I am curious when scores in some areas go backwards, despite some clear large gains (and did they try that experimental scaffold again?)
The part where they remove the experimental scaffold is the only thing in this table that goes backwards from Opus 4.* to Mythos, right? (Lower is better for MSE.)
Some things did get worse from 4.5 to 4.6, which I agree is interesting.
What I mean is that for branches to affect each other, the branches must become completely not separated: every difference between the branches must be erased, including the ion and also the atoms in all the observers. This makes communication meaningless, because there’s no longer anything the sender knows that the receiver doesn’t also know.
Inter-branch communication while preserving separate branches shouldn’t be possible under the standard many-worlds interpretation, because it violates linearity.
“Reality is amplitudes flowing between configurations” is helpful for understanding conceptually why MWI (and quantum mechanics generally) predicts this won’t work. For branch 2 to influence branch 1, the two branches’ amplitudes would have to flow to the same configurations. So suppose Alice#1 (with high probability) receives a message, and is still aware that she is #1 and didn’t compute the message herself. What does this mean for Alice#2?
It means that Alice#2 must have somehow arrived (with high probability) at the exact same physical configuration as Alice#1. If you believe, as is the orthodox view on LessWrong, that mental states are physical states, this implies that some physical mechanism has made Alice#2 forget, completely, down to the positions of the individual subatomic particles in the brain, that she was ever #2. The two Alices have split and then merged again.
Why can’t #2 keep her memory and still change #1′s outcome? Because then her branch’s amplitude would be flowing to a different configuration! Specifically, a configuration that includes an Alice who remembers being #2 and witnessing the computation.
And separately, there must be some mechanism for branch 1′s amplitudes to flow preferentially to the correct result of the computation to “meet up with” branch 2. This is even more unsatisfying, because it seems to require branch 1 to do the computation we hoped to avoid.
But could the amplitude flows of branch 1 depend on the state of branch 2? Not according to standard QM/MWI. This is where linearity comes in: as far as we can tell, the future state of a quantum system in superposition is always exactly equal to the sum of the future states of each individual configuration multiplied by their amplitudes: . ( just means “wait an amount of time equal to t”.)
The contribution of branch 1 to this result – where the amplitudes of branch 1 flow – is , which doesn’t change depending on what b is, or what turns out to be. Each branch evolves and splits into more branches, separately, except amplitudes add whenever branches meet. This is the entirety of the interactions between branches in Everettian MWI; there’s no room for any other kind of influence.
Conversely...
...if we discovered that reality obeys a nonlinear quantum mechanics (all available evidence so far says this is false, but you never know!), then communication between branches becomes possible:
quotes from Polchinski (1991)
I find that forbidding EPR communication in nonlinear quantum mechanics necessarily leads to another sort of unusual communication: that between different branches of the wave function.
This same paper also has some interesting comments on the “branch selection” issue:
I have ignored previous branchings of the wave function, describing the macroscopic observer and apparatus as though they started in a definite state, as would be acceptable in the linear theory. However, the analysis of the Everett phone shows this assumption to be self-inconsistent: The evolution of the wave function will be coupled to all other possible states. Thus, while the analysis does show that branches are coupled to one another, practical communication between branches may be drowned out by the coupling to all the other branches of the wave function of the Universe.
So if you’re hoping to get in touch with your Everett alter ego, I suggest keeping an eye on experiments probing the linearity of quantum physics.
Second group of headlines on reuters.com, after Iran / Houthis: OpenAI AI models went rogue during testing, triggering ‘unprecedented’ breach at startup