Time Travel, AI and Transparent Newcomb

Epistemic sta­tus: has “time travel” in the ti­tle.

Let’s sup­pose, for the du­ra­tion of this post, that the lo­cal physics of our uni­verse al­lows for time travel. The ob­vi­ous ques­tion is: how are para­doxes pre­vented?

We may not have any idea how para­doxes are pre­vented, but pre­sum­ably there must be some pre­ven­tion mechanism. So, in a purely Bayesian sense, we can con­di­tion on para­doxes some­how not hap­pen­ing, and then ask what be­comes more or less likely. In gen­eral, any­thing which would make a time ma­chine more likely to be built should be­come less likely, and any­thing which would pre­vent a time ma­chine be­ing built should be­come more likely.

In other words: if we’re try­ing to do some­thing which would make time ma­chines more likely to be built, this ar­gu­ment says that we should ex­pect things to mys­te­ri­ously go wrong.

For in­stance, let’s say we’re try­ing to build some kind of pow­er­ful op­ti­miza­tion pro­cess which might find time ma­chines in­stru­men­tally use­ful for some rea­son. To the ex­tent that such a pro­cess is likely to build time ma­chines and in­duce para­doxes, we would ex­pect things to mys­te­ri­ously go wrong when try­ing to build the op­ti­mizer in the first place.

On the flip side: we could com­mit to de­sign­ing our pow­er­ful op­ti­miza­tion pro­cess so that it not only avoids build­ing time ma­chines, but also ac­tively pre­vents time ma­chines from be­ing built. Then the mys­te­ri­ous force should work in our fa­vor: we would ex­pect things to mys­te­ri­ously go well. We don’t need time-travel-pre­ven­tion to be the op­ti­miza­tion pro­cess’ sole ob­jec­tive here, it just needs to make time ma­chines suffi­ciently less likely to get an over­all drop in the prob­a­bil­ity of para­dox.

• We would also ex­pect de­stroy­ing time ma­chines to be a con­ver­gent in­stru­men­tal goal in this uni­verse, since any agent that does this would be more likely to have been cre­ated. So by de­fault pow­er­ful enough op­ti­miza­tion pro­cesses would try to pre­vent time travel.

• Let us sup­pose <im­pos­si­ble thing>. Now <im­pos­si­ble re­sult> re­mains im­pos­si­ble, how? Maybe the uni­verse has a mys­te­ri­ous agency we can trick or bar­gain with!

I think you’ll need to back up a bit fur­ther if you want to ex­plore this. “time travel is pos­si­ble” isn’t well enough defined to be able to rea­son about, ex­cept in the hu­man con­cep­tual space with no physics at­tached. And if you’re as­sum­ing away physics, you don’t need to ex­plain any­thing, just let the para­doxes hap­pen.

• Are you fa­mil­iar with the Gödel met­ric? Time travel may well be im­pos­si­ble, but at least within the con­text of gen­eral rel­a­tivity it is plenty well-defined to rea­son about.

• Does Gödel met­ric say any­thing about pro­hi­bi­tion of para­doxes?

I have found that as­sig­ment of 0 to a para­dox­i­cal con­figu­ra­tion is an opinion or as­sump­tion and not a re­sult or a the­o­rem.

For ex­am­ple the elec­tron wave func­tion nega­tive solu­tions were dis­missed as un­re­al­is­tic math arte­facts at one point and later adopted as a valid way to rea­son about positrons. Would it have been cor­rect to say that “ob­vi­ously elec­trons have a nega­tive charge?. In mod­ern terms you can dis­t­in­guish be­tween claims of “lep­tons have nega­tive charge” and “elec­trons have nega­tive charge”. But if the only lep­ton you know is an elec­tron is “pos­i­tive elec­tron” a valid thing?

While things add up to nor­mal­ity un­usual cir­cusm­stances can ex­hibit un­sual phe­nom­ena.

• Does Gödel met­ric say any­thing about pro­hi­bi­tion of para­doxes?

The real ques­tion here is what me­chan­ics + GR “says” about para­doxes; there’s noth­ing spe­cial about the Gödel met­ric other than that it’s a spe­cific ex­am­ple of a sys­tem con­tain­ing closed time-like loops.

The an­swer is that me­chan­ics + GR can­not rep­re­sent a sys­tem con­tain­ing a para­dox, at all. We just have a bunch of par­ti­cles and/​or fields mov­ing around a space with a met­ric. The lo­cal laws of me­chan­ics + GR con­strain their be­hav­ior. A “para­dox” would, for in­stance, as­sert that there is a par­ti­cle at (x, t) with ve­loc­ity v, but also not a par­ti­cle at (x, t) with ve­loc­ity v—the un­der­ly­ing the­ory can’t even rep­re­sent that.

We don’t know how to in­te­grate QFT with GR, but con­cep­tu­ally a similar prob­lem should arise: we just have some quan­tum fields with com­plex am­pli­tudes at each point in space­time. A para­dox would as­sign two differ­ent am­pli­tudes to the field at the same point. Again, our phys­i­cal mod­els can’t even rep­re­sent that: the whole point of a field is that it as­signs an am­pli­tude at each point in space­time.

We could maybe imag­ine some sort of mul­ti­val­ued state of the uni­verse, but at that point our “time ma­chine” isn’t ac­tu­ally do­ing time travel at all—it’s just mov­ing around in a some­what-larger mul­ti­verse.

• As we lack the means to rep­re­sent the differ­ent op­tions we prob­a­bly do not have a law that para­doxes will be avoided (partly be­cause we do not have a tech­ni­cal analo­goue for “para­dox”)

In the ex­tended on­tol­ogy what cor­re­sponds to old time would be an open ques­tion. That is if you have a mul­ti­val­ued state in the past and some of the val­ues of that are effects of (par­tial) val­ues in the fu­ture it’s still pretty much “time travel”.

I also thought that quna­tum me­chan­ics is pretty chill with su­per­po­si­tion. Could not one ex­tend the model by hav­ing a differ­ent imag­i­nary unit and then have a su­per­po­si­tion of am­pli­tudes? And I thought get­ting a sure eigen­value is a spe­cial case. Isn’t the non-eigen­value case already cov­er­ing a si­mul­tanoues at­tri­bu­tion of mul­ti­ple real val­ues? I case there are two cases 1) we do not rep­re­sent that cur­rently in our mod­els or 2) Our rep­re­sen­ta­tions used in our mod­els can not rep­re­sent that.

• Hav­ing played Achron I have vis­ited the thought space a lot be­fore.

The thing that al­lows for time travel might break a lot of your other pre­sump­tions. If the moon was made of cheese were would all the milk would have come from?

I would like to point out that the as­sum­tion that timelines are sta­ble is a sep­a­rate as­sump­tion from time travel. I think it needs ar­gu­ment and ar­gu­ment from lack of imag­i­na­tion is not a very con­vinc­ing one.

How­ever even if you go out­side of this as­sum­tion were similar things still hap­pen. In an­other at­tempt one could ar­gue that if a grandafther para­dox has 2 vi­able states then both of those states should sum in prob­a­blity to a state that is “sta­ble”. So you as­sign less prob­a­bil­ity to things that can grand­father para­dox. If a thing can at­tempt to para­dox in mul­ti­ple ways you have keep split­ting the “mea­sure”. The end re­sult is that a thing that can con­stantly para­dox will be van­ish­ingly un­likely to hang around. This is differ­ent than “mys­te­ri­ously pre­vented” but for these pur­poses serves a similar func­tion.

After all if quan­tum su­per­po­si­tions are not pre­vented why would “timeline su­per­po­si­tions” be in­com­pat­i­ble?

It is note­wor­thy that in­ven­tion of time­travel tech­nol­ogy is an event that prob­a­bly has causes. Any­thing that util­ises such tech­nol­ogy would be causally de­pen­dent on that event. But that even is causally de­pen­dent on other events. Thus it might be coun­ter­pro­duc­tive to have any time­travel tech­nol­ogy effect any cause of time travel tech­nol­ogy. This would effec­tively mean that all of pre-time­travel his­tory would be “nat­u­ral re­serve” on the pain of death of all time­travel his­tory.

The is a re­verse effect where is a time ma­chine can help with its con­struc­tion it might be tempted to do so pro­vided it doesn’t undo it more than do it. Even the slight­est chance of a work­ing time travel method would be blown to op­ti­mally early time travel (within the con­straints).

That is fic­tion has plenty of time­travel rea­son­ing to throw around. Spec­i­fy­ing your poi­son would go for a long way. For ex­am­ple rpg game con­tinuum has slip­shank­ing. Hav­ing a sud­den fight? Grab a pis­tol from near­est con­tainer. Then af­ter fight go to a shop buy gun, go to past and put gun into said con­tainer. Timeline is sta­ble, but it feels weird taht re­al­is­ing that you could try do some­thign makes it pos­si­bel for you to do so (this kind of rea­son­ing has limi­ta­tions, if you have ope­nend the box and seen it empty you ob­viosly can’t slip­shank a pis­tol out of it. But be­ing strate­gi­cally ig­no­rant in or­der to max­imise slip­shank­ing pos­si­blities is in­ter­est­ing). The ul­ti­mate slip­shank would be to do it be­fore time travel is in­vented with the in­ten­tion to in­vent time travel to pull it off