Uploading your brain

I noticed that recently there was a dispute between Ray Kurzweil and PZ Myers over the prospect of reverse engineering the brain.  I didn’t read either’s argument beyond to find out that Kurzweil believes that we will soon be able to reverse engineer the brain and Myers does not.  Therefore, in this post, I’m going to take the philistine approach of stating my views on the topic without having read prior arguments that are readily available to me.  However, I have watched this Bloggingheads.tv discussion between Eliezer Yudkowsky and Massimo Pigliucci on the theoretical possibility of uploading one’s mind to a computer, so I’m not completely in the dark as to what prior arguments are on this topic.

As to my position on that matter, I favor Yudkowsky’s view that uploading one’s mind to a computer is theoretically possible.  However, I don’t think that the technology to do so will exist any time in the foreseeable future, which probably puts me closer to Myers’ view in his argument against Kurzweil (though I didn’t read either, so I don’t know for sure).   So, first I will  cover the theoretical level, since if brain uploading is theoretically impossible, the practical issues aren’t worth considering.

Massimo Pigliucci’s view on the theoretical possibility of properly emulating a human brain in silico was based on an analogy to photosynthesis.  In photosynthesis, a plant takes up light and through a complex array of steps converts carbon dioxide and water into carbohydrates.  We know all the steps in the process and can simulate it on a computer.  However, if you do so, you will find that you have not produced real sugar.

In much the same way, Pigliucci believes, a simulation of the human brain may accurately represent all the processes involved in cognition and consciousness yet not actually produce the real products of brain activity, among them being thought and consciousness (which are not necessarily mutually exclusive).  Like the sugar, consciousness would be unattainable through a computer emulating the processes involved in the human brain.

Now, I don’t think that Pigliucci’s point is, on its face, ridiculous, but I do think that it’s ultimately mistaken.  The human brain and a computer are run on two very different media, with our brains consisting of neurons, cells that fire action potentials and communicate through chemicals called neurotransmitters in a very biological manner, whereas computers run through the interactions of electric current with transistors, the reading of magnetic media or static charges, coupled with many other circuits.

So I think its a very good idea to ask whether the biological nature of the brain prevents its activity from being emulated in toto by a computer.  However, I think a little bit of reflection should reveal a fundamental difference between cognitive processes and photosynthesis.  In photosynthesis, a cell is taking in an energy source in order to interconvert chemicals, taking in the energetically stable carbon dioxide and water and using light energy to generate a molecule with a higher energy state that can maintain the organism’s life processes while still obeying the second law of thermodynamics.

Like photosynthesis, biomolecules are the currency of cognitive processes.  However, the analogy doesn’t hold up because the inputs and outputs are different.  In photosynthesis, the inputs are a set of chemicals and light and the outputs are a different set of chemicals and heat.  In cognition, metabolic processes work as the mechanism, but ultimately both the inputs and outputs are forms of information, and this is where Pigliucci’s analogy breaks down.  A computer cannot meaningfully emulate photosynthesis because both the inputs and the outputs are medium dependent–a process that does not process carbon, oxygen, and hydrogen atoms cannot be said to have performed photosynthesis.  However, cognition itself does not require a specific toolkit of chemicals.  Instead, it requires information to be processed in a particular manner.

It just so happens that what computers do is process information.  So in order to emulate cognition satisfactorily, a computer would need to process incoming information in such a way that it would output information in the same manner as a biological brain.  Computers are designed to perform information processing, so conceivably, if a computer was sufficiently complex as to emulate all the mental processes of the human brain (as distinguished from metabolic processes), the computer could be said to have achieved a successful emulation of a human mind and thus would be considered a conscious, thinking computer.

From that point it’s pretty easy to conceptualize a computer that successfully emulates a particular human mind, and that is what I believe it would mean to say that one’s brain has been “uploaded” to a computer, though I think it would be more accurate to say that one’s mind rather than brain has been uploaded.  This of course, brings us to the next question, which is: “Will we be able to upload our minds to computers?”  This is quickly followed by:  “When?”

Here I see some difficulty.  Actually, a great deal of difficulty, and perhaps an insurmountable level of difficulty.  There are multiple barriers to uploading one’s mind to a computer.  The first and most obvious is that we don’t have computers that are powerful enough to emulate a human mind.  The human brain consists of tens of billions of neurons, each connected to a multitude of other cells.  However, that’s not the full story.

They fire slowly relative to computers at about 100 Hz compared to my five year-old laptop’s processor, which runs at 1,500,000,000 Hz.  Of course the difference is that we’re comparing individual neurons, each acting in response to a different stimulus, to a processor that runs commands through single-file, but this raises a very important issue:  the “wiring” of the brain is fundamentally different than that found in a computer.  While that doesn’t mean that Massimo Pigliucci is right and the emulation of the human mind by a computer is a theoretically impossible task, this fact adds a substantial layer of difficulty to the process of emulating the human mind.

And it doesn’t stop there. Different neurons and different neuronal pathways use different neurotransmitters.  There’s serotonin, dopamine, acetylcholine, epinephrine, and glutamate to name just a few.  So an emulation would have to not only keep track of the massive number of synapses that form the intercellular communication medium, but also what type of neurotransmitters they use and what effects those have.

Look at drugs.  Alcohol, caffeine, THC, nicotine, cocaine, heroin, and methamphetamine are all chemicals that have different effects on the brain.  And it should be noted that these effects are dosage dependent and at the dosages most typically taken by drug users, they leave the brain as a functioning organ.  It’s function is modified, but in all cases, a user is still able to feel, walk, talk, breathe, eat, drink, hear, etc.  A computer that correctly emulated the mind would have to be able to emulate the effects of drugs on the mind.

Each little detail about the human brain makes the endeavor of emulating its main process look more and more taxing on human ingenuity.  There is a much longer list of characteristics and specifications that must be met before a computer emulation can even approach such a goal.

Once all these issues are overcome, what one has is a computer that can emulate the human mind.  However, what you don’t have is a computer than can emulate a particular human mind.  To do that you are faced with a new array of complications.

The biggest is the extraction of mental information from a living person.  Unlike a modern computer, brains don’t come with standardized input-output ports.  You can attach a hard drive to your desktop computer via USB port and copy all the information from its hard drive.  The human body has no such design feature.  A human body is build from a single cell dividing and organically assembling itself into a functioning multicellular organism whose functions are self-contained.  There has been no evolutionary impetus to make the mind a transferable component.  As such, copying the information therein is much more difficult than is suggested by the phrase that titles this post.

As of now, our brain imaging technology is, well, rudimentary doesn’t quite capture the magnitude of the challenge faced here.  You would need to be able to determine the location, connections, and neurotransmitters used by every single neuron in the subject’s brain and you’d have to do that with 100% fidelity.  This is made worse by the fact that the brain is continually operating and changes in brain state often involve the physical movement of neurons.  If the technology to do so is invented within the next century, I will be very surprised.

Of course, once you’ve solved that problem you have to look at why you’re attempting to upload a person’s mind in the first place and how that squares with your engineering effort.  Brains aren’t static entities.  They act differently with age.  A two year-old doesn’t see the world the same way that a twenty year-old does and the point of view of an eighty year-old is utterly alien to the previous two.  A brain changes with age and the mind it generates changes correspondingly.

Now, it seems to me that one of the strongest motivating factors to upload one’s mind is the prospect of immortality or at least greater longevity.  No one really wants a perfect emulation of his* mind, as such an emulation would age and die just as a real brain would.  A perfect emulation would indeed

Another point of interest is that the human brain was not designed in a vacuum, but rather as an organ in the context of the entire human body.  Without a surrogate body, the emulated mind would be under sensory deprivation, which can drive a person insane.  So some sort of surrogate body would have to be designed.  But it would be strange to attempt to emulate every aspect of a human body.  A robot would presumably have different needs and thus one would hope that impulses such as hunger and thirst would be treated differently.

All of these modifications, which seem to me to be necessary to make the whole endeavor worthwhile, raise the question as to how much such an uploaded mind would be “you”.  How we react to hunger, thirst, infatuation, lust, affection, and pain among many other drives and feelings forms an important, but not always considered aspect of who we are.  They fit in with a biological paradigm.  Our bodies, including our brains, have been developed over eons of evolution to perform the tasks necessary to maintain the information that builds them.  But once you’ve transfered the processes

Ultimately, beyond the technological challenges ahead of any set of engineer attempting to upload someone’s mind to a computer, there is a question of what would be achieved by doing so.  It seems to me that by the time we developed the capacity to emulate the mind of an actual person, we would have far more powerful artificial intelligences that are better optimized to their roles than an emulated human could ever be.

*English has no completely gender-neutral singular pronoun that can be correctly applied to humans.  In situations where gender is indeterminate as above, the use of the male pronoun is correct.  If you feel that this is a point worth discussing, feel free to comment on the matter and I will elaborate on my position.


About Meng Bomin

Real name Benjamin Main, I am a graduate of Grinnell College with a degree in Biological Chemistry.
This entry was posted in Science, Technology and tagged , , , , , , . Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s