The Future is Not Accelerating Wednesday, Sep 9 2009
futurism 3:17 am
I’m just writing this post to let my readers know that I no longer think that technological change is reliably accelerating. It seems to me that most major technological developments of the last century occurred mid-century, for instance. I agree with the recent BusinessWeek article that claims that innovation has slowed in the 00s. For what it’s worth, I stopped believing in reliably accelerating technological change before the recession started, I just didn’t bother to write about it here. I do think that the law of accelerating returns has some merit, but it’s less than implied, and it seems to me that society undergoes periods of both linear and lightly accelerating technological change. One deceptive confusion has to do with the difference between cutting-edge technological breakthroughs which are lab curiosities for the time being and breakthroughs which are actually commercialized and adopted.
There are extremely strong social and economic pressures to engage in research that looks trendy and produces a bare minimum return. I also doubt that most people at knowledge jobs are actually working at any given time. (There is another random blog post that I wanted to link on that, which I saw on Michael Silverton’s Twitter, but I can’t find it. See also Andy Rooney’s recent observation that everyone is carrying books to work and acting like they aren’t reading them.) This is radically different than in the past, when people had to work or it was obvious they weren’t. This has to be hurting our total productivity. To use Half Sigma’s terminology, I also think that people are making a tremendous amount of money on wealth transfer rather than wealth creation, when they are actually working.
It depends what qualifies as a major technological development. Kurzweil’s graph of number of US patents filed – if the figures are accurate – would tend to suggest that either people are becoming more inventive, or (more likely) they are increasingly gaining access to the means required to publish their ideas. At least in north America and Europe these changes can’t be explained purely in terms of population increase – i.e. an exponential increase in the number of inventors. One possible explanation might be that the average business size in north America might have been decreasing exponentially over the last century, but filing patents at the same rate, although I don’t know whether that’s really the case.
The “low hanging fruit” criticism has always applied, whether the research is being done in academia or industry. The main reason for the tediously slow progress in AI development, apart from computational restrictions, is that in most cases it’s not low hanging fruit. Also, in the rush to chase the latest trend older research can sometimes be forgotten, and there’s a long history in science of older work being rediscovered by new generations.
On individual productivity in the workplace, or lack thereof, this is just a function the strange way in which modern businesses operate. The way that most businesses are set up assumes that people’s output over time is constant. Most people work a fixed number of hours per week. In industrial production the constant productivity rate assumption usually holds true, but for “knowledge economy” jobs which involve creative thinking, like writing or software engineering, this assumption is often false.
Wealth creation I think only constitutes a minority of economic activity. The majority of the activity going on is, as you say, transferal of wealth from one group or geographical region to another. A well known institution whose primary purpose is wealth transferal is known as “the stock market”.
I believe that the dot for 2008 in the 6th slide in this presentation is right on track: http://www.slideshare.net/kkleiner/kain020109-tedu09-1
I also believe that we have seen numerous major techonological achievments in recent time like the invention of home computers, mobile phones, the Web and stem cell therapy.
On individual productivity – it just shows that thanks to modern tools people can do the same amount of work in less time. It’s not a concern for global productivity as the chart on the 22nd slide in the same presentation shows.
It’s a pretty gutsy move to disown your blog title!
I’m with you this far: that people by their choices have dragged the trajectory of technological improvement far below where it could have been (based from whatever date you choose — 1960 might be a good one). We consume diversions and goods that convey alpha-status and we gamble away wealth in casinos and in purely speculative investments.
At the risk of sounding tiresomely single-issue, to take best advantage of our technologies — especially as they become more powerful — we need to improve ourselves. At the very least, even barring alterations to our brains, we need to behave like people with something profound at stake. When we can do that, science will take its rightful place in our priorities (as will human rights, education, good governance and responsible stewardship).
Until then, we will have plenty of reality TV, impenetrable financial products and short-sighted parochial selfishness to obscure the gap between ourselves and our potential.
By the way, if people are listless and even passive-aggressive at work, it’s only partly from being lazy. Mostly it’s because they can’t believe in what they do, and that in turn is partly due to the lack of real satisfaction in sitting still and typing non-creatively all day.
I’d like to clarify your views vis-à-vis Kurzweil, Smart, and other. Are you rejecting merely double exponential progress, or the whole exponential thing entirely?
Interesting acknowledgment. Couple of considerations come to mind:
(1) are certain technological segments currently undergoing rapid progress, while others slow? would be interesting to note which fields fit in to which category
(2) could technological development be like the punctuated equilibrium phenomenon in evolution? ie technological progress can be measured in fits and starts
Woot, I get your domain name!
I’m disappointed in how many “cured in mice” stories I’ve read, knowing it will be years before any of those advancements make it to people.
It’s probably not accurate, but somewhere I got the impression that a major slice of the new patents were for things that shouldn’t be patented, like software, and many were defensive in nature, just to fend off patent trolls.
But finally… we’re in the worst slowdown in years and years. You can’t take a measurement now and call it average.
But super finally… I read lots of blogs from smart people. But it’s just talk. Who among us is doing something? As much as I like computer security, it’s not what I imagined I’d be doing when I was working on my degree, and at the end of the day my net addition to the world is pretty small.
It seems to me that much of the disagreement over whether technology is accelerating is really just disagreement over whether or not to use a log scale.
Yes, many patents filed may have nothing to do with inventiveness so there’s always a certain amount of noise. Bad patents aren’t really a new thing though.
On the hot air quotient, there’s always a lot of it – especially amongst artificial intelligence proponents. On a typical AI related forum I’d estimate that the percentage of hand wavers, dreamers, sci-fi survivors, clueless doomsters and armchair generals to people actually doing some non-trivial AI work is extraordinarily high. But you’ll find something like the Pareto principle in all areas of endeavor. It’s just particularly acute in AI because the problems are difficult to make progress on.
I agree with some other commenters that this is wisdom on Michael’s part (if I may talk about him in the third person). He’s putting a bit of critical distance between himself and something that is really futurological pseudoscience catering to the desire for ever-increasing intensity of experience. You can’t really hope to approach an event like the Singularity in a responsible fashion while also getting off on future shock.
But having rejected the notion that the shape of the future is adequately depicted by an exponential or even hyperbolic curve, that still leaves open the question of how to conceive of it correctly. I would in fact still defend the notion that the advent of the Internet represents a significant qualitative change in the world, and probably a penultimate stage before the actual Singularity. Also, that one should not be fooled by the number of tech developments which seem to have life only as lab stunts and Youtube videos, rather than as mass-produced pieces of everyday life… one should not be fooled by this into thinking that progress has stopped or slowed down. We’re in a different cultural configuration, that’s all. The past also had many inventions and technically ingenious constructs which went nowhere commercially.
It’s a futurist cliche, but it’s still helpful to think of the world in terms of “preindustrial”, industrial, and an information age. I put preindustrial in quotes because you can think about in various ways – agricultural, feudal, mercantile. I take the defining development of industrialism to be the factory, and the mode of transformation of daily life characteristic of industrialism as arising from the mass production in factories of some new class of artefact. But once you get personal computers, and especially once you get networked computing devices, you’re in a new phase. What distinguishes a computer from all preceding artefacts (except perhaps a book) is the very large number of states it can assume. Our computers are not yet people, but they begin to approach the complexity of people. And if the technical culture sometimes seems a little exhausted, a little deracinated, a little flighty and unfocused compared to previous generations of invention and accomplishment, it’s because we’ve never had toys like this before. They are turning the human mind inside out, cognition itself becoming situated in a network of autonomous artefacts rather than just in a human brain and body. It’s the biggest leap any natural intelligence can hope to perform and it doesn’t happen overnight. So while we endure all the epiphenomena of this enormous transformation, don’t mistake the absence of milestones in the image of yesterday’s breakthroughs for a general lassitude and lack of anything happening.
Hi Michael, you may have been referencing “Modern Work is Mostly a Lie” http://michael.silverton.palo-alto.ca.us/sweetcron/items/view/2711 or http://tr.im/modernwork
However, I don’t know if I agree that this means the future is not presently accelerating. I do very much agree with your observation — and I think Ray would too — that the rate of acceleration varies in the short term. I have a personal, mostly unsubstantiated bias toward believing that A(f) > 0 holds true — i.e., Acceleration of Future always greater than zero; a presumption sloppily derived from evolutionary pressure which I presume as a variable constant (always present and greater than zero). There may be times when either or both get very close to zero, indeed.
Perhaps Clay Shirkey’s Cognitive Surplus model also helps to explain the ebb and flow of acceleration. Maybe it’s sociologically evolutionary for groups of humans to reach tribal plateaus where more and more members keep banking this surplus until someone or some subset figures out, “oh, *this* is what could be doing next” and convinces the others. An interesting tangent here is: Why do some humans use their cognitive surplus to primarily keep consuming larger doses of repetitive romance novels and pr0n sites, while others attempt to employ that surplus to learn a new subject, another language, or experiment and explore other new possibilities at the periphery of their previous experience?
Of course, there are many additional fantastic comments here, as always, e.g., Thuris, “people by their choices have dragged the trajectory of technological improvement far below where it could have been.” I suspect that this observation is partly what has gradually drawn me into this current role of #MadScientistClown, as some kind of sociotechnological Circus Barker for proximate social changes like Basic Income and Universal Health Care that might conceivably better structure human society to engage it’s fallow collective Cognitive Surplus and re-accelerate progress to some extent. But like a circus barker, WTF do I know? Often, I can only point to the next interesting diversion that I discovered while out tramping the dusty roads and byways.
One of my recent side-shows has been exploring reasons why Basic Income could be a powerfully accelerating influence. A full-on ramble exceeds time and space here, however, and Jamais Cascio’s recent work in Fast Company http://tr.im/fastcobasic1 http://tr.im/fastcobasic2 is surely pre-requisite to anything useful I might add.
Again, drawing on that Thuris quotation for a tangible example, it seems to me that the still-too-centralized social cognition process, by which Single Payer health care got removed from the debate entirely — and by which even the likes of NPR has now constructed the Public Option as “far left” — illustrates to me the frighteningly stagnating power of the 20th century centralized global cognition grid (mass media messaging, controlled by a few) vs. the disruptively democratizing potential of the nascent, distribute global cognition grid that we’re all in the process of defining, designing, and implementing in various ways.
Perhaps such events could be interpreted as evidence that we are indeed near an inflection point, or just past one; remember the overlapping “S” curves … there are indeed relatively flat segments of those curves … and they don’t overlap neatly or cleanly like in the PowerPoint presentations. Sometimes, we have to remember that our own symbols are just that, symbols to represent things that we know we can’t actually measure directly, just yet.
So all this free association gets me thinking that I want to ask Eliezer, “why does hope spring eternal” considering that it is often so clearly irrational within the context of our everyday experience? For millions of humans, nay billions, quotidian observation should lead to Accelerating Disillusionment and dismay, shouldn’t it? I mean, in so many cases, the current situation just keeps sucking; even, or especially, where it seems that there is less than zero excuse for it to keep sucking. The dumb choice, the backward choice, the fear-coddling choice, the redundant choice, the flat stupid choice wins the day so often, why do some humans keep trying, anyway? Should optimism be clinically codified in the upcoming DSM-V?
One accelerating trend which can be refuted is microprocessor clock speed. In chapter 2 of The Singularity is Near, Kurzweil shows clock speeds continuing to rise exponentially up until at least 2015. Probably he was writing the book in the early 2000s, and asked chip manufacturers what they thought would be the clock rates of in development CPUs to be marketed over the next decade – hence there are data points well beyond the publication data of the book.
Other metrics from the same chapter seem harder to dismiss. There’s still room for the number of internet hosts to continue expanding, since many areas of the world still have little or no internet access and there’s ample scope for a growing “internet of things”, but human population will mean that a saturation point will be reached in the foreseeable future and the exponential growth will end.
While clock speed may not have risen so dramatically since the early 2000s, instructions per second have. See the Wiki entry for a chart:
http://en.wikipedia.org/wiki/Million_instructions_per_second#Million_instructions_per_second
In similar fashion, see the cost per GFLOPS chart:
http://en.wikipedia.org/wiki/FLOPS
The meaningful measures of computing power continue to increase exponentially.
Basically on-topic… if you have not yet seen this beautiful animated plot of GDP against life expectancy, please set both to logarithmic, scroll it back to the year 1800 and click play:
http://graphs.gapminder.org/world/
George Dvorak’s comment seems apropos: “are certain technological segments currently undergoing rapid progress, while others slow? would be interesting to note which fields fit in to which category”
I don’t think all sectors of technology and science are accelerating exponentially, and I don’t think Kurzweil claimed that.
Genomics, brain imaging, brain science, computing hardware, communications infrastructure … all these and more are clearly accelerating exponentially. IMO that is all we need for Singularity.
Specifically, all we need for Singularity is dramatically superhuman intelligence … and it seems clear that, as Kurzweil points out, the acceleration of brain imaging and brain science combined with the acceleration of computer hardware will get us there — if AGI innovations don’t get us there first.
Yes, the technology of refrigerators, automobiles and T-shirts are advancing relatively slowly … but so what?
I think one should reason backwards from the Singularity. What do we need to get to a positive Singularity? Superhuman, beneficially-inclined AGI. What do we need to get there? Either
A) computer science theory, cognitive science and computer hardware
or
B) brain imaging, brain science, and computer hardware
Are the directly relevant technologies exponentially accelerating?
Yes.
Bingo.
I was under the strong impression that Kurzweil implies that all of science and tech are accelerating exponentially.
I’m not talking about a Singularity in this post here, so the comment about reasoning backward from the Singularity is off topic. Still, since you bring it up, yes many of the directly relevant tech for AGI is accelerating, except what we need most: AGI funding, AGI theory, and FAI theory.
Clock speed increases do seem to be gradually slowing down. This effect alone blows a fairly spectacular hole in the thesis of “singularity” enthusiasts – that the same amount of IT progress will take 1 year, then 6 months, then 3 months, etc, etc.
From the long term historical perspective, the rate of change in systemic complexity has been accelerating for many billions of years. This is most clearly evident and internalized through devices such as Carl Sagan’s cosmic calendar which allow one to mentally map the universe’s developmental history to date.
Space and time started out small, dense and fast in the Big Bang, expanded and slowed down, and now is accelerating inward in localized pockets of evolutionary complexity such as on earth.
Human progress of all forms is limited by the net productive output of human intelligence. Human population has been increasing exponentially and new technological developments have conterminously improved our productive efficiency, so you have a clear double exponential right there, possibly only slowing recently with the leveling off of population growth.
Of course, just as the human S curve is leveling off, we have a new source of computational intelligence in the form of our computers, which are increasing exponentially in circuit complexity per capita, quantity, and higher level algorithmic efficiency.
Clock speed may have leveled off, but it leveled off about 1 million times faster than human clock speed, and isn’t the primary limiting factor regardless.
Gamefly offer a wide variety of games, with over 7000 titles to their credit. It’s very easy to make use of their services. Just register and pick your plan and begin to rent! When you gain access to their services, you can browse their selection of games; each game page has its own summary, reviews, and even cheats.