Welcome! These forums will be deactivated by the end of this year. The conversation continues in a new morph over on Discord! Please join us there for a more active conversation and the occasional opportunity to ask developers questions directly! Go to the PS+ Discord Server.

Questions regarding fork merging

7 posts / 0 new
Last post
DivineWrath DivineWrath's picture
Questions regarding fork merging
Hi. I have a few questions regarding fork merging. I'm hoping to hear what your thoughts are. I understand the rules of the game reasonably well, but I'm having troubles with how the science and technology would work. When merging 2 forks, what exactly does one do with them? At first I thought that they took the 2 egos and try to merge both to make a 3rd ego. Kinda like the average of the 2 egos, but with pains taken to make the new ego work correctly and keep the memories intact (a simple average might break stuff). Basically, the egos would have equal weight in determining what the new merged ego is like (give or take based upon the extra time that one fork might have been running that the other hasn't). However, I wasn't sure if that was the way forks were merged in the setting. I've been considering alternatives. In the new Transhuman rules, they seem to strip down one fork to get the memories, and then add those memories to another fork. All this talk about prioritizing memory retention seems to support that method. A blending merging shouldn't have those problems. Things like memories, skills, and everything else aught to be preserved in a blending merge, where as a grafting merge might only retain select components. Another question is what happens to stress when you fork merge? Do you retain the sum total stress of both forks, or do you keep the worst of the 2 forks? Would there be a difference if the stress came from 2 different sources, or from the same source? By the same source, I mean what if you forked and sleeved into 2 morphs to go kill the same TITAN monster? What if you ran a simulspace simulation at the same time, in the same world, where both forks did the exact same thing? ---- I'm not sure if I had more questions. I've been putting off asking these questions for a while now so I might need a bit longer to think things over.
nerdnumber1 nerdnumber1's picture
I don't think the exact
I don't think the exact methodology of merging has ever been described in much detail. It is likely that there are multiple tricks and methods that are used in concert based on the specific case and individual style, with the psychosurgery check signifying how well these methods are chosen and executed. I would say that the high degree of similarity in short-term fork merging makes the former "averaging" method is more useful and easier, while more divergent ego merges (such as separate-ego merges) necessitate grafting sections of one ego onto a somewhat pruned primary ego. Still, this is sci-fi brain-surgery, so I think there is more finesse and nuance involved in merging software. As for stress, I think this question came up before. I don't remember how it was resolved, but I doubt blending with an insane version of yourself is going to be good for your sanity. I think it might be somewhere between worse of the two stress and sum of both stress, based on how different the two stimulii that caused the stress were, GM discretion (if you combine two minds with different mind-rending experiences then you are worse off than if they experienced the same experience).
Arenamontanus Arenamontanus's picture
In my essay on merging here
In my essay on merging here is my take on how it is done:
Quote:
The merge process has to figure out how to modify the low-level neural pattern so that the high-level pattern of memories makes sense. This is done by a process of averaging, pattern-matching and reconciling that can become arbitrarily complex: the full merging problem is equivalent to understanding both egos completely and composing a new one sharing all relevant properties. This is of course far beyond any transhuman cognotechnology, but crude approximations are surprisingly often effective. Many merge systems make use of a pre-fork ego recording to detect differences that have appeared in the egos being merged: this is why most everyday merges can be fast and reliable. Various heuristics have been developed for memory precedence, coincidence detection and even ways of remapping conflicting memories onto “free” neural networks.
Note the pre-fork trick. You calculate the differences between the original and version 1 and version 2, and then combine the differences rather than the entire states. This is equvalent to a three-way merge in current revision control software. The other methods mentioned in Wikipedia no doubt have their counterparts in EP cognoscience. I would use the worst stress of the two egos, or even add it if the sources were very different.
Extropian
nerdnumber1 nerdnumber1's picture
Arenamontanus wrote:In my
Arenamontanus wrote:
In my essay on merging here is my take on how it is done:
Quote:
The merge process has to figure out how to modify the low-level neural pattern so that the high-level pattern of memories makes sense. This is done by a process of averaging, pattern-matching and reconciling that can become arbitrarily complex: the full merging problem is equivalent to understanding both egos completely and composing a new one sharing all relevant properties. This is of course far beyond any transhuman cognotechnology, but crude approximations are surprisingly often effective. Many merge systems make use of a pre-fork ego recording to detect differences that have appeared in the egos being merged: this is why most everyday merges can be fast and reliable. Various heuristics have been developed for memory precedence, coincidence detection and even ways of remapping conflicting memories onto “free” neural networks.
Note the pre-fork trick. You calculate the differences between the original and version 1 and version 2, and then combine the differences rather than the entire states. This is equvalent to a three-way merge in current revision control software. The other methods mentioned in Wikipedia no doubt have their counterparts in EP cognoscience. I would use the worst stress of the two egos, or even add it if the sources were very different.
This is exactly the sort of thing I was thinking but was unable to so eloquently articulate. Excellent work. I wonder what happens if you change your heuristics to prioritize the personality and memories of one of the ego's over the other, essentially attempting to turn the merge into a more permanent/stronger multi-skillsoft of the other party? I could see some finding value in acquiring the skills, knoweledge and memories of other egos (either with or without their consent/knowledge) without completely losing their own individuality. You would probably even be able to soften the blow of an alien merge since the program is prioritizing keeping your ego whole over conserving the other. This is a good explanation for why Transhuman's "merging different egos" rules has one primary ego, since the partial prioritization of one ego gives a significantly increased chance that the result will be viable.
Arenamontanus Arenamontanus's picture
Thanks!
Thanks!
nerdnumber1 wrote:
I wonder what happens if you change your heuristics to prioritize the personality and memories of one of the ego's over the other, essentially attempting to turn the merge into a more permanent/stronger multi-skillsoft of the other party? I could see some finding value in acquiring the skills, knoweledge and memories of other egos (either with or without their consent/knowledge) without completely losing their own individuality.
You could likely do this. I doubt it is as easy as turning a fork into a skill- or knowledge-soft, but you might not want to merge in the emotional associations of a heavily traumatized fork, and selectively weakening the episodic memories of a fork in a merge while keeping some of the motor-skills is likely entirely doable. Character idea: he runs many forks, trying out many approaches to projects and learning from them, then quickly merge them back in order to boost his skills. Essentially he is training himself in parallel. The downside is of course that he is constantly doing rather iffy neuroscience to his ego, over and over again. Slowly he is going mad, but he is going to be *really* good at his core skill...
Extropian
CodeBreaker CodeBreaker's picture
Arenamontanus wrote:
Arenamontanus wrote:
You could likely do this. I doubt it is as easy as turning a fork into a skill- or knowledge-soft, but you might not want to merge in the emotional associations of a heavily traumatized fork, and selectively weakening the episodic memories of a fork in a merge while keeping some of the motor-skills is likely entirely doable. Character idea: he runs many forks, trying out many approaches to projects and learning from them, then quickly merge them back in order to boost his skills. Essentially he is training himself in parallel. The downside is of course that he is constantly doing rather iffy neuroscience to his ego, over and over again. Slowly he is going mad, but he is going to be *really* good at his core skill...
"Sounds a bit like how the TITANs got started. We watching this?" "Yeah, yeah. Don't worry [Proxy RetCon], we have a team on standby. If we see signs of an uncontrollable lift-off we can nuke the site from orbit." "It's the only way to be sure!"
-
nerdnumber1 nerdnumber1's picture
CodeBreaker wrote:
CodeBreaker wrote:
"Sounds a bit like how the TITANs got started. We watching this?" "Yeah, yeah. Don't worry [Proxy RetCon], we have a team on standby. If we see signs of an uncontrollable lift-off we can nuke the site from orbit." "It's the only way to be sure!"
You think that someone with a dozen minds hasn't made redundant backups and alpha forks working on other habs? Nuking the site would merely be an inconvenient loss of progress and resources, throwing the project back while tipping your hand. __________________________ It is unfortunate (yet necessary) that the "rez point debt" system for gaining skills from forks in Transhuman makes this parallel training strategy ineffective for PCs (at most allowing for quicker training or going into heavy rez debt). The Exhuman faction in Transhuman has made me interested in making a (moderate) exhuman character. Basically, he (it?) will use psychosurgery on his own forks (and any other consenting party member) whenever he deems it necessary. The character wants to transend humanity, but on HIS own terms, so no jumping into any Titan or alien tech without serious analysis and experimentation (likely involving his own forks). Might ask to merge one of his forks with a respected party-member's fork who has skills that he lacks to get an improved version of himself. I was thinking he might have some disturbing quirks, such as a habit of keeping a back-up of anyone he operates on, without necessarily informing them of this, not out of malice, just in case there is an accident during a mission, or they have need of a quick fork while the individual is sleeved in a biomorph. Other Sentinel: "I hate _______." or "I'm terrible at ________." Exhuman Sentinel: "I could fix that for you. With time-acceleration, the process should take under 4 hours with risks within acceptable margins." Other Sentinel: "No way I want you messing with my head!" Exhuman Sentinel: "... I could fix that for you." Other Sentinel: "NO!"