Just What Is Our Future Anyway?
A lot of my current thinking has been influence by Ray Kurzweil and his concept of the Singularity, and the Law of Accelerating Returns. If one looks back at trends in technology of all kinds, they follow an exponential curve. There is no reason to believe (barring a catastrophy), that these trends will not continue in the future. Limits in one technology give way to the potential in the next paradigm, which fuels continued exponential growth. Technology, says Kurzweil, is evolution by other means. Ultimately, and soon, as we round the knee of the exponential curve, we will reach a point where significant transformations in our way of life will happen so fast, that ordinary humans will not be able to participate in a meaningful way. I’ll try to explain here what this means for us “ordinary humans” and where this will eventually lead us in our place in the universe.
Exponential Growth
Exponential growth is not very intuitive. We humans tend to think linearly. We look back in time at the progress we’ve made in the last 50 years, and expect to make similar progress in the next 50. People who try to predict future trends often think linearly. This leads to optimistic predictions in the short term (because the details were not well understood), and pessimistic predictions in the long term (because the growth was far greater than a linear view).
When something grows linearly, it means it grows by adding in some amount over each period in time. If you save $100 a week by stuffing it in your mattress, your accumulated wealth will grow linearly. Every week you’ll add $100 more. Every year, your savings will grow by $5200. After 20 years, you will have accumulated around $100,000.
When something grows exponential, however, the growth is multiplicative rather than additive. It increases by a percentage of the current amount. Imagine that instead of stuffing your $100 in your mattress, you give it to an investment banker who puts it in mutual funds and equities that return 12% annually in growth and dividends. At the end of 20 years, you’ll have more than $350,000. In another 20 years, the difference is even more stark. After a total of 40 years, your mattress savings will be up to $200,000, while your investment account will be worth an amazing $3,800,000. After 60 years, compare $300,000 to a startling $38,000,000! (This is assuming, of course, that you don’t have to pay taxes along the way. J )
Exponential growth can be a very powerful thing. In the above example, with 12% interest, we double our money about every six years. If you look at compute power historically, we are doubling at a rate greater than every 2 years! This is true even if you go back to the early 1900’s to mechanical computing devices, and follow the trend through relays, tubes, transistors, and integrated circuits. However, it is only in the last 20 years that computers have begun to have a significant effect on our lives, and the trend is only accelerating.
Exponential growth can sneak up on you. It seems very slow and insignificant at first, but that can change dramatically as we round the knee of the curve. Consider the following graph showing the exponential nature of human population growth over the last 2000 years:
Note that it took us tens of thousands of years since modern man came into existence just to reach the first billion of population around the year 1800. In just a little over 200 years since then, the population has exploded to nearly 7 billion. (Of course, with population growth, there may be resource issues and population controls put in place that act to curb this growth, as can be seen in future years where the graph is expected to roll in the other direction).
I had my own experience with exponential growth sneaking up on me recently. Before I went on vacation for a week, my kids’ above-ground pool had a few insignificant splotches of algae that had taken at least a week to grow on the floor of the pool. If I’m gone for a week, I shouldn’t have to worry about it, right? Well, not exactly. When we returned, the entire pool was neon green! Two gallons of shock, and a week later, we finally had clear water again.
When it comes to technology, we are currently at the knee of the exponential curve. We know this because paradigm shifts are occurring multiple times within our lifetimes, and occurring more and more frequently, and adoption is taking less and less time. Just look at how fast the adoption of the world wide web took place!
GNR
GNR refers to 3 major areas of technology that Kurzweil talks about – “Genetics, Nanotechnology, and Robotics (specifically Artificial Intelligence – or AI).” These are the areas of technology that will have the most profound impact over the coming decades.
I’m not going to spend a lot of time talking about the specific technical advancements that have been made in these areas. Kurzweil spends a good chunk of “The Singularity is Near” covering these in detail.
I will talk about some of the major impacts, and what it means for us going forward.
Out of the three, genetics/biotechnology will have the greatest near-term impact. We have already seen real applications in this area, from cochlear implants, to treatment of Parkinson’s disease using electrodes in the brain, to more directed forms of cancer treatments, to genetically modified crops. There have been experiments in neural implants that interface the brain and nervous system to electronics. A monkey was trained to control a robotic arm with the power of its mind alone. Over the next 15 years, we will see rapid advancements in treatment of disease, helping those with disabilities, and using circuits to overcome damage in the brain and nervous system. We will even make real progress toward modifying all the mechanisms that cause aging.
While the power of controlling biology has its benefits, it also has its risks. Genetically modified viruses, or even viruses created from scratch are a dangerous possibility that we need to be prepared for. We need to keep moving forward, though. We can’t turn our back on technology and bury our heads in the sand. As TR said, this only exposes our backsides to the world. This is true – since those with malicious intent would continue to develop the technology, while we would now be powerless to combat it.
While nanotechnology is a little further behind, its implications are even more profound. We’ve been able to do some relatively simple things so far – make simple gears and motors and carbon nanotubes, move individual atoms around, etc. Eventually, we’ll be able to make nano-sized robots that can perform all kinds of functions – from manufacturing everyday products to making cellular repair in the human body. Ultimately, material things will become very cheap to manufacture. The true value will come in the information – the pattern of how to make them. As I had mentioned in another post, the world is moving ever closer to one based entirely on information/IP.
Nanobots in the human body will be able to carry out a whole host of other functions – from delivering medicine, to making repairs, to replacing blood as the carrier of oxygen. Nanotechnology will also eventually allow scanning of the human brain from the inside, which will complete the details we need to create computational devices that accurately replicate the function of the brain. Nanotech will also ultimately allow for direct neural stimulation that creates virtual environments that are as convincing as true reality. Just imagine the possibilities there! Kurzweil believes that the nanotechnology revolution will be in full swing, come the later part of the 2020’s. Here’s an interesting related article -- The Human Machine Merger: Why We Will Spend Most of Our Time in Virtual Reality in the Twenty-first Century
The dangers associated with nanotechnology are even more ominous. The “grey goo” scenario is what happens if an uncontrolled and malicious self-replicating nanobot gets loose that can exist and replicate itself in a normal environment. If there is not a nanobot “immune system” already in place in our bodies and in the environment, it’s all over. The entire biosphere could be reduced to dust in a matter of weeks – much faster (hours) if a smart “Trojan” version invades the biosphere undetected before full-scale replication begins.
Does this scare you? It should. It scares the heck out of me. It has also scared professionals in the nanotech field to already start thinking about measures to help make sure this never happens. Certain ethics must be followed – such as no self-replication allowed that can take place in an uncontrolled environment. But this doesn’t eliminate the possibility of an accident, or malicious intent. This is exactly why we need to stay on top of it and have countermeasures ready. What may help us out of this predicament, according to Kurzweil is the next technological revolution – Robotics – in particular, AI.
As I alluded to before, compute power is growing at an exponential pace (it actually appears to be double-exponential, but let’s not go there for now). Kurzweil’s books have all kinds of data, graphs and references to back this stuff up. It won’t be long before we have enough compute power to fully simulate a human brain. In order for this to be possible, we also need to understand how the brain works. There has been a lot of progress in this area as well, that Kurzweil spends considerable effort documenting. Here’s a really cool graph that shows how compute power is growing. It shows the amount of compute power that can be purchased for 1000 (normalized) dollars over the years and projected into the future:
Note that this is a logarithmic plot (see the increasing exponential scale on the left). If a trend is exponential, it will show up as a straight line in a logarithmic plot. Because it is curving up in this case, it means that the growth is actually double-exponential – which means the rate of change is actually increasing. Note that for $1000, you’ll be able to buy compute power equivalent to the power of the human brain around 2025 – and equivalent to the power of ALL human brains around 2055. For supercomputers, this capability will be achieved sooner. The following graph projects that we’ll be able to simulate a human brain by 2013. It also projects that we’ll be able to scan and download the exact details of a particular human’s brain in 2025 (in other words, a simulation that will be indistinguishable from the individual on which it was modeled). This assumes that we’ll also have the technology to scan and understand every necessary detail in a human brain (which will be aided by nanotechnology in this timeframe).
Once we create intelligence equivalent to a human, says Kurzweil, we will necessarily soar past it. This is because a computer that has human intelligence is actually far superior, as it will be able to think relatively much faster. It will be able to modify its own design, becoming increasingly more capable. Machines will also have the advantage of being able to easily share their knowledge with one another. These machines will plead their case that they are conscious – even human – and we will believe them. They will have the full range of human emotion, creativity, and appreciation for the arts.
Once artificial intelligence takes hold, our ability to solve the worlds problems with its help will seem to become unlimited. While machines become ever more capable – so will humans. Neural implants, and brain-enhancing nanotechnology will vastly improve our ability our brain power and memory. Those people choosing to remain fully biological (MOSH’es – “Mostly Organic Substrate Humans”) will have an impossible time keeping up with the changes.
Once again, with great power comes great responsibility and great danger. A malicious AI may not care what happens to humans. This is why we must “create them in our own image” and build in safe guards so that they will revere humans and their origins. I’m sure some of you are familiar with the Asmovian “Laws of Robotics”:
First Law: A robot may not injure a human being, or, through inaction, allow a human being to come to harm.
Second Law: A robot must obey orders given it by human beings, except where such orders would conflict with the First Law.
Third Law: A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
It may not go exactly like this – but you get the idea.
The Singularity
The emergence of strong AI (AI at or exceeding human intelligence) will be a primary driving force of the singularity. Things will evolve so fast that MOSHes will effectively not be able to participate in any meaningful way. Those choosing to leave technology behind, says Kurzweil, should not fear. Just as we have natural parks to preserve our environment and our heritage – so will the advanced societies of the future. (Still, one can not help feeling that the human race will be reduced to pets or zoo animals).
By the time the singularity happens, people and machines will spend most of their time in virtual realities. Machine intelligences will be able to manifest bodies in real reality – through nanotechnology. Biological humans – who will have most of their biology replaced by more permanent and reliable systems by then – will be able to change their bodies in the same way. This won’t really be necessary, though, as things will be so much easier and more flexible in virtual reality – and well, more “real.”
Human suffering will be a thing of the past. Naturally, there are a lot of moral dilemmas to face along the road to the singularity. We can discuss these later. For now, I’m just trying to communicate the mechanics here.
So when will the singularity happen? I think Kurzweil predicts around 2045. There are some more graphs to support that we are very near the singularity. Take the following:
Once again, this is a logarithmic plot. It marks events considered by various experts as paradigm shifts in evolution of the universe, life, and technology (Kurzweil explains this stuff in a lot greater depth in “The Singularity is Near.”). You can see that as we go from the creation of the universe 13 billion years ago to the present, the time between paradigm shifts is becoming exponentially smaller. What it’s showing is that as we approach the present exponentially, the time between paradigm shifts is exponentially approaching zero. Thus, it is expected that soon things will be changing faster than we can possibly comprehend today.
Post Singularity
As Kurzweil says, it’s difficult to see what lies beyond the singularity – just as it is difficult to see beyond the event horizon of the singularity known as a black hole. Still, he offers us some predictions.
First of all is the possibility that we will be able to download our minds and consciousness to a computer (far more capable than what we have today). By then, we’ll have quantum computing based on some of the fundamental properties of matter and energy. The computing substrate won’t look anything like what you’d call a computer today. By moving our human “patterns” from a biological to a non-biological substrate and ditching our bodies and inefficient brains, we’ll be able to fully participate in whatever happens next. The natural question is whether or not we’ll be ditching our humanity in the process. Kurzweil thinks not. I’m not sure myself. We’re free to discuss the possibility and what it means.
As things progress further, we’ll continue to find more and more efficient and fundamental ways to utilize matter and energy to perform computation. At this point, information and computation will be the only things in the real world that underlie our existence. The patterns inscribed on this base of computation will be what defines our consciousness and who we are. Kurzweil and others would argue that this is fundamentally who and what we are right now. That sounds like a very sterile and boring existence on the face of it – but it’s fundamentally not. There will be worlds of experience within this virtual reality that not only meet our own wildest dreams of existence, but far exceed them. And there will be nothing stopping us from interfacing with “real” reality in whatever way we want. (Note: This is the point where I start wondering about our current existence, and just how real it is. Are we just a simulation running on someone’s computer? Are we in a virtual reality inside of another reality? Maybe a reality inside of a reality inside of a reality?)
Kurzweil projects that by the end of the century, we will have fully saturated all the matter and energy in our immediate vicinity (i.e. solar system) with computation, and be spreading outward at a speed close to the speed of light.
Long Term
What happens next depends on whether or not it is possible to send information faster than the speed of light. If there are ways, our descendants will find and exploit them. Open up a small wormhole, for example, and send through some nanobots and some info and continue the process. If we can get around the speed of light, it will take 100’s of years to saturate the entire universe with “intelligent matter” and energy. If the speed of light is truly a limit, it will take billions of years.
Kurzweil believes that the universe currently does not have a consciousness – but it will once our intelligence spreads across the cosmos. Kind of a romantic thought (or empty, or disgusting, or blasphemous, depending on your particular point of view). This begs the question (and he addresses this in his book) as to whether we are the first intelligent life in the universe. If we’re not, wouldn’t the matter of our solar system have already been absorbed by some other spreading intelligence? Well here are some of the possibilities:
- We really are the first. This doesn’t seem likely, because that puts us in a privileged position. Most theories about life in the universe assume that we’re not “special” because that’s far more likely. We could fall back on the anthropomorphic principle though. It’s the same theory that’s applied to the universe in general. Why are the physical laws so exact as to lead to life in the universe? Well, if they weren’t, we wouldn’t be here discussing it. Same argument with us being the first intelligent life – If we weren’t, we wouldn’t be here discussing it.
- We can’t overcome the speed of light. Each intelligence is confined to a region of the universe. We just haven’t met up with any yet because of the distances involved.
- There are other intelligences in the universe, and they have fully saturated most of the matter and energy in the universe. They are just leaving us alone.
There is one other possibility to consider that Kurzweil did not think of. Perhaps intelligence has found a way to saturate some of the most fundamental properties of matter and energy with computation while leaving alone the more macro properties we are familiar with (chemical, nuclear, maybe even quantum). In this case, we could coexist and we wouldn’t even know it (well, they would – but not us). In this case, even the atoms that make up our very existence right now would be teeming with intelligence and consciousness. And we won’t recognize this until the day we evolve to the same level.
The End of the Universe?
Does it really matter what happens if some day the universe just expands forever and cools until there is no activity left, or just crunches into another big bang and all is lost? That is somewhat depressing. But there are other possibilities. Perhaps the intelligence of our progeny will find a way to create other, new universes and extend their life and intelligence to them. Thus, intelligence and consciousness will continue to evolve forever.
References
Recommended reading
Kurzweil, Ray – “The Age of Spiritual Machines – When Computers Exceed Human Intelligence” - 1999
Kurzweil, Ray – “The Singularity is Near: When Humans Transcend Biology” – 2005
Links
Kurzweil’s web page – KurzweilAI.net
My Related Articles
Trying To Understand the Big Picture
jbravo
5 Comments
To be blunt, I'm not sure I believe that human beings have the capability of behaving responsibly with the kind of power that these technologies represent. As you said, it's very hard for us to comprehend exponential growth; in fact, a fair number of well-intentioned technological "breakthroughs" have turned out very badly already, based just on our failure to recognise one of the important factors. Killer bees would be the first example that comes to mind here; a second would be the nuclear bomb (not simply that we made it, but that when the first bomb was detonated there was a possibility, which had not adequately been ruled out, that it would chain-react with the matter that comprised the atmosphere of the earth and kill us all). The occasional person pays attention to the long-term impact of a particular event or set of events, but on the whole, people don't spend very much time thinking about the consequences of their actions, and much less time taking responsibility for them. In fact, based on ad-hoc personal observations, another thing that grows exponentially is the frequency with which people say "it's not my fault, I didn't know" when confronted with the consequences of their actions.
This doesn't take into account malicious or pseudo-malicious intent. I say pseudo-malicious, because I don't think there are very many people whose intentions are truly malicious, just some combination of emotionally damaged, selfish and short-sighted.
Basically, I guess my thoughts are that our best chances for survival are that Kurzweil is wrong, and the new developments level off, or that some smaller catastrophe occurs which sets us off of the exponential track before we go beyond the line where a big catastrophe occurs (small = doesn't render the earth entirely unfit for habitation).
On a more practical note, I have a couple things to add about computing technology. Quantum computing has run up against a wall of non-scalability; aside from that, the notion of harnessing the quantum superposition principle for computation is based largely in the existence of the nondeterministic model for analysis of algorithms--basically, it's a US military pipe dream because it would allow virtually instantaneous brute-force decryption. The only proof-of-concept we've come up with can factor the number 15, but isn't scalable (= you can't make a bigger one that will factor 16).
Conventional processors have hit a wall as well. Computational power hasn't been doubling every 18 months for the last 3 or 4 years; about 4 years ago the news I heard was that they halving the feature size on the processors would only yield about a 10% performance improvement. Right now speedup is through parallellism--dual core, quad core, multiprocessor mobos, and so on--and through memory-processor speed matching (bringing bus, memory, etc. speeds up to be closer to the processor's speed, reducing wasted processor cycles), but the actual micronisation of processors has slowed to almost stop. This is not necessarily a permanent barrier, but we don't have a clear philosophical reason to suppose that we can get feature size anywhere near the actual molecular line or below it--in fact, we have several clear philosophical reasons to suppose that we can't. Machinery depends on mechanical forces, and mechanical (Newtonian, or even Einsteinian) forces become increasingly nonexistant as things get smaller--you just don't have the sample size with which to receive predictable mechanical results.
Another issue I have with the technology has to do with artificial intelligence. I do personally believe that we will probably produce an 'intelligent machine', whatever that means (noone has every offered me an adequate definition of 'intelligence' against which to compare people, computers or left rubber boots). I think I believe that 'sentience' is actually a sliding scale, on which everything features at some level (left rubber boots pretty far down, computers somewhat above that, most people somewhat above that); I don't think I have a really reliable reason to assume that we humans exhibit any special 'sentient' quality which differs fundamentally from qualities exhibited by animals, insects, or computers (left rubber boots I leave as an exercise to the reader). So I think that some kind of 'intelligence' already exists in computers, and that this will continue to increase; but I think it's a product of information association and re-association, not a product of processing power. A faster processor just makes the same program run faster. To make a smarter computer, you need smarter software.
On the other extreme, though, the human brain has defied attempts by psychologists to map it. There's an evident connection between certain base-level responses and certain glands or cranial features, but the actual 'mind' doesn't seem to be cleanly embedded in the parts. Theories of the holographic nature of the mind have arisen, and I'm quite intrigued by this idea; but I'm not sure how much actual information is contained within the term 'holographic'.
Anyway, I'll stop there because my thoughts get a lot more hazy as we get into thoughts of 'the meat versus the mind'. But thanks for putting this out there, because it's real cool like. :)
Kurzweil definitely tends to be on the optimistic side when it comes to these predictions. What he doesn't take into account are other forces that might be growing exponentially in the opposite direction of progress that could come to dominate the situation. He does counter various forms of criticism in his book -- but I'm not sure they're all dealt with realistically. The possibility of catastrophy is certainly real, but he tends not to dwell upon it. I hope it is possible to stay ahead of the curve and prevent something really bad from happening -- but like you said, the ability to do that may be decreasing (and probably exponentially at that). This is the reason I am so adamant about getting humans on other worlds. We need to have a backup plan if we really screw up.
"On a more practical note, I have a couple things to add about computing technology. Quantum computing has run up against a wall of non-scalability"
I hadn't heard that one yet. But then, my understanding of quantum computing is pretty shallow. I am always skeptical, though, when someone says somthing is impossible. More often than not, it really means "not possible now." It seems we always find creative ways to get around what seems like insurmountable barriers.
"Conventional processors have hit a wall as well"
Well, not quite. I am far more familiar with this topic, since I work for a major chip manufacturer. :-) It's true that we are finding other ways to keep making progress -- as you said, multicore, etc. We've definitely taken the silicon based IC process much further than people would have expected in the first place. It's still got a little life left. But you're correct, it will eventually trail off at the top of the "S-curve". But just as IC's replaced discreet transistors, replaced tubes, replaced relays, etc, I'm confident that we'll find a new technology to pick up and run with. Vertical fets are probably a good first contender, where we can pack more of them on a chip. Seems that molecular or carbon nanotube-based computing are some good longer-term possibilities. Also, going to 3 dimensions of fet placement might be a possibility. Some form of optical computing might take us quite far.
"(noone has every offered me an adequate definition of 'intelligence' against which to compare people, computers or left rubber boots). "
What do you think about the "Turing Test" as at least a relative measure of intelligence vs a human?
"A faster processor just makes the same program run faster. To make a smarter computer, you need smarter software."
Absolutely. But I think we are making progress there, as well. Chaotic computing techniques -- neural nets, genetic algorithms, etc, are making a comeback.
"There's an evident connection between certain base-level responses and certain glands or cranial features, but the actual 'mind' doesn't seem to be cleanly embedded in the parts."
Agree there. A lot of the properties of the mind are emergent. You can't understand the whole by looking at the individual parts. I'm certainly not an expert here, but Kurzweil spends a lot of time on this issue. You should read his book and then see what you think. He says that there are a lot of aspects of the brain that are reasonably well understood. Things like certain patterns of organization repeated over and over. We don't necessarily need to know every detail in order to simulate a brain that acts just like a human. We do need to be able to scan every relevant detail in order to exactly replicate a particular individual's brain. In any case, I think you're right that at this point, there's a lot more we don't understand about the brain than we do understand.
Don't have time to post, so I say, good topic, good night.
;-)
BTW, wow. Someone's thinking around here. Deep thoughts.
Hey you know AdGuy always gets the last word! ;)