Is it realistic to think that, given the inefficiency of conveying information to another human through verbal or written language, it may one day be possible to "download" knowledge directly into a human brain?

Basically, human A wants human B to know something, some complicated state of affairs that may be misunderstood if communicated through the ambiguity of language, or maybe it's a simple packet of information, but human A wants human B to absorb it as quickly as biologically possible. It would also have the benefit of being completely unambiguous. The new knowledge human B has is identical to that of human A.

Note that I don't necessarily mean Matrix style "I know Kung-Fu" learning. Nothing that requires muscle memory, just the transmission of semantic knowledge, or facts about the world, that one human wants to transfer to another.

The information also doesn't need to be "true," it's up to the receiving human to verify it, but the receiver will be under no uncertainty as to what the sender means.

Edit: I really don't know how much more specific I can make this question.

  • 1
    $\begingroup$ what is the technological level of your society? $\endgroup$ – Olga Dec 2 '17 at 16:11
  • $\begingroup$ @Olga Late 21st Century with significant advances in high resolution brain imaging and brain-computer interfaces. $\endgroup$ – OptimalSolver Dec 2 '17 at 16:17
  • $\begingroup$ Based on what I learned from the original Star Trek's episode "Spock's Brain," it's clear that "a child could do it." $\endgroup$ – Don Branson Dec 2 '17 at 22:19
  • $\begingroup$ Do you want real implanting of knowledge, which has to be done in whatever encoding that brain uses, or do you want "fake" implanting, where we figure out a way to encode the information and force the brain to live with whatever interface we provide? $\endgroup$ – Cort Ammon Dec 2 '17 at 22:52
  • 2
    $\begingroup$ 1. You should edit the 21st century comment into the question. 2. You should read the JDlugosz link and edit into the question how it does or does not relate to what you want to know. 3. You should read the Cort Ammon and Justin Thyme comments and in the language they use, explain how it fits your question. (It seems obvious to me that you want real implanting aka thought transference, but apparently it was not obvious to them. Please make it obvious.) Those are the three things that people have asked you to do to clarify the question (not necessarily make it more specific). $\endgroup$ – Brythan Dec 3 '17 at 18:23

There are really two questions in one:

  1. Is is plausible to have a high-speed data interface to the brain, and
  2. Is there a method to convey information unambiguously.

A high-speed data interface to the brain...

... is probably not possible. Or is probably possible. Take your pick. I believe it is not possible, because the brain does work that way. The brain is a very slow, very asynchronous, very highly distributed data processing machine. There is simply no one data connection in the brain through which all memories pass; in computer terms, there is no system bus.

But. Maybe we really cannot make a high-speed data interface to the brain, but most likely we can make a slow speed data interface. That is, although we most likely cannot pour knowledge into the brain, maybe its not necessary. Maybe we can make a suitable digital device with a suitable biological interface which works as if it were part of the brain, retrieving information at the low speed at which the brain can process it.

At our current technological level we already have something very very close to this, in the form of keyboards and the world-wide web, and Google, and Wikipedia. Anybody can find out just about any little fact in mathematics, physics, chemistry, or history quickly, by using the brain to formulate a query, typing that query on a keyboard, using Google's vast computing power to search Wikipedia's vast data repository. It is perfectly possible to bypass the keyboard by capturing motor neural impulses, and it is perfectly possible to dispense with the screen by feeding neural impulses directly into the optical nerve.

Conveying information unambiguously...

... is perfectly possible. We do it routinely in mathematics, in physics, in chemistry, in engineering, and in information technology. For this purpose we use special-purpose languages (such as mathematical notation, chemical notation, technical drawings, and programming languages) and well-specified databases. We can specify colors unambiguously using either well-defined color spaces, or the Pantone Matching System. We can specify shapes unambiguously using technical drawings or computerised 3D models represented in well-defined formats such as AMF.

For more complicated uses there are special-purpose subsets of natural languages using controlled vocabularies, such as Simplified Technical English.

On the other hand, ambiguity does not necessarily arise from the imperfection of language; quite often, ambiguity is inherent in the information. For example, Julius Caesar (the Julius Caesar) was born on 13 July 100 BCE, and that's a fact. It's also a very ambiguous fact; that nice specific date is misleading: for it was Julius Caesar himself who, in the fullness of time, would regularize the calendar so that all years had 365 or 366 days, and so that 1 January would invariably fall about 10 days after the hibernal solstice. All we know is that he was born on a day which in that specific year the Romans counted as 13th of Quintilis (as the month was called then, the name "July" being given to it much later and in his honor); we don't know to what day in our calendar corresponds that specific 13th of Quintilis 100 BCE; we can say that it was most likely during the summer, and that's about all.

  • $\begingroup$ Law also goes to great lengths to unambiguously convey information using natural, but not uncommonly specialized, language. The term "legalese" is often used in a less than flattering manner, but it's hard to deny that many of the hoops taken particularly in legal texts are there to ensure that the information can only be interpreted in one way. Of course, this still requires a context for such a "language" to exist within. My old question What conditions might cause a natural language to be completely literal? may be relevant. $\endgroup$ – user Dec 2 '17 at 18:02
  • $\begingroup$ I think you need to read up on the Roman, Julian, and Gregorian calendar and edit your answer. We do know the conversion between these date systems. In fact every programming language with date types has built in conversion functions. By my understanding, Julius recorded his birthday in his own system, so it's safe to assume he knew the conversion from the Roman system. $\endgroup$ – Stephan Dec 2 '17 at 18:04
  • $\begingroup$ @Stephan: The entire point is that the pre-Julian Roman system had years of unequal length; some had twelve lunar months, some had thirteen; and the length of each particular year was essentially arbitrary. It is clear that we have different knowledge of the pre-Julian Roman calendar: which of us needs to read more will remain an unsolved ambiguity. Moreover, I cannot see how C.J.C. could have "recorded his birthday in his own system" -- his birthday was a. d. III Eid. Quint. (pre-Julian) since he was a toddler. $\endgroup$ – AlexP Dec 2 '17 at 18:20
  • $\begingroup$ Not so. The Roman calendar was anchored on the summer and wi ter solstices just like almost every ancient calendar system. The "different lengths" part comes from the fact that the Roman calendar didn't really count winter as days of the year. A year was from summer solstice to winder solstice. $\endgroup$ – Stephan Dec 2 '17 at 18:26
  • 1
    $\begingroup$ The trouble with this plan is that conveying information unambiguously does not always result in the subject learning the topic perfectly. $\endgroup$ – Ross Presser Dec 3 '17 at 2:15

First: can we know what part of a brain is performing which task? Yes.

from https://www.mayfieldclinic.com/PE-fMRI_DTI.HTM functional mri image

This is now routinely done. MRI can tell which parts of the brain are more active when performing a given task. Interestingly (and something I just learned) the patients do not actually perform the tasks, because they move around too much when they do. They just think about performing the task. As the technology improves, fMRI has become able to detect and discriminate between different word recognition / generation activities, and different activities with a given body part.

2: Can one reverse this, and use magnetic stimulation to provoke activities within the brain? Yes.

Transcranial magnetic stimulation is as yet a pretty blunt instrument, but it is being used clinically to stimulate learning, ameliorate pain and other more global brain activities. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3532848/

Recent studies have suggested that exogenous manipulation of cortical activity with repetitive transcranial magnetic stimulation (rTMS) may improve naming in persons with chronic left hemisphere stroke and nonfluent aphasia.... The specific mechanisms by which rTMS administered to this region produces beneficial changes in language ability are debated.

Transcranial magnetic stimulation can produce the sensation of movement and block the ability to speak (but leave the ability to sing), among other things. This is not future speculation but established and in clinical use.

Could this sort of thing evolve to placing memories in a mind as the OP proposes? I see no reason why not. A 19th century surgeon would scoff at the thought that one could reconnect severed nerves or reroute a blood supply from one part of the body to another but these things are now routine. If the electromagnetic nature of memory as stored in the brain is better understood, one could replicate the generation of memory with extrinsic magnetic stimulations - directly implanting knowledge into the brain.


Technically yes, but today no.

We know barely how a humans brain work and therefore we know information is beeing stored using electric impulses. These electric impulses can be triggerd from outside as well, but the point is, that (today) we just don't know exactly enough, which points to trigger to get a specific information saved, and also we (today) can not exactly know, what information the brain is actually processing, when we see, where the information (voltage) flows.

We already know, which areas do which things, but that is definitifely not enough. Its like a computer. You know the brain is the RAM and you also know, that something between 0x0000001 and 0x0000007 to something between 0x00006F0 and0x000B71A is somehow connencted to language and speaking and you even know some other vague areas and a part of what they do.

BUT even if you capture "ahh this does language" you don't know "this is working on the word "pineapple" at the moment"

Saving information to a RAM/brain where you already know that much and nothing more ist just not possible. Also Reverse Engineering is somewhat difficult as we are talking about a living brain inside a human. You can not rewire this for try and error, because of moral aspects in as far as I know every country of the world. But even if someone says "hey, take my brain and do what you want even if i'm still living" this doesn't help you that much as you have to exactly know, which information in this RAM does what and how to access it.

Also every RAM can be accessed the same way and can be called the same way (mostly) as you have standards like DDR2 and the given size of the RAM (eg 2 GB) and input voltage and so on, but a brain doesn't follow some ISOsor even vague standards and therefore every brain is diffrent. Also there are tumors etc, so even if you could access a brain it, the way of how to do that would be different from human to human.

Also the brain is not entirely RAM and it is not entirely a processor. It works like a mixture of both of them. I'm not very into this, so ask a doctor for details, but a brain does (in my oppinion) not have a processing unit and a saving unit, but both things can be done somehow simultaneously from different regions in the brain. Given that you can not easily save information as you have to follow the brains "workflow" as the brain needs to get input to a specific time at a specific location based on what else is processed at this time and shot before as this most probably affects the way the brain stores this information correctly or not.

So putting together all this: Yes, one can possibly do this in the future, but nowadays this is somewhat impossible due to the reasons above.

Additionally: Also if you successfully manage to enter some information into the brain (let's say 0011101100001011 at some place) you can definitifely not know, how exaclty this affects the brain. Sure if you reversed the brain enough and you know that memory address of the word "banana" and you manage to change it to "apple" this does not mean the person is from now on saying apple to every banana it sees for as I said above the brain is also a processor and not just memory, so maybe this throws an exception (apoplexy in worst case), because the brain expected a different word or abstract datatype. -> good point. Abstract datatypes - those are things the brain uses to store information in and it does this, because it defined those datatypes based on its experience/knowledge influenced by the very personal perception, which means that every human saves his/her information in a slightly or heavily different way.

We know the brains error handling in some aspects. If we stay outside in the cold for a few days for example the brain may respond in shaking your body and felling pain or we know that getting a virus may lead to overheat the head, but we don't know yet, what the brain does (to us) if we manipulate thinking.

Back to the areas: Even if we know, which areas are responsible for doing things like speech, listening, seeing etc. we dont know, whether this covers the brains ability to save/store information or whether this is processing those information. So we know a brain acts as RAM and processor and we know some areas and their job, but we can't tell if the job is of the type RAM or procesor.

An additional thought on this: If humans think about bungling things in our brains - what do you expect from an AI? I think an AI would try to do similar things if it is bored. Before the AI does this you don't know if is good or not. Good means for us humans. The same applies to your question. We don't know what humans will do with the ability to change other humans (and their own) thoughts. It could be either good or bad. Using this feature beeing bad is probably more likely to happen as most humans are assholes and they would try to use this opportunity for whatever, but that is not part of the answer.

Coming back to your question: You wrote

it's up to the receiving human to verify it

I disagree. The verification of information stored in the brain does also have to be stored. Think about An apple can be eaten, which you KNOW is true and an apple is the same as a banana, which you KNOW is false and an information an apple is Gk(g5sP, wich you cant answer. Then there is an apple is red, wich you don't KNOW - you have to say "possible". But then there are information like "cousin Anna is 42 Years old", wich you probably just think might be correct as you didn't saw her for a few years. So if you just enter raandom information into a brain I would expect the brain to say something like "nope, that is false" or "don't know - might be fake news". Think about it: If you know, that in the past you never heared the word "Gwabbrügtz", but now you know it is a city with 81234 inmates and [...], but you cant combine this with experience, then you'll most likely say, that this information is not true as it might be a part of a dream last night you can't remember. So if you place information into the brain you have to ensure yourself the brain is 1. thinking the new information is true and 2. dealing with the new information in a healthy and good way. So no, it's not on the person to decide, whether the information in the brain is true - this is always determined previously by experience. You dont decide every time your asked, whether an appöle equals a banana -> you know this for sure to be false. So the knowledge whether this information is true is already in your brain. Therefore if someone manipulates the brain with information the validity of this information mus be transmitted as well as otherwise the brain would never think the new information is true.

Matrix style "I know Kung-Fu" learning

If it is already possible to make a brain believe things you entered into it, then there is absolutely no problem to apply this for muscle navigation as well. I wrote about those abstract datatypes and that every brain has to be accessed different, so the information, what to do with muscles is not one of the hard problems. It in fact should be really easy in caomparison to what it needs to successfully change a single bit on purpose (with knowing previously what this will change).

Also: Nowadays it is possible to make a brain do things with muscles. This has beend tested on animals (mice) and it works, but i dn'ß think it would be that easy with a human brain (without causing damage in any way / while this feature is under development)


I called the brains memory RAM as all unsaved changes will be lost if it is shut down. Nevertheless this can be changed. If one manages to change the brains memorys content the brains memory capacity can be extended as well so you can for example connect somehow the brain with some SSDs to make the brain store its information in there. Also this could result in "sharing memories" in a different manner we usually do. Instead of using speech the brains memories could be saved onto physical devices such as SSDs. To do this you have to implement sort of a driver to let the brain have read/write acces to this storage and then you can define how the information have to be saved on this external storage. This would make it easy to build a knowledge database for every (connected) human to make use of. Also you could stream information from this storage or from another brain directly into a "new sense" (a few of them we know as touch, smell, taste etc.). This would allow direct communication between two brains, so you could for example use the left arm of another person as you do it with your existing two ones. To enable this feature you have to write a driver for this into the target brain. This driver should be different from human to human, but the data inside the storage can have a standardisized format to reuse this fpr other brains.

To take those thoughts a bit further: If we think Matrix style leaarning a huge amount of things in a few seconds this could be possible - or not, because if you hear someone speaking you're fine, but if you hear his/her voice 20 times faster it gets confusing and you will most likely dont get all the information. I think it should be worth a thought if that could apply on the brain as well. As the brain is not just RAM, but also a processor it could be happening, that entering information triggers some kind of thinking (which is very likely) at least in the way of "what information I already have can I connect this with?" as we usually try to do this with normal learned things. So as a processor we have to ensure the brain does not overheat on the amount of things it is confronted with, so maybe it would be very slow to enter information and maybe not. I don't know and I don't think there is evidence for any answer to this as there is no such experience yet. It's just very likely, that in addition to the other circumstances the speeed will also be kind of bad for things humans would like to do. Also it would be difficult to get a save place to enter information as the brain can't be shut down without having exceptions thrown a lot after a reboot. Save place means, that the brain does not try to use the involved neurons at the same time as we modify them, because that would probably cause the brain to do some weired things. So you should make the area you want to edit not available for the brain for the time youre modifying information, which could be done by cuttin some synapses and "rewire" them after modification. To make the brain work properly in the meantime you could mak an exact copy of the area to modify, build a temporary storage, which the brain accesses as it was the space it would normally be and then you have to "rewire" the synapses. This is just an optional step, but for a good working solution this is a must-have.

So if you want to know, whether such a thing can be done SAFELY the answer will be no for a few hundred years. Nevertheless it is possible, but it will take very long to be able to avoid errors and it will be very expensive as well and at least the beta will still cause some mental deseases and deaths.


To sum it up:

Yes. To change a brains memory or adding new information should be technically be possible (magnetic stimulation or you're really using wires^^) but given the problems I mentioned above it is much too hard to do. Especially that what programmers understand as "abstract datatypes" may be a real hard problem for people trying to manipulate brains as their declarations/definitions will (most likely) vary from human to human significantly.

*I didn't read the other answers entirely before posting, so maybe there are some acpects already coverd by some other answers


Not a practical proposition
One of the big problems here is assumptions of prior knowledge. For example there would be no point in trying to transfer the knowledge of what diastereoisomers are without explaining what chirality is and no point in explaining chirality without explaining broader concepts in organic chemistry and three dimensional geometry and so on. The problem is that some of the required underlying knowledge is not at all obvious and it may not be obvious who has it or what is needed.

Another problem is the storage of knowledge. It might well be that the same knowledge is stored differently by different people depending on their lifetime experiences making the whole idea exceedingly challenging to say the least.

On top of all them there are the physical problems associated with reprogramming synapsis at a distance in a safe reliable and specific way. So whilst theoretically such a transfer might be imagined I suggest that in practice it would be very difficult if not impossible to achieve.


There is an decades-old Sci-Fi classics for this.


Also called hypnopedia. The basic (reasonably hard Sci-Fi) idea is that human brain registers some information even when asleep. Through hearing repeating the messages in their sleep an individuum may learn this information. It is less efficient than "normal" learning, but uses the time when people normally do nothing, so any additional amount of learned information is a plus.

You wanted uploading of a helicopter pilot skill as in Matrix? Sorry, that's much, much "softer" Sci-Fi.

  • $\begingroup$ Brave New World used this on children to teach propaganda $\endgroup$ – WorldCraftTrainee Dec 2 '17 at 21:35

It is virtually impossible to do without a complete map of every neuron (100 billion of them) and synapse connecting neurons (up to 1000 trillion of them).

See Here for a description of the biology of human memory.

Then you must have the ability to rewrite all of those, physically breaking and reforming synaptic connections in the brain.

The brain is not a computer with a blank slate of hardware that can contain data, knowledge, programming or instructions. What we are in personality or knowledge or skill is 100% encoded in the physical arrangement of cells, the precise way neurons fire and activate other neurons due to physical wiring from one to the next; those physical wires connect through synapses where one neuron 'touches' another (with a small gap for atoms of chemicals to flow between them).

To know something new requires a rewiring of this tangled mess, a literal growing process like growing vines. There is no PLACE in the brain to "download" anything, even when you hear a sound it is converted into electrical stimulation of cells, that fire if they recognize the pattern to stimulate other cells, and eventually stimulate cells to reconfigure their connections (synapses) so you know something new.

Worse than that, HOW you know something is unique to you and your experiences. What you think of when I say "car" is unique to you and your lifetime of experiences with cars. For non-English speakers it may mean nothing, for a car designer or mechanic it may mean far more. This is because in your lifetime your brain has developed a "model", encoded in neurons and synapses, of what a "car" is, how they differ, the appearances, how they work, and so on.

Any new thing for you to learn about cars has to fit into your existing model and modify your existing tens of thousands of neurons and millions of synapses devoted to your current understanding of cars, AND how cars link to everything else in your life.

Even if we had the tech to map the neurons and synapses (we don't), that is just analogous to a microscope, seeing them is not enough. We don't have anything remotely close to being able to manipulate the brain or create a new synapse, certainly not without surgery.

There will never be a high speed download of data, or skills, or knowledge to the brain. Thinking and learning will not get faster than the biological limits. That said, you CAN train your brain, through exercises, to remember far more than most people do naturally, learn faster, read far faster, and so on. Speed reading is a real thing, people really can train their brain to read and recite a book, and so on.


For starters, "muscle memory" is a misnomer. It all happens in the brain.

In the 1980s (I think) I read a study where the researchers trained mice to perform a task. Afterwards they harvested rna from the brains of the trained mice and injected it into untrained mice. The untrained mice exhibited the "trained" behaviors very shortly thereafter.

Of course, this meant that the trained mice were sacrificed in the process.

  • 6
    $\begingroup$ Problem is, that's pretty much completely wrong. In the 1950s to 1960s researchers fed parts of flatworms (not mice) to other flatworms (they're canabalistic) and this seemed to allow the latter to aquire knowledge that the former had. However their conclusion is treated with much skepticism and there has been a lot of problems with reproducing their findings. One alternative explanation to the one the original researchers had is that flatworms simply gain the stress hormones from the flatworms they eat making them more active and thus complete tasks slightly quicker. $\endgroup$ – AngelPray Dec 2 '17 at 16:39
  • $\begingroup$ I'm going to down vote my own answer because your comment provides a good context. $\endgroup$ – pojo-guy Dec 2 '17 at 16:52

In a related question I go over some details. In conclusion,

So if you numbered them, and noted the activation particulars, that would represent the thought. It's only meaningful in the context of the brain it was read from, and that changes over time: you can’t play it back later because the specific pattern matchers have changed their topology and detailed weightings, so the “same” pattern matching unit has a slightly different meaning if stimulated again in the same way.

Recording thoughts means using language, which is something the brain is already wired for. Only consider a language that is far more precise than any normal human language, and a custom system to decode it into the context of the current brain receiving it. This system would need to be more powerful and hold more data than a human brain! And the result would still be imperfect.

In general, it’s not a simple thing. The complete details of the wiring is specific to a single brain, and even representing this abstractly is a significant problem. (see chapter 12 in GEB).


Not the answer you're looking for? Browse other questions tagged or ask your own question.