Slashdot Mirror
Downloading The Mind
bluemug writes "The Canadian Broadcasting Corporation's popular science radio show Quirks and Quarks aired a piece this weekend about Ray Kurzweil's ideas on downloading human minds to silicon. (The interview is available in MP3 or OGG.) Kurzweil figures we'll have strong AI by 2029 and be able to copy a human mind about a decade after that. Book your appointment now!"
I have been thinking about this for a while now.. If you can download the mind - will we be able to upload it as well at some point in the future? I'm thinking along the lines of falling asleep in a body that's in its 70s, and then waking up in a body in its teens. It would certainly be interesting to relive my teens. A few things that could be done differently..
Stop the brainwash
I actually read about half of the book. I could not finish it as I was unable to read cause I was laughing too hard. I am not saying he's TOTALLY wrong. There may be a time when we will have computers that will be smarter than we are. When we will be able to download our minds into the computer. All of that is fine, his timeline is totally unrealistic.
;-) ). There was a conference, where one of the scientists started making wild predictions. Something like Kurzweil. Computers are supposed to be able to see (image recognition) as well as humans in 20 years, think in 30, etc. One of the other scientists has asked that guy:
A couple of points:
1. The estimates as to how much processing power is in an average human brain vary quite a bit. Is each neuron a bit? It can have multiple inputs - maybe it's something closer to a byte or a word? How and where is memory stored? Just haveing the raw processing power does not mean we will have the knowledge to USE it. We are seriously lacking in the knowledge departament.
2. Social implications. How many good technologies are set back, or even stopped because the people are not ready for it? Do you really think that an average person will simply accept and approve of the ability to live forever in a computer? All the religions of the world are going to have a field day with that. Don't think so? We've had genetically modified crops for a while now. They're safe and far more efficient. Why are there still countries that will not allow such crops to be used for human consumption?
In the end it reminds me of a story I've heard of a long time ago. I'm going from memory so you'll have to forgive me if I get the details wrong.
It happens during the height of Artificial Intelligence (when a lot of people thought we will have talking, seeing, thinking computers in just a few decades
"Why are you saying this? All of those problems are quite hard. It is unlikely anyone will achieve those things in that time."
The first scientist answered:
"True, but notice that every date I've given is AFTER my retirement."
What a way to generate funding, eh? This kind of things simply hurt the field in general.
And that's my gripe for this week. I feel a LOT better now, thank you!
Oliver Sacks' "A Leg to Stand On" illustrates how great an effect the loss of a single limb can have on the psyche of the victim. What would be the effect of the loss of the entire body? Kurzweil makes no mention of it.
I don't know about Ray Kurzweil, but I sometimes pay attention to parts of my body that are below my ears.
-----
Sorry, I'm only a 1336 h4x0r.
Kurzweil argues that strong AI will preceed the ability to download minds, which does not seem logical. It has been reasoned (by Pinker and others) that AI will be developed by reverse-engineering the brain, and artifically replicating its processes. The evolution of strong AI is thus dependent on technology to copy, and trace the functions of the human mind.
Your essence is trapped within the electrical/chemical field of your brain. Simply compying what a brain knows wouldn't do. You have to copy how it reacts. Even then, your brain's copy may or may not be imbued with its own intentionality.
:-)
Metaphysically this is about as practical as putting your soul in a brass pot for storage until you get your new body ready.
Maybe as a backup - then in the case of brain damage, memories could be reinstated.
But for my money - I think I'd prefer to be a brain in a tank mounted on a giant robot.
Say you have a class of nanobot which can absorb and replace the function of a single neuron.
You inject yourself with a load of them, and it starts absorbing neurons and taking their place. Eventually, your entire mind ends up running on these replacements, each of which behaves just like the organic neuron it replaced. You've been concious all the way through.
Now, assume each of these is able to communicate it's inputs to a machine on the outside which is able to simulate neurons en masse. They start to disable themselves and telling those around them to get their signal from this machine instead of them.
Eventually, you end up with a load of simulated neurons which are running on this machine, linked to the nerves through whatever method they use to communicate and a bunch of these neuronbots.
The simulated one is functionally identical to the original organic brain, except now it's got the potential to be pysically a lot more robust. Continuity was never lost, and all that was destroyed was a few neurons at a time, who's function was replaced.
If you're interested, check out "The Ophiuchi Hotline" or any of his short story collections. Unfortunately, most of his older works are out of print, but can be found at used bookstores and half.com.
He has one hell of an imagination and I highly recommend him.
"Downloading" a brain is a lot more complicated than copying a harddrive. Even if we figure out how the brain works, and then figure out how this contributes to a mind (neither of which we are close to understanding at all), downloading a brain is just a duplication of you. You yourself wouldn't notice anything, but your copy's memories would depart from your at the point of the brain scan from which the copy is instigated.
Ugh, there are so many loose ends its hard to pick one to pull on. Someone mentioned before, but your body is more than just a bunch of neurons floating in fluid. Your mind, your person, your sanity rely on constant bodily feedback. Your mind isn't just the brain, its the entire nervous system, head to toe. (check out Antonio Damasio's books Decartes's Error and The Feeling of What Happens for a thrilling discussion of this).
George Dyson's book Darwin Among the Machines doesn't address the stupendously anthropocentric idea of human intelligence on silicon but does explore some possibilities behind the emergence of intelligent (not necessarily conscious) systems on their own.
I read Dyson's book after stumbling across it browsing at a bookstore, only to learn that he lived about 2 miles from me! I went down to his boat shop and introduced myself and have had a few chats with him. He talked about Kurzweil a little bit and he actually gave me a copy of The Age of Spiritual Machines. At the time I was a naive fanboy (as opposed to the seasoned fanboy I am now) and asked him if he could write something in the book (I had him sign the Darwin book earlier). He declined, asking me with the ever present Dyson eyesmile, "What am I supposed to say? Sorry this book isn't as good as mine?" It was very humble humor, don't read it wrong.
I read Spiritual Machines and enjoyed it, if for no other reason that it provided a fun exercise in saying "that's a nice idea, but it won't work for these reason..." It addresses a lot of concerns and the whole identity dissolution theme was rather interesting to play along with. Still, I don't think that his future is a likely one.
Bah, I'm just rambling. Short end to a long story: Kurzweil's ideas are fun to read and worth the time spent if you have time to kill, but are highly unlikely. Copying humans into computers is a much bigger problem than just raw clock speed, which is what he boils it down to.
Here's a link to a page about Kurzweilian Singularity. Its worth checking out if you haven't read any of this stuff before.
Etc, etc, ad nauseam, and so on and so forth.
I agree with you. To be able to solve a problem using a computer, you have to know how to solve the problem in the first place. Kurtzweil (and his string AI buddies) are counting on some "emergent miracle" to occur.
Now, such an event occured in nature (i.e. evolution), but it took little longer than 30 years. :-)
...richie - It is a good day to code.
I don't know the history of this idea, but the book Mind Transfer (1988) by Janet Asimov was about the exact same thing - building a robot to hold you "self" that lived on after your biological body died.
Turing proposed that the ultimate test for an AI was to behave in a human-like manner such that a human observer could not discern the behavior of the machine from the behavior of another human.
Still, there are many who argue that although machines may one day pass Turing's test, they will nevertheless lack the essential consciousness or awareness that humans possess. See John Searle's paper, "Chinese Room". Nobody knows of a good, direct test for awareness.
Still others (Roger Penrose) do not rule out the possibility of genuine machine intelligence, but think that we have much to learn about our own minds before we can consider it seriously. Penrose specifically argues that our current understanding of science is too weak too incorporate an accurate model of conscious thought. But our science may change and one day become sufficient.
In any case, 2009 sounds like a very optimistic (pessimistic?) estimate.
...richie - It is a good day to code.
It seems to me that the ability to copy a human mind is almost prerequisite to strong AI. Sure, the "great AI winter" is at least partially due to the crash government funding the field enjoyed in the late 80's / early 90's drying up as suddenly as it emerged, but AI has always been a field prone to too-early predictions. It seems that with each new metaphor we invent for describing the human brain, we also convince ourselves that our minds really are as simple as our metaphors suggest. But Turing thought that human-level mimicry would be possible by 1990 (while at the same time vastly underestimating the quality of hardware that would be available in 1990).
There's a real possibility that we just aren't smart enough to figure out how we work, and so the only route to strong AI is to make monkey-see, monkey-do copies. And while procreation is a time-honored method of doing that, the structure of the brain suggests that serialized output was not high on God's list of priorities, and the biological format rather resists studies. So, I often think that we might have to be able to emulate the brain in silico or some other more easily-studied medium before we have a chance of understanding what makes that brain tick.
If you're claiming that we don't know that much about how the brain works, I'd agree with you. If you're claiming that it's going to be tough to figure out how it all works, I'd probably agree with you there as well.
However, if you're claiming that science can never understand the brain, I'd have to strongly disagree with you. As an atheist, I don't think there's anything so special about the brain. There's no soul there, put there by some random deity. There's no magic. It's just a lump of protein mixed with water, in essence. Sure, it's a marvellously complex lump of protein. but it's still a lump of protein. We've made a heck of a lot of progress understanding the behaviour of lots of other types of stuff using science. What makes this particular lump of protein any different?
Can anyone give me a non-religious argument why, at some stage in the possibly distant future, that the workings of the brain won't be entirely comprehensible to humans?
Any sufficiently advanced technology is indistinguishable from a rigged demo
--Andy Finkel (J. Klass?)
He concludes that physical continuity doesn't matter. He's got some good arguments (along the lines of some made in this thread) to back this up...
Grr! Arg!
On hearing the program, I'm feeling cranky about two things (and I speak as someone who was interviewed by Quirks & Quarks about studies in measuring brain activity).
First, I don't think Kurzweil has said anything that Hans Moravec ("Mind Children") and Marvin Minsky didn't say a long time ago. Minsky contemplated about machines transcending us, and Moravec long ago used Moore's law to predict when computers will be as complicated (he thinks) as human brains. Kurzweil is recycling other people's ideas.
Second, Kurzweil (like other MIT hardware guys) talks about the brain with the underlying assumption that it is just a collection of processing units (neurons) connected by simple electrical contacts (dendrites and synapses). In fact, the entire body of a neuron is chock-a-block full of calcium channels and tiny pores that are regulated by hundreds of different chemicals. Every year, new processes are discovered. Some chemicals are moved into the cell by active molecular transporters. Some chemicals may move between regions of cells by gaseous diffusion. Not only will you have to scan the connections between each neuron, but you're going to have to mimic the action of all this oozy stuff in real time using silicon.
And what about hormones and polypeptides that regulate all kinds of activities at short ranges, and also throughout the body? "Thinking" and decision-making involve lots of input from centres that excrete tiny quantities of chemicals -- all of this will have to be "scanned" (whatever that means) at a molecular level. It won't do to merely list the size and position of 100 billion neurons and their 100 trillion connections. You'll have to model the far greater number of wet chemical processes on every neuron.
In the 1940s some people thought everything would be "atomic" by 1990. Atomic rockets, atomic cars, atomic radios. Today, just substitute the word "computational" or "silicon" for atomic and you can blather about nonsense in the year 2040 without having a clue of what it means.
I think the brain's "wetness" is an integral part of it's operation, and this makes it a very dynamic and complicated thing. To simply see the brain as a collection of tiny silicon CPUs wired together is naive. It's a theoretical model straight from the 1960s or earlier, before we knew much about the brain at all. A real breakthrough in Artificial Intelligence will probably arrive slowly, and probably be stimulated by people who learned modern (i.e. post-20th century) physiology when they were young.
Hence, I think the term "an expert in computers and artificial intelligence" is an oxymoron at this time.
While we're on the tack of old science fiction, I vaguely remember a short story from my middle school years, so let me paraphrase:
Old Man "Bob" is wheelchaired into the waiting room of the hospital, where we find the rest of his family dressed in black, obviously in mourning. "Why is everyone so sad" he asks. "We just came from your funeral."
You see, "Bob" had a stroke, and died, however thanks to recent technology, he was able to save a copy of his brain about 3 months prior. The doctors cloned his old body, reloaded the brain. Of course the tech doesn't copy that well, so the life expentancy of the replacement is about a month because of cancers, but its enough time for the family to "bring back the dead", so they can all say their goodbyes in a way they couldn't the first time around. The only problem is that "Bob+1" didn't know he was only a copy, destined to die (again)...