if the some constant of nature was so and so?

There's no machine that even in principle

can change one of the constants and say the standard model.

Or many fine tuning arguments

or anthropic reasoning are about this.

If we're going to limit our thought experiments

to what we have somehow in principle access to,

I imagine that greatly limits.

Yeah.

So I think if there is, at the end of the fundamental reason