Future Current

John Smart and Eliezer Yudkowsky at the 2006 Singularity Summit at Stanford

John Smart is a developmental systems theorist who studies accelerating change, computational autonomy, and the singularity. He is President of the Acceleration Studies Foundation, a nonprofit community for research, education, consulting, and selected advocacy of communities and technologies of accelerating change. He also co-produces the Accelerating Change Conference, a meeting of 350 change-leaders and students at Stanford University, and edits ASF’s free newsletter, Accelerating Times, read by future-oriented thinkers around the world. He is a member of the Association of Professional Futurists, the FBI Futures Working Group, and the editorial advisory board of Technological Forecasting and Social Change.

In 2006, he presented the talk “Systems Theories of Accelerating Change” at the Singularity Summit at Stanford. There he looked at accelerating change from universal, biological, human cultural, and technological perspectives, and introduced a few well known and unorthodox ideas in acceleration mechanics.

The following transcript of John Smart’s 2006 Singularity Summit talk “Systems Theories of Accelerating Change” has been edited for clarity with footnotes by the speaker. An audio version is available at the Singularity Summit website.

Systems Theories of Accelerating Change

I want to thank everybody for coming to this event today. As Ray says, coming together and taking the issue seriously helps get past the silliness and we advance the dialog. So, that’s what this conference is all about. If you want to go into these slides in more detail, they are available at accelerating.org/slides, along with my email if you would like to discuss this further.[1]

I’m going to give you a story in three acts with three actors. We are going to talk about the framework, which are the rules in our universe. We are going to talk about the picture, which is the reality we are creating. And we are going to talk about the painter, the actor: you and I. An analogy from astrophysics would be laws and constraints (the framework), the universe (picture), and the intelligence (painter, actor) that emerges within the universe.

So, act one is the framework. What is the framework of accelerating change? I am going to discuss three things in brief: intelligence, then something I call MEST compression–matter, energy, space and time compression–and then a paradigm that I think is very important for understanding, called evolutionary development. Only a small number of people are working in this latter area of evolutionary development right now, but I think it is really important to the future.

So, intelligence first. In many ways intelligence is the driver of accelerating change. Not driver in terms of the energy, but driver in terms of the steering mechanism. The word “cybernetics” comes from steering, and that’s really what intelligence is doing.I think the best place to start, if you want to get a sense of how the universe is involved in creating information is Seth Loyd’s book Programming the Universe. It’s a quantum physicist’s perspective on how even at the quantum level, computations are going on everywhere. The whole space-time continuum is in the process of creating information, and we on this planet, in this particular location, appear to be on the leading edge of not an anthropomorphic acceleration but an infopomorphic acceleration.

We are not the center of the system, but information might be. It’s also a biofelicitous system. The book Biocosm by James Gardner is a great way of understanding this concept called the Anthropic Principle, which physicists love to hate but they can’t get rid of. It is a really interesting idea about how life-friendly and intelligence-friendly our universe appears to be. And finally The Accelerating Universe by Livio. Every complex adaptive system goes through cycles, and one phase of the cycles is senescence and death. This book and the concept of the fifth force, the lambda energy, is early evidence that even our universe is about a billion years into an accelerating disintegration. Where is intelligence going to go? Possibly somewhere we are not looking right now. Not outer space but inner space. That’s my theory and I’m sticking to it.

To start with, to think about what these issues are, Wigner in 1960 talked about the unreasonable effectiveness of mathematics. If you apply a little bit of intelligence with these unreasonably simple equations, rules emerge that help us make intelligence a strongly non-zero sum game. Carver Mead made another wonderful observation. He said the microcosm is also unreasonably efficient. Whenever we play around in the physics of small scales, amazing efficiencies emerge. We don’t understand why. These efficiencies are not so much due to human creativity as they are due to human discovery. And that’s a really interesting thing when you consider the dynamic in human intelligence between creativity and discovery.

A guy who saw this even earlier was Teilhard de Chardin. In The Phenomenon of Man he said the universe was starting with the Chemosphere, and then a Biosphere, and is moving now to the Noosphere. This memetic technologic substrate that is happening now on our planet. He said that technology is in the process of cephalizing the earth, giving it a head. And a really simple way to understand this phase transition, he said we have a planet with finite sphericity and accelerating linkage density amongst all the actors. So, what’s actually happening, you put those two together, the constraint and the acceleration and you get a phase transition. Eventually you get a king of ’skin’ on the planet that is so dense that a new set of dynamics emerge. Kind of like moving from a gas to a liquid.

So, what is acceleration studies? It’s in its very early stages right now. Here are three good books to start with to really get a sense of the history. Don’t look for any understanding of it in your physics or information theory textbooks in college. You won’t get it right now because we have not yet at the academic level given ourselves permission to even study it seriously. Ray Kurzweil pioneered bringing out the data and helping us understand how powerful it is, this fifty years of Moore’s Law. The data tell us just how predictable is this shrinking of transistor density, which gives us increasing speed and power. I think it is actually more a result of human discovery than human creativity. Software, which is on a six-year doubling, is maybe a lot more human creativity. There are still fundamental algorithms being discovered but in a lot of cases we’re still creating, and we’re not really sure what the optimization is. The universe isn’t giving the optimum algorithms to us.

A guy named Richard Dickerson showed that solved protein structures are a Moore’s-dependent process. So there are all kinds of interesting secondary computations that are just as exponential once you have this fundament of accelerating computation in the hardware. As Ray noted in 1999, generalized technology capacity, and in particular, information and the communication sciences, are the only things independent of socio-economic cycles. And I think that is because in the Great Depression a few curious people playing around with the physics of small scales found these amazing efficiencies to emerge. You didn’t need a lot of resources or human creativity. You just needed people looking in the right place.

Now, ninety years earlier than Ray’s work, I think the first singularity theorist that I have been able to discover was Henry Adams, one of our first historians of technology. And he said there was a rule of phase applied to history. In 1909 he published an essay which said that sometime between 1921 and 2025 we are going to see a singularity, a time where change from our perspective looks almost instantaneous. I think that is a really interesting bit of history. There are two other books here from 1998 and 2000, The Evolutionary Trajectory by Richard L. Coren and Les Arbres De L’Evolution by Pierre Grou. They predict a singularity in 2080 or 2130, respectively. Two more recent books, including Ray’s The Singularity is Near from 2005, predict a much closer singularity with even narrower confidence intervals. Of course, this is all a guessing game at this stage. But the more people that get into this, the better the wisdom of crowds gets.

So, MEST compression (Matter-, Energy-, Space-, and Time-compression, our next topic in the ‘framework’ of accelerating change) is the engine, the “motive force.” MEST compression appears not to be evolutionary but developmental. It seems implicit in the physics of the special universe we inhabit. This intelligence and life-friendly universe. We have these four things–matter, energy, space and time–which are increasingly understood by physical science and they are producing information, a fifth phenomenon which is still very poorly understood. As Tim Ferris would say, there is no Einstein of information theory yet. What we do know is that every new substrate that emerges in the universe in evolutionary developmental history can use MEST resources to do any standard physical process or computation with astoundingly fewer resources for every new substrate.

So, MEST compression is about continuing to do more (computationally speaking), better, with less physical resources. And that seems to be the fundamental driver that is pulling us into these faster domains. Our machines are becoming dramatically more MEST efficient with each new generation. As Ray has noted, our main candidate for future computational technology: nano, molecular, quantum, even reversible logic, require little or no energy. We are all moving to increasingly energy-efficient, sustainable and virtual cities. The weight of our GDP per capita goes down in all developed service economies over time. Global energy intensity, the energy consumption per person, has been flat for almost three decades in the developed world. We do not use more energy; we do not need to for multiple interesting reasons. So, human society itself is apparently within a couple generations, 50 years, of becoming resource independent, moving beyond scarcity economics, in regards to energy.

What are the physics of a MESTI universe? How do Matter, Energy, Space and Time conspire to generate Information? No one really knows. We know that this compression is driving acceleration, and there are emergent properties of information. The leading edge of Universal information is more intelligent, independent, and immune, and yet persistently incomplete over time.

It was not always like this. Originally, we had the Hot Big Bang. MEST processes were going very fast. Basically, in the first half of the universe’s apparent developmental life cycle processes were going slower and slower, until we had the emergence of the first galaxies. And then, as Sagan noted in Dragons of Eden, in the second half, we have this astounding acceleration. The closer we get to the present time, the faster and faster things go.

So, basically what we have here is a U-shaped curve of change. It looks like a developmental cycle. No one really understands why it exists. No one even understands whether it is an artifact or not. I don’t think it is. Cycles suggest recurrence. Probably the best person physically who has attacked this problem is Eric Chaisson in his book Cosmic Evolution from 2001. He says that there is a free energy rate density curve that looks like a Moore’s Law curve writ large on the universe. And he notes that suns, like our star, have actually four orders of magnitude less free energy rate density in them than our human bodies do. That’s because suns have a lot of energy, but they are huge, fluffy, non-dense systems. Our Pentium chips have six or seven orders of magnitude more free energy rate density in them than human brains do. That’s the way it seems to work. We don’t understand why, but that’s how it works.

I have some laws of technology. These are tongue-in-cheek. They are just rules of thumb. But I would say the first thing to realize from Chaisson’s model here is that technology learns 10 million times faster than you do, and it always will.[2] If you recognize that, it is incumbent upon us to make technology an extension of our humanity. Because it is going to continue to learn and pick up these cognitive, modular singularities 10 million times faster than we do. This calculator watch I’m wearing does calculations; it is a calculational singularity for me. I have no idea what is going on on the other side of it, underneath the interface. But human culture has crafted the structure of this watch so that it really serves my ends. And more and more of those calculational singularities are going to get picked up modularly as we move toward this thing we call general AI.

Another thing to realize is that humans are selective catalysts, not controllers of our technological evolutionary development. And, the first generation of any technology is often dehumanizing. So it is incumbent upon us as designers to try and get beyond that first generation hurdle.

There are five classic strategies to discuss for technological development: Nanotech, Infotech, Biotech, Cognotech, and Sociotech (NIBCS). Only two of those are lead drivers in in this acceleration phase: nanotechnology and information technology. As Archimedes would say, they are the levers that are moving the world. And we need to realize how special they are. As one example, some researchers were playing around with a structure called a photonic crystal in 2004 and they figured out that they could make micro-lasers a million times more energy-efficiently than we had made them in 2003. That is the jaw-dropping efficiency advance that happens in the nanospace. We don’t understand why this happens. We still do not put the kind of resources and funding into the nanospace or the infotech space the way we do other spaces listed above. When we do, these amazing advances occur.

A more pedantic example, Toshiba introduced the lithium ion nanobattery last year. They announced a nanobattery that recharges 80% in sixty seconds. That is almost 80 times faster, nearly two orders of magnitude, than the lithium ion batteries we have in our cell phones today. They did this simply by playing around with the physics at small scales using nanoparticles at the anode and cathode. You might not think that is so disruptive, but put it together with PHEVs (plug-in hybrid vehicles) and you have a car, when hooked up to a high-amp charger at a filling station, that could fill its nanobattery gas tank faster than it can fill its gasoline gas tank.[3] That gives you an oil-independence transition for the next twenty years.[4]

With MEST compression, the thing to understand is that technology is now the fastest system of change in terms of hierarchical systems of social change. It used to be all about the power, then it was all about the money, and now it is all about the technology. These layers have reorganized to fastest first and we need to understand just how powerful the technological hierarchical system as a tool for social change has become. And we need to understand the power of technological benevolence, as Buckminster Fuller would say, using an incredibly compelling technology, like a cell phone, and pushing it out to everybody for free.

Probably the first person to really see MEST compression clearly was Buckminster Fuller and the MEST compression he called “ephemeralization.” He said that we are moving away from needing physical resources. All progressions in his developmental scheme are from more material, to more abstract. They eventually hit what he called the electrical stage, sending electrical bits to do physical work. This trend has also been called virtualization, weightlessness, or as I call it: MEST compression, MEST efficiency, or MEST density.

The last thing that we should see, with regard to MEST-compression is that every new substrate uses vastly less matter, energy, space and time to do its encoding. Everything in outer space is slow space compared to what is in inner space. Outer space is computationally slow, very simple. It is rapidly encapsulated by our simulation space–inner space. I put it to you that the brain is basically a massively efficient simulator by comparison to previous ‘computational substrates,’ modeling the birth and death of the universe on a few watts. Now, I look at structures like the international space station, which is a non-autonomous system and I don’t think it is anywhere near as interesting as the brain. And I think it is only when our technologies become as organic and as completely simulating as natural, as fault-tolerant, and biologically inspired in structure and function as this system (the human brain) will we finally see a technology that becomes what we would consider a natural extension of ourselves.

The handle on this, the third element of the ‘framework’ for accelerating change, that I think everyone needs to get, is something called evolutionary development. I think it is a new paradigm for understanding Universal change. Evolution is random, chaotic, unpredictable, expansionary. Development is the opposite. It is self-ordering, directional, predictable, and convergent. It involves the calculation of optima.

So, what happened in the Cambrian explosion? We had a rapid evolutionary expansion of differential multicellularity,[5] and then within the next 10 million years we had 35 body plans that were computed. There have been no new body plans for the last half-billion years.[6]

The development of brains undergoes a very similar process called ‘neural Darwinism.’ See Gerald Edelman’s, Neural Darwinism, for one model. Like the emergence of body plans in the Cambrian, brain development is also exactly a ‘left-hand’ evolutionary expansion, ‘right hand’ developmental pruning process.[7] And so is the way evolutionary, experimental unconscious thoughts compete with and converge to the thoughts you think on the conscious level. (See William Calvin’s, The Cerebral Code, for one such model.)

So evo-devo is extremely helpful as a tool for understanding change, both on the biological level and in the universe at large. If you want to see the difference between them, between the processes of evolution and development, look at two genetically identical twins. At the microstructure level, everything is molecular chaotic selectionism and unpredictable. And at the macro-level, there are amazing similarities between those systems, between the two twins. Both evolution (more specifically, chaotic molecular adaptation) and development are working in that system. One drives unique local and one drives predictable global patterns.

Marbles, landscapes, and channels is another way of looking at the interaction of evolution and development in the environment. The marbles roll around on a landscape (which is the system), they take unpredictable paths, but together they converge on these low energy channels, and that is the MEST compressed computational attractors at the bottom of the basin.

This model is being used right now to attack biogenesis. Eric Smith at Santa Fe says look at potential pre-biotic chemical reactions. They form a vast and evolutionary possibility space and energy landscape. A subset of these, what we might call an eary developmental subset, makes self-reproducing, self-varying chemical cycles. Those produce information, which permanently modifies the future selection environment: so called ‘niche’ construction. (See John Odling-Smee’s Niche Construction for a good book on the latter topic). That now modified selection environment has a series of low-energy paths (or seen another way, ‘high-information’ developmental paths) constraining the landscape. Now, can we one day soon develop a model robust enough to understand how biogenesis is going to be an inevitable developmental emergence from molecular chaotic determinism? I think we might. And that does not take away the incredible generativity of evolution itself as a system. It is just the other half of the picture that is not being done.

Evolution does not understand how many eyes you need. It throws as many as it can at the wall, like this three-eyed skink with an apparent vestigial third eye. But development pares it back to the optimum two eyes for stereoscopic vision. Same with four-wheel automobiles.[8] Probably the most aggressive example of this developmentalist way of thinking is Russell’s Dinosauroid hypothesis. If you remember those raptors in Jurassic Park, they had opposable thumbs, forward-facing eyes, and were tending massively towards the “anthropoid form.” And that is where they would have gone, if there had not been a K-T meteor in this model. Will we be able to simulate this hypothesis that the anthropoid form is a developmental optimum in future simulation science? I think so.

So evolution and development are two fundamental ways of looking at change in the universe. The deepest question is when, where, and how they interrelate. And that is what we need to get to. And I think for political dialogs, innovation (an evolutionary, creative perspective) and sustainability (a developmental, conservative perspective) are two really important left and right hands of change. If you only have one in your rhetoric, models, and plans, you are missing half the picture. Because Evo Devo needs both.

Development also follows trajectories. Angus Maddison, showed phases of capitalist development that were moving toward a networked society. And now we are moving beyond the network to what several people call a metaverse or an “attention economy.” This is an ICT-enabled services sector.

The picture is acceleration mechanics, immune systems, intelligence amplification, and AI.[9] You need to understand things like the S-curves and how they saturate. Things like the B-curve, seen in all biological systems, which moves through and past the S-curve toward decline and recycling. All living systems do this. I would like for you to think about the difference in the S and the B-curve (logistic growth and recycling), between the germline cells in us that are immortal and the somatic cells in us that are continually recycling. They follow two different sets of dynamics.

Developmental systems highly protect the immortality of their parameters, but they have evolutionary extendability in the expansion of those parameters into the computational net. If you understand that difference, then I think you understand deeply why the only kind of immortality we might see in the coming decades, and here I differ with some of our speakers, is what we might call a cybernetic immortality. At the biostructural level we are built to fall apart at an accelerating rate after sexual maturity. It is encoded in our fundamental structure and it is actually valuable to us for that falling apart to occur in a competitive evolutionary environment. We are in a process of transitioning to a system that is, in many ways, biologically independent. Nanotechnology may get us there, but I think information technology is getting us there even faster.

The last curve I am going to leave you with is the J-curve. Here is where we have second order hyperbolic growth. It is made of first-order components, but there are emergent singularities all the way along. And at each level, some aspects of the post-emergent, post-limit systems cannot be understood or guided by pre-singularity systems. That is kind of the state that we are in today. We see this threshold coming up and have tremendous opportunities for the future, and I think the important thing is to develop the awareness and have the courage to advance the dialog. Thank you.

Footnotes:

1. See also accelerationwatch.com for more on this topic.

2. Smart’s ‘first law’ of tech is that is that technology learns 10 million times faster than you do, and it always will. The ’second law’ is that humans are selective catalysts, not controllers of our technological evolutionary development. The ‘third law’ is that the first generation of any technology is often dehumanizing. These ‘laws’ are more rules of thumb, but are useful in understanding the notion of accelerating change.

3. Ed Note: Or nearly as fast: 10 minutes.

4. Natural gas and alternative electric generation for transportation might become steadily more attractive than oil, in this future scenario.

5. Perhaps driven by the first emergence of eyes.

6. The implication here is that a number of developmental optima were computed, which henceforth prevented further evolutionary experiments with body plans as further evolution became ‘path-dependent’ on these optima.

7. John Smart holds up his left hand, fingers spread, read left to right, as a symbol for evolutionary diversification. He doees the same with his right hand, also read left to right, as a symbol for developmental convergence (of evolutionary paths).

8. Hypothesis: Three or less, and five or more-wheeled automobiles will be ‘developmentally’ likely to be minor forms, wherever the automobile first emerges, independent of culture or planet.

9. The second element of Smart’s (framework, picture, painter) model of change. See slides for detail.

This entry was posted on Saturday, February 16th, 2008 at 4:09 pm and is filed under Metaverse, Singularity Summit, John Smart. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.