Thursday, October 8, 2020

How to Prepare Your Data and Brain to Have Your Mind Uploaded to a Computer

This blog post will discuss life extension by means of artificial intelligence. There are two main methods. You can 1) use data collected about you to build an avatar of yourself, or 2) use actual memories from your brain and upload them to a computer. Both of these methods are currently in the realm of science fiction, but they may be viable within our lifetime. Thus it might be worth your time to act now to increase the odds that you achieve digital immortality. Namely you could choose to preserve your data and your brain so that they are available to be uploaded to a computer when the technology matures.

I am quite confident that even if you were to die next month, current techniques are capable of preserving your brain in enough detail that you could be resurrected from the data in the future. Just paying to have your brain preserved though, does not ensure that it will be uploaded. However, it is quite possible that merely preserving it will increase the odds that it will be uploaded by altruistic scientists in the future. As we will discuss, it is not clear if this upload will be a vessel for your current, sentient consciousness or just an uncanny replica of your memories and personality. Because brain preservation is an opportunity that is just now becoming available to humanity I think we should all look into it, and consider it. However, because the outcome is uncertain we should also make an effort to make peace with the fact that we may be mortal and will never have our mind uploaded.

Creating an Avatar from Data

You certainly don’t need a physical brain to construct a decent simulation of a human being. Instead you could take data from a person, living or dead, and attempt to reconstruct their personality and likeness. You could do this for a person like Aristotle (385-323 BC) by programming it to reflect his writings, and everything we know about him. But a much more precise simulation could be made of someone like Bertrand Russel (1872-1970) who we have actual photographs and video of. Even using today’s technology specialists could take recorded conversations, photographs, and video footage of a person and combine it with a generic chatbot to create a convincing digital simulation of someone. Just a minute of video footage of someone talking can be fed into an artificial neural network to create a highly realistic “deep fake” of both their voice and their appearance. So if you want to have a chance at recreating a parent or grandparent (or yourself) you ought to start collecting data on them now. There are many kinds of easy-to-procure data that could help to create a digital recreation of someone in the absence of a brain specimen.

A default AI capable of holding a conversation and reasoning effectively could form a starting point. It would amount to a generic reproduction of the brain’s algorithms and data structures. Then it could be updated with a specific individual’s characteristics, and eccentricities. The more data you have on that person the more realistic the simulation. Here is a list that I brainstormed of sources of data that could help to describe someone as an individual and thus could be used to help fashion their digital identity:


photos,

home videos,

emails,

letters,

art,

creative work,

diary,

voice memos,

SMS texts,

phone call log,

search engine history,

internet browsing history,

books read (Kindle, Goodreads),

their videogame records (trophies, scores),

music playlists,

movies and television  (Netflix viewing history),

youtube history,

recorded phone calls,

social network,

social media posts and likes,

map and GPS history,

travel history,

vitals & body measurements,

medical records,

structural brain scans (MRI),

functional brain scans (fMRI),

psychological evaluations,

personality tests,

psychometric tests,

school records,

standardized testing,

yearbooks,

surveillance records,

legal and medical history,

wardrobe,

childhood toys,

belongings and property,

online purchase history, 

genome sequence,

epigenome sequence,

videos of their gait,

habits, mannerisms, posture,

terms, phrases and colloquialisms,

a list of their values,

a description of their morals,

ethical stances,

spiritual convictions,

recounting of fondest memories,

pictures of their homes,

historical setting,


You could certainly start gathering, saving, and safeguarding this kind of data for yourself today. I am sure that in the near future there will be businesses that help people to curate their data. Such a business might also help to create a structured self-report questionnaire that asks people questions about their likes and dislikes, and interrogates them about what makes them unique, interesting, and sets them apart from others. Such a questionnaire should ask about things that would not be obvious from the person’s data. You could fill out the questionnaire about yourself, or about your loved one. Some of this data might equate to a binary setting (e.g. I like loud people or I don’t), but some of it could be used to train artificial neural networks to create a convincing simulacrum of the person in question.

You could also collect data using detailed 3D photographic scans and motion capture video in a performance capture environment. This could help the system to accurately recreate the person’s lip movements, facial expressions, microexpressions, eye movements, pupil dilation, sweating, intonation, speech patterns, voice stress, and much more. Clearly this method could create a convincing copy or mimic, but without a brain specimen it certainly wouldn’t qualify as a form of life extension.

Creating a Digital Reconstruction Based on Brain Data

Life extension through brain preservation began in earnest with the freezing of bodies. Cryonics is the low temperature freezing of a human corpse or severed head with the hope that resurrection will be possible in the future. Robert Ettinger first discussed it when he published “The Prospect of Immortality” in 1962. Since then there have been many companies that offer the service of keeping corpses in vats of liquid nitrogen. Cryonics is often characterized as pseudoscience, and those that practice it have been called quacks, but as the relevant technologies develop this will change.

There are currently three such cryopreservation companies in the U.S. and one in Russia. As of 2014 about 250 corpses had been cryogenically preserved, and around 1,500 living people had signed up for preservation. When a customer opts to just have their brain and not their body preserved it is called “neuropreservation.” Of course, this is cheaper. Depending on the company and the method, cryopreservation can set you back anywhere from $28,000 to $200,000. There are a number of costs: medical personnel have to be on call for death, the body must be transported quickly to the facility, the preservation process has to be performed by medical experts, and the body must be stored indefinitely. In many cases individuals set up a trust fund to cover storage and revival costs. The cost of cooling and storage have already shown to be substantial. Many cryonics corporations have gone into bankruptcy. In fact, as of 2018, all but one of those that came before 1973 had gone out of business and were forced to thaw and dispose of the corpses that they stored. Consider the fact that most businesses have a one in one thousand chance of surviving even one hundred years and you get an idea for how tenuous this is in its current form.

Initially the idea was to freeze a body until medical science advances to a point where it can be reanimated and treated medically. However, freezing temperatures cause damage to tissues and cause individual cells to break, destroying the information stored in the connections between them. Certain chemicals called “cryoprotectants” can prevent ice formation during cryptopreservation but they also cause damage making it so that the corpse cannot be reanimated. These limitations of cryonics have led to other options.

A company called Nectome uses a chemical called glutaraldehyde to perform a “100% fatal brain preservation” procedure. It is fatal because the procedure creates chemical crosslinks between protein molecules that eliminate biological viability. So unlike with freezing, the brain can never be resuscitated… but it can still be mapped. It must be doused in the glutaraldehyde quickly after death though. The brain’s cells start to break apart and die (the cell membranes rupture) soon after death due to lack of oxygen. This is why this procedure must be performed either immediately after death or on a live person under general anesthesia. In the second case it is a form of assisted suicide or voluntary euthanasia. Why would someone want to undergo this procedure? So that their brain could be mapped, transcribed into digital data, and reconstituted within a machine.

The molecular details of a brain would probably not be needed to create a high fidelity map, but the cellular details would be essential. All of the person’s neural connections (their connectome) including cell type and location, information about the cell membrane and intracellular structure, as well as the positions of hundreds of millions of axons and trillions of synapses would have to be scanned by a computer. There are around 100 billion neurons in the human brain. Given that each of these can have up to 1,000 connections to other neurons each, that gives us 100 trillion neural connections to hold in computer memory and model. A brain map or connectivity database of the anatomic connections of a human brain has been estimated to occupy less than 20,000 terabytes. Today this would cost over $300,000, but in a few decades this price will drop significantly.

To upload the connections the brain must be cut into extremely fine slices and scanned with an electron microscope. This is done today, but currently it is done very slowly, and no human brain has ever been mapped in its entirety. If the procedure could be fully automated like gene sequencing has been automated in the last two decades it could become quick and cheap. It is pretty easy to envision a distant future where there are automated factories that takes preserved brains from the past and reanimate them digitally. It may sound absurd, but this could be our salvation: a fountain of youth, a portal to immortal godhood, and a source for eternal life.

This process of scanning a brain and copying or transferring it to a computer is called whole brain emulation or mind uploading. The neurons could be simulated by hardware and the software they run would come from the mapped brain. If this was attempted today it would have to run on silicon microchips, but by the time this becomes viable other forms of hardware may be available such as optical, neuromorphic, or quantum computing.

You wouldn’t want to be among the first to be resurrected because the tech will exhibit exponential improvement and the early methods will be less effective, and more destructive to brain tissue. So you would want to stipulate in your will, trust, or contract that you would like to wait until the technology fully matures. This may be hundreds or even thousands of years after it becomes technically feasible.

Most leading experts today believe that advancements in AI, computer science, and brain mapping will come together to result in artificial consciousness. An uploaded mind with artificial consciousness could inhabit a robot situated in the real world, or could be situated in virtual reality or cyberspace. Transhumanists and futurists see mind uploading as the most viable form of life extension technology. There are definite benefits in leaving our organic components behind. If we were made of metal, fiberglass, and silicon we could better withstand accidents, damage, the vacuum of space, and the passage of time. Mind uploading could even help humanity survive a catastrophe on Earth if it were to become inhospitable to biological life.

To help you recreate a person it might also help to mine your, or another person’s, memories of them. For example, if your brain had been uploaded you could use your memories of your grandparent to help recreate them, or to help train the system that is attempting to emulate them. If a pet’s brain could be restructured digitally it could be mined for a very high level of detail about someone’s movements and emotional reactions. The digital memories of whole families could be used to reconstitute each other with even higher fidelity. One could also collect information about (or mine the brains of) teachers, classmates, childhood friends, roomates, coworkers, etc., and use these to refine and reconcile a model of someone.

Once you are dead, you won’t be in any hurry to be brought back to life. You will be unconscious until you are resurrected so it will feel as if no time passed at all, even if it took one thousand years to resurrect you. If the procedure is done properly the new version of you should remember his/her last day like it was yesterday. However, you may not be able to recall the last few hours before your death because the chemical and electrical changes responsible for short-term memory would not have had the time they needed to be changed (consolidated) into physical changes responsible for long-term memory. These traces are very subtle and are likely to be very difficult to ascertain and record.

Would It Really Be You?

Would minds that have been uploaded retain a sense of historical identity with their past self or would it be like being replaced by a twin or doppelganger? If performed exactingly the mind upload technique would probably result in a person that even our friends and family could not tell apart from us. This new person themselves might even be convinced that they are us, but that doesn’t mean that we are them. The important question would be: “Would we feel like our sense of consciousness and personal continuity continued to live on in this system? Would we experience identity perseverance through time after having our mind uploaded from our body? Or would we be effectively dead?” Personally I don’t think any kind of technology developed in the next hundred years would allow us to feel like we have woken up within a machine. But thousands of years from now, that may be a possibility.

Take the teleporter from Star Trek for instance. When Captain Kirk steps into the machine his cells and molecules are read by it and the data it collects is used to create a perfect copy of him in some other location. As this is taking place his original body is destroyed, but the new copy has all of his knowledge and memories, and even perfect recollection of the thoughts he was having before he entered the machine. But let’s be honest, the original Captain Kirk was killed. If his body had not been destroyed in the process, there would be two of him, but neither would feel like they were two people. In fact, it would be possible to convince the original Kirk that the teleportation hadn’t worked and he would have no way of knowing that he had an exact double on another planet. He would feel no psychic connection with it.

Some scientists have speculated there are two ways to overcome this problem: 1) you could slowly turn off a person’s brain as you slowly turn on its digital recreation as if you were pouring the contents of one container into another, and 2) you could gradually replace neurons with their electronic equivalents one by one until the entire brain had been replaced. I don’t think the first solution solves the problem. I think the second one might if done properly, but is too messy and complicated to be feasible in the next several hundred years. The second technique is analogous to the thought experiment of “the ship of Theseus.” Theseus gradually replaced the parts of his ship until the whole ship had been updated, but the question considered by philosophers is: “Is it still fundamentally the same ship?” It is important to point out that both of these methods requires a living person and could not be performed with a “neuropreserved” brain. Can there be continuity between two selves separated by death?

The feeling that you are the same person that you were 5 or even 50 years ago, involves an illusion. You are not that same person today. Your interests, memories, and values have changed substantially. Keep in mind that every year 98% of all of the atoms in the body are replaced. In fact, when you wake up in the morning you are not the same person you were when you went to bed. Billions of connections throughout your brain were altered while you slept. We even lose continuity every time we take a mind altering substance. Accidental drowning, near death experiences, coma, anesthesia, hard drugs like psychedelics, and the passage of time all cause the erosion of neurological continuity and personal identity. What would it mean for you to die, and then have a version of you be resurrected 10 or 100 years later? Would that be similar to the case of Star Trek teleportation? Or would it be a just another example of imperfect continuity that we already accept and take for granted?

Ok, here is a simple hypothetical. Imagine being teleported to a base on the moon. If your previous body was destroyed your personal conscious awareness would be annihilated with it despite the fact that your duplicate will behave as if it was you. But now imagine that rather than being transported to the moon, you are merely transported three feet to the right. Same outcome right? Ok, imagine that you are transported one nanometer (a billionth of a meter) to the right. What if your duplicate was compiled in the exact space that you take up now, perhaps using the same atoms and molecules? It would feel like it was you, but would you feel like you were it? This hypothetical scenario tells me that there is something illusory about the persistence of personal identity through time. You are not the same person who started reading this paragraph, or even this sentence. You might as well have been duplicated multiple times.

Not only has your brain changed physically and chemically every morning when you wake up, but countless physical vibrations and electromagnetic rays have passed through it. Because the Earth revolves around the sun, and the solar system revolves around the Milky Way your brain is now in a vastly distant location than it was when you went to sleep last night. During your eight hours of sleep trillions of tiny microscopic changes have taken place in your brain due to metabolism, homeostasis, learning, and entropy. Because of these constant changes, continuity in your personhood is broken down on submillisecond time scales even when you are awake, so it’s not clear how important it is that your uploaded AI brain feels 100% continuity of consciousness with your previous biological self. What is really important is to preserve as much of the identity, values, creativity, intelligence, and humanity as possible. 

The Costs and Benefits to Humanity

To some people mind uploading sounds creepy and unnecessary, but even these people must admit that it accomplishes the goal of preserving interesting and important data as well as preserving the diversity of our species and of intelligent life on Earth. Each person has their own perspective, insights, and intuitions. Is it wise to just let these decompose with the rest of the body? Historians go to great lengths to preserve books, cultures, languages, movies, and even videogames. Our societies preserve mummies in museums and academics try to make all possible inferences about historical figures and events from prehistory through antiquity, and on through to the present age. Why wouldn’t we want to preserve brains and minds?

Because uploaded minds could be run within a simulation there would be few costs. If it costs pennies, creates no waste, and does not contribute to pollution or overpopulation, wouldn’t it be preferable to have a digital version of your grandparents? Why not? In the future, if these uploads didn’t require a lot of energy or take up a lot of space, I can see corporations or even the government getting involved and making sure that corporations in the business of brain preservation that fail do not trash their corpses because of the value of the information that they contain. If two heads are better than one, doesn’t that mean that we want as many heads as possible? Why not have a communal system of digital intelligences? In comic books and science fiction it is already a staple. For instance Marvel comics alone has the Xandar World Mind, The Kree Supreme Intelligence, the Phalanx hive mind, and the Eternal’s Unimind.

If the costs to running uploaded brains are inconsequential then there will not be many barriers. In the distant future the computer memory and processing resources needed to run the equivalent of a human brain will be very small. Your brain uses about the same amount of electricity as a 60 watt lightbulb and this could be reduced dramatically by technology. Moore’s law and Kurzweil’s law of accelerating returns suggest that this could happen in just a few hundred years. If you could run millions of lives in their own chosen simulated realities on something like a phone, wouldn’t you? Of course it would have to be overseen by some kind of ethical governing body because there are ways it could go wrong.

There are certainly some risks and downsides. Your brain could be destroyed accidentally (or on purpose) before you are brought back. This is a problem because, as of today there is no way to back it up. Even more disturbingly, your brain could be stolen and placed inside of a nightmare simulation. It would be possible for a malicious person to upload you to a computer, enhance your senses and intelligence one millionfold and then subject you to the worst torture imaginable for the rest of eternity. You would be totally disembodied, with no way to reach out to your own hardware, so you could not commit suicide, and there would be nothing you could do to escape. This is a serious concern, especially given that this could happen to people, or even to copies of people, and the device involved could be hidden so that any kind of law enforcement that exists at the time would have trouble finding it.

Acclaimed futurologist and inventor Ray Kurzweil believes that he will live long enough to be digitized before he dies. He is currently 72, but he points out that technology and medicine are advancing exponentially and thus should be able to keep him alive long enough to see the next major advancement. He uses the analogy of a bridge to another bridge to describe how future medicine will be able to keep extending lifespan until brain emulation technology finally matures.

Let’s assume that brain emulation technology will be feasible by the year 2100. For someone who is 70 today to reach this time the medicine of the future will have to be able to extend their lifespan to 150. Experts in gerontology estimate that almost everyone would develop Alzheimer’s if they lived to be 130.  However, even if you died with profound Alzheimer’s and severe memory loss, the data could still be mined, and your synthetic brain could be free of Alzheimer’s and have full recall. Much of Alzheimer’s disease is an issue of data access due to reduced brain metabolism. In most cases it is not necessarily an issue of complete loss of memory traces (although cell death and brain shrinkage are issues). Thus much of the mental aging or cognitive morbidity you suffered in old age could be reversed, because increasing the energy output of an artificial brain would be as easy as turning a knob.

Upgrading Your Own Hardware

I would want to live forever, even if it meant that I had to remain in virtual reality within a computer. This is partly because very soon, virtually reality will be much more interesting and stimulating that actual reality. Just look at the progress in videogames and computers in the last 40 years. Think about the jump from Pong to a game like Red Dead Redemption 2. We have gone from simple sprite graphics to vast, photorealistic, polygonal worlds. Virtual reality will become insanely immersive within our lifetime and its quality will continue to increase exponentially. But the digital environment won’t be the only thing that shows exponential progress. Your mind and consciousness will too.

Today our brains are stuck at a fixed energetic capacity because our hunting and gathering ancestors could only find so much food in a day. Our neurological blueprint is encumbered by the metabolic constraints of our past. Not so for machines. You could easily turn up the juice on an AI. You could increase the processing power and speed easily. You could add as many neurons and synapses as you want. This would expand the level of consciousness.

Working memory, and intelligence, could be expanded millions of times. There would be many ways to do this but one of the most interesting ways would be to manipulate something called “sustained firing.” The ability of neurons in the brain’s association areas, such as the prefrontal cortex, to engage in sustained firing allows them to maintain whatever information they encode for as long as they keep firing. This firing can last up to a minute at a time and allows us to keep specific representations active for sustained periods. If we didn’t have this ability we could not have a train of thought and thus. Sustained firing in humans lasts for longer than any other animal, but if it was made even longer then we would be less forgetful, near-sighted, and impulsive, and far more intelligent. Merely, increasing the duration of sustained firing in an AI could vastly expand its awareness and mental capabilities.

Using techniques like these to amplify the working memory of a digitally reanimated brain would result in interesting abilities. A digital brain that could coactivate many more parameters and specifications (memory fragments) would perform searches for associations with much more specificity. This could result in the ability to completely recall events that a biological brain could not. This could allow you to recover distant memories that you have long since forgotten. In fact, you might be capable of remembering virtually any semantic knowledge that you had acquired before (such as facts about the world), and also a great deal of episodic memory (such as minute details about every birthday you ever had).

If the hardware and software is consistently upgraded your intelligence and knowledge will grow geometrically. Constant tweaks, additions, and improvements to our artificial minds would cause us to rapidly gain computational power in the same way that computers did in the last 70 years. Thus as we aged we would consistently get smarter. Each day you would think faster and more comprehensively. But even more excitingly, this computational power would enhance our very sentience. We could become incomprehensibly intelligent and commune with other superintelligent beings in fantastic ways. Imagine a hyperintelligent future version of yourself that was able to engage in a form of profound post verbal communication with hundreds of other entities simultaneously.

Imagine learning new fields of science, mathematics, and engineering in seconds and having the mental wherewithal to put them to use creating not only new theories, but practical uses as well. You will be free to learn about, explore, and contribute to whatever endeavors, or lines of progress you wish. You will also be able to watch all of the fantastic social, scientific, and technological advancements and breakthroughs being made by others. Imagine living in the year 30,000 and having encyclopedic knowledge of everything that has happened and everything that has been discovered. The neural circuits associated with physical and emotional pain could be cut out, and those associated with pleasure, excitement, and love could be amplified. Imagine living in a cyberspace afterlife paradise with trillions of times more cognitive resources than you have now. I want that.

It is pretty clear that missing out on brain uploading may be like missing out on heaven. It has the potential to provide everyone alive today with eternal bliss, replete with endless knowledge, superintelligence, limitless growth, incomprehensible beauty, and unlimited connection. So you might want to start looking into curating your data, and preserving your brain. As we have discussed though, for people alive this century the desire for a postmortem existence could all be in vain, and for that reason we should also come to peace with the idea that we may not be immortal. Also we should not let the specter of digital immortality in any way diminish the value of a good and worthwhile life, or our sense of peace with the natural way of things.

No comments:

Post a Comment