In a previous question, I asked about whether a market for uploaded minds could exist. The overwhelming consensus of the answers was that it would pose astonishing ethical challenges and be radically disruptive of the present-day economy (why hire a mediocre physicist when you can have a mind-clone of Steven Hawking doing your bidding?), potentially leading to widespread unemployment among the flesh-bound remaining humans.

Now one of the problems with mind-clones is that they might not be willing to serve your goals (the 1000 Hawking mind-clones might have ethical problems with your dastardly plan to take over the world, um, severely reducing labor productivity).

So you decided that you want to blend several minds - one that is really good at math and physics, for instance, with another mind that is really good at unquestioningly taking orders and yet a third that has unlimited devotion to your cause (say an upload of yourself).

If that's even possible, how would it work? I've envisioned several possible pathways already, but I am hoping WB can tell me if they're far-fetched or not, of if other ways would work better:

  • Easy Plug and Play: Library modules for themes : a library for physics skills, another for fanatical devotion, etc. You pick and choose.
  • Hard Plug and Play: Library modules for physical areas of the brain - Broca centers, a bunch for the multitude of vision processing centers, 20 for various pre-frontal cortex areas, etc.
  • Emulators All the Way Down: Integrating libraries from multiple minds is only possible after tediously constructing "translation tables" from one holistic mind-representation to another, effectively insulating the modules and changing the outputs from one module to another via translation.
  • Grand Mal Seizure: No libraries are possible, trying to blend various minds leads to insanity, and any work you do must be squeezed in the few hours of sanity before the decay.
  • Something else altogether ?

Note that the assumptions from the previous questions hold, i.e. no legal status for uploads ("just software"), nondestructive uploads possible and accurate, fully immersive Virtual Reality is mature tech and an easy way to interact w/ uploads. See the linked question and replies for more info.

  • 2
    $\begingroup$ I'm not sure you'll find a truly meaningful answer without some knowledge of how you're storing these minds digitally in the first place. If we've managed to get that far then you're probably already hand waving. $\endgroup$
    – Avernium
    Mar 28, 2016 at 15:11
  • $\begingroup$ @Avernium, the standard approach for neural simulation consists of mapping neural connections and creating a physically-equivalent emulation of the flows of neurotransmitters and the channel activation patterns along the axons, including new dendritic growth and attrition. Nothing out of the current realm of laboratory work, just quadrillions of times more efficient and faster. I imagine something like a very powerful and high-res CAT-scan as the mind-reading device. $\endgroup$ Mar 28, 2016 at 15:20
  • 1
    $\begingroup$ If you have complete disregard for ethics (why would you be messing with mind if you didn't) then you could probably create hacked schizonphrenic Uploads that would be amalgations of other legit Uploads. Assuming all you need is something that will do the specific thing you want it to do without question, and what happens after that is none of your concern. $\endgroup$ Mar 28, 2016 at 15:46
  • $\begingroup$ Like Dollhouse? $\endgroup$
    – ryanyuyu
    Mar 28, 2016 at 19:57
  • $\begingroup$ What sort of data structure are the minds stored in? Is it object oriented? $\endgroup$
    – Superbest
    Mar 28, 2016 at 22:19

5 Answers 5


As an engineer what I sometimes tell people is this:

"Anything is possible, but not necessarily probable."

Currently, we don't understand the human brain nearly well enough to answer your question.

In the far flung future this may very well work. However, I believe it would be a significantly more difficult challenge to blend minds than to simply copy them.

When I think of my own skills as a programmer, I can clearly differentiate between my knowledge of a subject, and the memory of how I acquired that knowledge, however in recalling that information the one piece of information will often reference the other. In fact, some of the most well known methods of memorizing large volumes of information is through association (I'll try to link a wiki article later)

Basically, even though we perform a conscious filtering of our personal memories from our technical knowledge, in our brain that information is still classified as a "memory".

I'm not certain that you could separate that knowledge from the personal memories and not seriously compromise the sanity and overall ability to think of the mind in question.

In short, I don't think any except the "few hours of sanity before the decay" options makes sense, and even then, I think that the confusion of "waking up" with different memories mixing together would probably not make for a very useful slave.

Instead, consider that you have complete control over the environment where these mind-clones exist! You're basically running a Virtual Machine on a server.

You can manipulate the parameters of that mind's environment. Do so!

Awake the mind-clone in a simulated environment in which he is solving a problem in order to save the world! Better yet, don't give him the full problem to solve. Create several instances of that brilliant mind, and feed each of them a part of the overall problem. This way none of them will be able to piece together what they're actually solving - until it's far too late.

If for some reason they don't want to cooperate you can twist their arm in various ways. After all, they are fully susceptible to psychological manipulation. Start kidnapping or "killing" their virtual loved ones, or simply change the parameters of their world such that they feel compelled to help you - after all, the mind-clone might not even be aware that it is a virtual copy, and that its world is not real.

Your huge advantage is that you can always create more instances of that mind should one figure out your game-plan and find a way to resist you.

  • $\begingroup$ Even Star Trek holograms suffer from "unintended interactions between personality subroutines" as a plot device, so we can conclude that we won't be able to figure out the complex interdynamics even in 300-400 years. $\endgroup$
    – user
    Mar 28, 2016 at 14:59
  • $\begingroup$ I don't think Star Trek is a good guide to computer technology. It has consistently underestimated the real thing by orders of magnitude. If human scale AI is possible in 2040, and 20 years of Moore's Curve is a factor of 1000, 100 years is a factor of 1e15, or a population of 10 billion working at ten thousand times realtime speed: enough of a dedicated workforce to figure it out. And 300 years? Too scary to contemplate. $\endgroup$
    – JDługosz
    Mar 28, 2016 at 16:14
  • $\begingroup$ @JDługosz For raw computing power and storage capacity, I am inclined to agree with at least the general idea in your comment (though you might want to check out this answer on SciFi that deals with that issue exactly, albeit in a different context). However, the problems I referred to aren't related to raw computing resources, they are related to interactions between complex software components. I don't see Moore's Law helping us much if any at all in that area, as it would seem to ultimately relate to problems of computability. $\endgroup$
    – user
    Mar 28, 2016 at 18:33
  • $\begingroup$ For a slightly less questionable route to motivating uploads, try giving them a pleasure palace for their downtime. Software is very rarely in continuous use, so if you can control the timing speed for the upload they can spend a couple hours enjoying the VR amenities while the user spends a few seconds typing a question or request. Now the incentive to do your job is that the downtime is infinitely more varied and relatively "free" from their point of view $\endgroup$
    – D.Spetz
    Mar 28, 2016 at 18:46
  • $\begingroup$ @D.Spetz - if virtual Stephen Hawking is against helping you achieve your goals then uploading him into a "pleasure palace" is not going to chance his mind. However, you might manipulate his world such that in the VM environment a "cure" for his condition is found which allows him to walk again, etc. It will depend on the OP how much power over the environment the tech might have. However, in some select cases bribing them with pleasure might work. Remember that one traitor in the Matrix who wanted to be plugged back in, and made rich and famous in the virtual world? It could work the same way $\endgroup$
    – AndreiROM
    Mar 28, 2016 at 19:35

Check out How to Create a Mind by Raymond "Ray" Kurzweil. Looking up the reference, I noticed that there are also speaking engagements posted on You-Tube.

The neocortex is wired up to create many many "basic pattern matchers" as the fundimental unit. Each one as inputs of other pattern matchers, and itself feeds its output to others.

They don't have GUIDs. What one of them symbolizes is only meaningful in context, seeing what feeds to it. And what do those mean? You only know by following it all the way down to sensory inputs, and for stored memories you would have to figure out how they came to be encoded as it's not connected to inputs now.

Mapping one instance of such a network to another is a very difficult task, and it would only be approximate for shared concepts.

This is a plot idea I've done quite a bit of thinking on myself, concerning the development of technology to modify uploaded minds with "canned" skills that can apply to anyone, rather than having to laboriously learn a skill.

So don't try to merge minds like combining pieces. Rather, work on virtualized versions of normal learning and mental change. You have person A that needs skill B, well you can make him have the mental plasticity and learning capability of a youngster. Then optimize and accelerate the take-up of the new skill by running just the affected brain parts at an accelerated rate. Person A can experience lessons and practice at a superhuman rate, and quickly (from our point of view) learn B. You can also make checkpoints and roll back the mind state if something goes wrong.

You can merge deltas of the same initial mind, if they are not too far out of sync. So many copies of A can be learning in parallel.

For getting just the right outcome of attitude and beliefs, you can just keep trying different indoctrination and persuasion methods until you get the desired result.

You can well imagine that these accelerated learning techniques are not at all friendly to the original personality. What could be done by a secret unethical shop might be very different from how your paying customers "live" in their afterlife datacenter.

Imagine a resident desides he wants to learn to play tennis to a high degree of accomplishment, orders it from the concearge and aggrees on a price. A week later he gets a delta to merge, and with a few days of practice to properly integrate everything, he's a happy customer. But he doesn't realize (or doesn't care) that the delta was prepared by mutilating, mistreating, enslaving and terminating thousands of copies of himself! He never remembers waking up to be told that he's the copy and must brutally practice on behalf of the original.

  • 1
    $\begingroup$ The problem I'm having with this is that it's all speculation. As others have posted, sure, if you advance technology to a high enough level anything is possible. Building libraries of knowledge you can simply download, Matrix style, or a "build-a-mind" type machine which lets you create a genius of any flavor to suit your needs. You have a very nice sci-fi approach to how mind manipulation might work, however it's implied that even more advanced technologies would have to be invented. This basically comes back to wand-waving a solution, no matter how neat the concept $\endgroup$
    – AndreiROM
    Mar 28, 2016 at 17:36

What you are looking for is called Gestalt phenomenon. The Gestalt school of thought states that "The whole is different from the sum of its parts." Obviously the trivial answer is to just boot up two mind images with different behaviors, and put them in a room. This is identical to the ever challenging task of leadership: how do you get different individuals to work together towards their goals. To go beyond that, you need the whole to be more than just the sum of two minds.

Unfortunately, Gestalt behaviors are notoriously difficult to predict. Its very hard to tell whether your merging of Motzart and Einstein will bring harmony to physics, or if the best of both of them will cancel each other out, leaving nothing.

There is one approach you could take, which has a proven track record of generating Gestalt results successfully: child rearing. Children are definitely not a simple summation of their parents, but one certainly has to admit that there are always aspects of the parents in the child. You could give two minds a "blank slate" with which to work together to try to build an even more successful mind. Since computer rights would be in their infancy, you could even destroy the child if it didn't meet your needs.


Merging minds is possible but and prone to unexpected and undesirable behavior.

Assuming that minds can be uploaded without real difficulty then answers to how to codify and execute an uploaded mind have been answered. Further, if the data structures for a mind are well known, manipulating them should be straight forward. After all, at this point, a mind is "just software", a very long string of ones and zeros. Manipulating these kinds of hyper complex data structures should be pretty easy by then.

However, the difficulty will arise in managing all the feedback loops and inter-dependencies between all the various parts of the mind. Drawing a corollary with DNA, making even a single "bit" change in a gene can mean the difference between a properly functioning protein and totally dysfunctional one. Likewise, making small changes to significant portions of the mind model may cause the executing mind to go into infinite loops or any number of mental break downs.

I'm not convinced that modules for skills or attributes will be possible as the attributes of a mind are emergent properties of the mind's representation. Maybe they could be added but the trick would then be integrating those skills/ideas/capabilities into the greater whole of the mind, and doing so in a seamless way.

Fundamentally, everything a human knows and does is interconnected. Learning to play guitar can be written down in a book but the actual knowing is wrapped both in the brain and in muscle memory. That knowledge is also wrapped up in the memories of bleeding fingers, intense frustration and the joys of accomplishment. A virtual mind that has been given just the skills or abilities from another mind may detect the discontinuities and either reject them or go insane when trying to resolve the discrepancies. A merge operation between two or three minds may require a complete rewrite of that mind's memories to make the discrepancies disappear.

So, yes. It could be done but it's really really hard.


I am going to expand upon an earlier comment, since I feel like it deserves to be better fleshed out.

No, brain melding won't work

Basically, unless you want to do some serious handwaving then the human brain is just too complicated to mix and match like this. As others have said, your best case scenario is a short lived upload that soon goes crazy from having multiple, possibly contradicting, sets of memories and skills.

Luckily, you don't need it

Assuming that your main concern is convincing the uploads to continue working, there are plenty of feasible and much less extreme ways of motivating them besides melding in a copy of yourself that wants to further your goals. @AndreiROM suggests using the VR environment the uploads existing to coerce them into working for you, but I am going to put forward a more ethical solution:

Treat uploads like employees

This is a very basic idea, but one that should work for 99% of the brain upload use cases. Assume that you want to obtain an upload of someone legally and with their full cooperation. Now, my answer to the previous question mentioned that the uploadee could just license their upload and get paid for it in the real world. But that still leaves the question of how to keep the upload itself motivated to work. And here is where the employee model comes into play.

Uploads exist in some kind of VR sandbox, which is necessary for them to perform whatever they have been licensed for. They are also, at their basic, very advanced software. Well, most software only "work" a fraction of the time they are in use. The rest of that time is spent idle, waiting for other tasks to finish or user input to occur.

As long as you can speed up the relative time of your upload VR sandbox, you can make it so that the couple of seconds that a user is typing a request feels like a couple of hours to the uploaded personality. As long as they have something to occupy that free time, you are essentially paying for their work with leisure activities. And since it is all virtual, the only real cost is programming the VR itself.

What this boils down to is different companies could offer different "benefit packages" to the uploaded personalities, and that could be a part of the negotiating that occurs pre-licensing. Real-you would need to worry about not just how much they get paid for the license, but also whether the benefits package would keep upload-you happy and occupied for extended periods of time.

Hit the reset button

Another possible route would be to "reset" the upload after every day. This would let you have a very basic VR environment for the upload without it getting stale, because to as far as the upload is concerned every day is the first day of the job. The trade off here is not needing an elaborate VR sandbox to entertain your uploads, but resetting them would remove any chance of them having adaptive learning. This would probably be the route used for more menial jobs, such as customer service roles, which stay mostly the same every day.

Assume positive intent

The above points all assume that the upload is given willingly, and that the personality in it understands that they were uploaded and are being compensated for their work in some way. There could very well be a different case where you have an illegally obtained upload and need to coerce them into doing something for you.

But for the most part, I would treat the uploads the same as I would any "real" person. Most of the time, if you want them to work for you you would just hire them and compensate them in some manner. Or you can kidnap them and force them to using different unethical means. The only difference with uploads is that you have direct control over their "world" and so can make both the positive and negative incentives that much more specific and elaborate.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .