Bits, Atoms, and the Future of Manufacturing

(Updated 9 Feb 2010 with new estimates of value captured by Chinese manufacturers.)
(Updated 12 Feb 2010 with cost estimates for the iPad.)

Wired's cover story this month proclaims that "In theNext Industrial Revolution, Atoms Are the New Bits".  As I address this "revolution" in the last chapter of my book, and as we are thinking hard about manufacturing issues as the LavaAmp moves forward, I have a few observations about the topic.

Chris Anderson, the Editor of Wired and author of the piece, asserts as the core of his story that

The tools of factory production, from electronics assembly to 3-D printing, are now available to individuals, in batches as small as a single unit. Anybody with an idea and a little expertise can set assembly lines in China into motion with nothing more than some keystrokes on their laptop. A few days later, a prototype will be at their door, and once it all checks out, they can push a few more buttons and be in full production, making hundreds, thousands, or more. They can become a virtual micro-factory, able to design and sell goods without any infrastructure or even inventory; products can be assembled and drop-shipped by contractors who serve hundreds of such customers simultaneously.

To summarize (and oversimplify) Anderson's article, the future is about innovators -- individuals, really -- having access to manufacturing for small runs that can be scaled up as needed.  Design will be digital, and as the appropriate machinery continues to increase in capability and fall in price, manufacturing will increasingly be digital too.

In a full snark-on mode response at Gizmodo is Joel Johnson, with a blog post entitled "Atoms Are Not Bits; Wired Is Not A Business Magazine".  The snark distracts from some otherwise interesting analysis, which you can read and which I will get to below.  The comments on the post are unusually perceptive and many were made by people who are clearly plugged in at a professional level to the atoms-and-bits manufacturing story.

The summary of Johnson's argument is that small production runs equal small money, prototyping is not manufacturing, labor is cheap in China -- so what?, and some/all of the small production available now may be the result of oversupply due to the recession.  Many of these points are probably cogent.

But even when it comes to demonstrably successful examples of bootstrapping from rapid prototyping to international sales, Johnson is skeptical.  Wired's Anderson points to Aliph, makers of the Jawbone line of headsets, as an example of a virtualized business (ie heavy on creativity and IP, but with minimal capital infrastructure) that supports his case.  Johnson shoots back with this:

It's great that hobbyists can make ever more complex items, sell them on the internet, and have a small business. But the same process used by Aliph to manufacturer Bluetooth headsets (and bear in mind it takes 80 people just to coordinate this!) is exactly the same outsourcing process used by Apple to make iPhones.

Here Johnson makes a very interesting point, but is so full of kvetch that he trips over it and misses the significance of his observation.  Of the 80 or so people working at Aliph, only 8-10 will be actually working directly on engineering or manufacturing.  At least half the full count will be in sales, marketing, and customer service, with the rest distributed in IT, administration, and support, and finally with a few (probably 5-8) executives atop the whole thing.  The interesting bit is that those 8-10 people are able to coordinate the same sort of production infrastructure used by Apple and its many hundreds (thousands?) of staff in engineering and manufacturing.  

When it comes to manufacturing labor and cost, there are a few other observations that are worth pulling into this discussion.

  • First up is an article from The Economist last week about the results of recent teardowns on smartphones.  The numbers from iSuppli are interesting: of the four leading phones they took apart, the cost of components falls in a very narrow range of $170-180.
  • Next is a blog post from Slashdot a couple of months ago pointing to stories about the value breakdown on the retail sale price for consumer electronics.  The post refers to some analysis by Edmund Conway at The Telegraph suggesting the value added for an iPod assembled in China is only "a couple of dollars".  On a ~$200 widget that, let's call that $2, or 1%.  That number should hold for anything resembling a smartphone, which means assembly labor plus overhead (and local profit) adds only about $2 to the phone, too.  (Update 2: iSuppli evidently already has done a teardown on the iPad, and "manufacturing costs" are about 5% of the total component costs.)
  • Minimum wages in China range from ~$.4 to ~$.7 per hour, so that $2 in labor would pay for several hours total time.  This has to be an overestimate, by a long way.  I'll bet the assembly doesn't require more than a few minutes of labor per unit, with the rest going to overhead and profit.  Another issue is that Apple is probably paying Foxconn and its employees above minimum wage in order to retain trained labor, keep IP inside the company, and keep down the "fair day's wage" complaints from shareholders and critics in the US.  But that is just a guess, of course.  (Update 1: In a February 8th column at the NYT, Roger Cohen reports that a watch manufacturer in Dongguan is pulling down 5-8% of the retail price of various brands.  Wages are running $150-200 per month in that part of China.  I don't think this changes the numbers I've used above and below.)
  • With so little labor involved in the assembly, what other options are available to manufacturers today?  At the US minimum wage, spending that same few minutes assembling a doodad in the States would add a few more dollars to the cost -- not so much.  I'd pay that difference to know something was made here in the States.  The overhead on the factory is another matter, though.  It could be that paying for the real estate and the rest of the capital equipment makes assembly in the US uncompetitive.  (Though it would be the construction of the factor and the initial installation of the equipment that made the difference in cost -- the material cost of the building and the equipment would be roughly the same in China.)
  • That said, there is plenty of industrial land in Detroit laying fallow at the moment.  Tax breaks to build assembly plants in depressed US cities could probably bring a lot of those jobs home.  Yes, they would be minimum wage, but there are a lot of people here waiting for any job.

If you add the value of the software to the cost (say ~$30 on an iPhone), then you dilute the value of the assembly labor even more.  Thus even more of the value of the object is in bits rather than atoms and their arrangement, and the difference in wages between China and the US is diluted further still.  Even the arrangement of the atoms is really about bits, since all the sub-components of an iPhone roll off of manufacturing lines with minimal labor involved.

Around the office, we have been pondering many of these issues as they relate to biological technologies and the LavaAmp in particular.

We continue to refine the hardware design of the LavaAmp, and it looks like we have the production hardware down to 5 or 6 components, 4 of which are injection molded plastic.  The labor will only be in assembly of the final box, as all sub-assemblies should all come off automated fab lines of one kind or another.  All the real cost is in the design and tooling -- once we get up and running the per unit costs should be quite reasonable.

The reason that this is worth a larger discussion is that Biodesic is exploiting all of the trends and resources that Anderson writes about, however we are building not a consumer electronics widget but rather a tool that will facilitate the manipulation of biological systems.  As the boundary between bits and atoms blurs in one area (consumer electronics), the resulting improvements in design and manufacturing capabilities create opportunities for further blurring the boundary between bits and atoms in biology.  The LavaAmp should enable many more people to query DNA in their environment, and possibly even to play with PCR assembly of genes and genetic circuits, which is an experiment I am keen to try.  Trends like this will continue to put technologies into the hands of an ever wider range of people around the planet.

If you work with this sort of technology on a daily basis, what I wrote above comes as no surprise.  But that describes a very small minority of hardware and wetware hackers.  Many more people will come to realize it soon.  New manufacturing realities and the resulting new tools are about to contribute substantial change to our economy.

Here Comes the Wolf Pack

DARPA has just awarded $32 million to Boston Dynamics to build a deployable version of BigDog, the Legged Squad Support System (LS3).  Here is coverage at NetworkWorld, and here is Gizmodo's short note.  BigDog is part of the inspiration for my thinking about Cowborgs that wander around fields munching grass and producing biofuel (see "The New Biofactories", from last year's McKinsey Quarterly special What's Next).

But there is another obvious application that has been lurking in the back of my mind.  The spec calls for the following capabilities:

  • The robot must support all manner of walking, trotting, and running/ bounding and capabilities to jump obstacles, cross ditches, recover from disturbances and other discrete mobility features. The LS3 must be able to follow a leader between 5m and 100m ahead, in dynamic, cluttered environments with other moving soldiers in close proximity.
  • It must have the ability to perceive and traverse its immediate terrain environment autonomously with simple methods of control.
  • The robot must understand simple soldier-to-LS3 interaction with minimal direct control of the platform's speed and heading (joy-sticking and tele-operating are examples of direct control). The vehicle must require minimal oversight or direct control (e.g. joystick control) from an operator.  Direct control modes should only be used for error recovery, and should not be needed more than 3 times per 24-hour operational period, for no more than 5 minutes at a time.  

With the sort of stability and mobility required to meet these specifications, there isn't any reason you couldn't mount weaponry on the LS3.  You could imagine all sorts of science fiction scenarios with Miniguns or Vulcans, which for all I know might have serious mass issues for either the armament or the ammunition, or might sport too much recoil.  But it would probably be enough to mount cartridge fed shotguns or smaller grenade launchers, or any other weapons platforms now fielded on ROVs.  The combination of autonomous terrain traversing, navigation, and operation from a distance suggest that even early versions could be directed to walk into hostile situations while troops remained out of harm's way.  The requirement to "follow a leader" could be altered to "home in on a transponder" delivered by one of the many ROVs already in the field, whether RC car or airplane.

And just as there is no reason to think the US military won't be mounting weapons on BigDog, there is no reason to think the robots won't be operated in groups.  Imagine for a moment that you are a Taliban or Al Qaeda fighter hanging out in a cave, and you probably don't have a lot of exposure to technology other than what the US military is throwing at you on a daily basis.  Into your cave walks a Wolf Pack of armored DevilDogs armed to the...teeth (?) and probably  demanding your surrender in Arabic or Pashtun.  You say no.  They open fire.  Alternate scenario: You are a militant in Afghanistan and your exquisitely planned ambush of NATO troops is interrupted by a Wolf Pack chasing you up into the hills.

I can imagine that both of these scenarios would require a serious programming effort before a BigDog becomes a DevilDog.  But if either scenario works even once, just imagine the impact on enemy morale.  And how long before the LS3 becomes an ordinance delivery platform, walking into an enemy camp with a 400 lb bomb on its back?  Powerful stuff, that, both as a tactical weapon and a morale buster.

I don't know how I feel about this.  Yes, a Wolf Pack of DevilDogs would probably keep our troops safer.  And this might be a more effective way of hunting down bad guys.  But the spectre of increasingly autonomous weapons platforms should make everyone a little uneasy.

Then again, this sort of investment will make the Cowborg happen that much sooner.  Is that worth it as a sort of "peace dividend" spin off from military spending?  Hmmm.

25% of US Grain Crop Used for Biofuel

The Guardian UK reported today that 2009 USDA figures show 25% of grains grown in the US were used to produce liquid biofuels.  The typical food vs fuel story follows.  And it is mostly on point, if tinted by The Guardian's usual populist tone.  Yes, all the grain could in principle be used to feed people.  No, it isn't clear that grain-based ethanol is in fact better than burning petroleum when it comes to total greenhouse gas emissions or energy content.

The story ends with a nod toward "continued innovation in ethanol product" that supposedly is increasing yields and reducing costs.  Huh.  No mention, though, of the fact that any starch crop used to make fuel starts at a major disadvantage with respect to sugar crops, nor that there is an ethanol glut in the US due to construction of too many ethanol production plants.  Neither does the story get into why ethanol isn't a very good fuel to begin with (wrong solvent properties, low energy content, water soluble).

I go into detail about this in my forthcoming book, but the upshot of the argument is that the US is investing quite a lot of money in ethanol production technology and infrastructure that will never be competitive with sugar derived fuels.  And then relatively soon we will get butanol, longer chain alcohols, and true drop-in petroleum replacements made using modified organisms.  In the meantime, I suppose we will just have to suffer through the impact of decisions made more for political reasons than for competitive or national security reasons.  But grain to ethanol isn't really good for anybody except US Senators from farm states.

Good Climate Data, Bad Climate "Data" -- Science Always Wins.

This week brings news of 1) a dramatic improvement in the estimates of how soil carbon content is related to atmospheric carbon concentration and 2) the exposure of some really crappy work on the rate of melting of Himalayan glaciers by the International Panel on Climate Change (IPCC).  The soil carbon work is Good Data, but Bad News if you care about the effects of high atmospheric carbon concentrations, while the Himalayan glacier story is all about terrible peer review and Bad Data (non-existent data, actually), which doesn't help anybody figure out the real story on water supplies in Asia.

First up, a paper from this week's PNAS by Breeker et al at UT Austin, "Atmospheric CO2 concentrations during ancient greenhouse climates were similar to those predicted for A.D. 2100".  Already from the title you can see where this is going.

The problem Breeker and colleagues address is the following: how do you correlate the carbon content of fossil soils with prevailing atmospheric carbon dioxide concentrations?  Well established methods exist for measuring the carbon content of compounds in fossil soil, but less certain were conditions under which chemical reactions produce those particular compounds.  It turns out that model used to infer atmospheric CO2 contained an error.  Breeker determined that the primary compound assayed when determining soil carbon content forms at much lower atmospheric CO2 concentrations than had been assumed.

Prior attempts to correlate soil carbon (and by proxy atmospheric CO2) with greenhouse periods in Earth's climate had concluded that warm periods experienced CO2 concentrations of much greater than ~1000 parts per million (ppm).  Therefore, one might conclude that only when average atmospheric CO2 spiked above this level would we be in danger of experiencing greenhouse gas warming that threatened glaciers.  The correction supplied by Breeker substantially lowers estimates of the average CO2 concentration that is correlated with continental glacial melting.  Eyeballing the main figure in the paper, it looks to me like we could be in real trouble above 450 ppm -- today we are at just shy of 390 ppm and there is no sign we will be slowing down anytime soon, particularly if India and China keep up their pace of development and emissions.

Looking forward to 2100, things get a touch squiffy because Breeker relies on an estimate of CO2 concentrations that come out of model of global economic activity.  So the title of the paper might be a tad alarmist, simply because 2100 is a long way out for any model to be taken too seriously.  But the correction of the paleodata is a big story because at minimum it reduces the uncertainty of atmospheric CO2 levels, and it appears to clarify the connection between CO2 levels and continental glaciation.  More work is needed on the later point, obviously, or this paper would have been on the cover of Science or Nature.

Now on to a serious screw-up at the IPCC.  Elisabeth Rosenthal at the NYT is reporting that "A much-publicized estimate from a United Nations panel about the rapid melting of Himalayan glaciers from climate change is coming under fire as a gross exaggeration."  Here is Andrew Revkin's take on DotEarth, and anyone interested in this story should read through his post.  The comments are worth perusing because some of the contributors actually seem to have additional useful knowledge, though, of course, nut jobs aplenty show up from both sides of the debate over climate change.

In a nutshell, the issue is that the most recent IPCC chapter on glaciers contained a conclusion, advertised as real analysis, that was in fact a speculation by one scientist promulgated through the popular press.  The authors of that section of the IPCC report may have been warned about the unsubstantiated claim.  Contradictory data and analysis seems to have been ignored.

So, to be frank, this is a giant, inexcusable fuck-up.  The IPCC is composed of so many factions and interest groups that this may be a case of simple blundering or of blatant politicization of science.  But here is the beautiful thing about science -- it is self-correcting.  It may take a while, but science always wins.  (See also my post of a couple of years ago, Dispelling a Climate Change Skeptic's "Deception".)  Every newspaper story I have seen about this particular IPCC screw-up notes that it was brought to light by...wait for it...a climate scientist.  It is an excellent public airing of dirty laundry by the community of science.  So while this episode demonstrates that the last official IPCC report on glacial melting in the Himalayas should not be used for any sort of scientific policy recommendation or economic forecast, you can bet that the next report will do a damn fine job on this topic. 

Finally, whether or not the IPCC gets its act together, there are plenty of good data out there on the state of the planet.  Eventually, Science -- with a capital S -- will get the right answer.  The same methodical process that has resulted in computers, airplanes, and non-stick fry pans will inevitably explain what is really going on with our climate.  And if you use computers, fly on airplanes, or eat scrambled eggs then you are implicitly acknowledging, whatever your political or religious persuasion, that you believe in science.  And you better, 'cause science always wins.

Video from The Economist's World in 2010 Festival

The Economist has posted video from the World in 2010 Festival, held in Washington DC in early December.  The Innovation panel is below, with me (Biodesic), Dean Kamen (DEKA Research), Dwayne Spradlin (Innocentive), and Kai Huang (Guitar Hero), moderated by Mathew Bishop (The Economist).  (Here is a link to video selections from the rest of the event.)  I was chatting with a reporter a few days ago who observed that everyone else on the panel is quite wealthy -- hopefully that bodes well for me in 2010.  But maybe I am destined always to be the odd man out.  C-Span is re-running the video periodically on cable if you want to watch it on a bigger screen, but I can't seem to find an actual schedule.  (Here is their web version: Innovation in 2010.)


I have a couple of general thoughts about the event, colored by another meeting full of economists, bankers, and traders that I attended in the last week of December.  I met a number of fantastically accomplished and interesting people in just a few hours, many of whom I hope will remain lifelong friends. 

First, I have to extend my thanks to The Economist -- they have been very good to me over the last 10 years, beginning in 2000 by co-sponsoring (with Shell) the inaugural World in 2050 writing competition.  (Here is my essay from the competition (PDF).  It seems to be holding up pretty well, these 10 years later, save the part about building a heart.  But at least I wasn't the only one who got that wrong.)

Here is a paraphrased conversation over drinks between myself and Daniel Franklin, the Executive Editor of the newspaper.

Me:  I wanted to thank you for including me.  The Economist has been very kind to me over the past decade.
Franklin: Well, keep doing interesting things.
Me:  Umm, right.  (And then to myself: Shit, I have a lot of work to do.)

On to the World in 2010 Festival.  The professional economists and journalists present all seem to agree that we have seen the worst of the downturn, that the stimulus package clipped the bottom off of whatever we were falling into, and that employment gains going forward could be a long time in coming.  Unsurprisingly, the Democratic politicians and operatives who turned up crowed about the effects of the stimulus, while the Republicans who spoke poo-pooed any potential bright spots in, well, just about everything.

At the other meeting I attended, last week in Charleston, SC, one panel of 10 people, composed Federal reserve and private bankers, traders, and journalists couldn't agree on anything.  The recovery would be V shaped.  No, no, W shaped.  No, no, no, reverse square root shaped (which was the consensus at The World in 2010 Festival).  No, no, no, no, L shaped.  But even those who agreed on the shape did not agree on anything else, such as the availability of credit, employment, etc.

Basically, as far as I can tell, nobody has the slightest idea what the future of the US economy looks like.  And I certainly don't have anything to add to that.  Except, of course, that the future is biology.

Here is John Oliver's opening monologue from the Festival.  He was absolutely hilarious.  Unfortunately you can't hear the audience cracking up continuously.  I nearly pissed myself.  Several times.  (Maybe the cocktails earlier in the evening contributed to both reactions.)

Back to Innovation in 2010.  Dean Kamen had this nice bit in response to a question about whether the imperative to invent and innovate has increased in recent years (see 36:20 in the C-Span video): "7 billion people can't be recipients, they have to be part of the solution.  And that is going to require advanced technologies to be properly developed and properly deployed more rapidly than ever before."

To this I can only add that we are now seeing more power to innovate put into the hands of individuals than has ever occurred in the history of humanity.  Let's hope we don't screw up.

Quote of the Day

The semi-ancient electrophoresis power goes on the fritz, so of course we open it up to see what's what.  Lot's of discrete logic, some massive (1 milli Farad each!) capacitors, a couple of Motorola 68000s.  Nothing obviously amiss, and no hope of figuring out what has gone wrong.

But we did get this gem: "Oh, that's where the beeping noise comes from."  That's what makes it all worthwhile.

In the News

The December issue of Nature Biotechnology contains a special focus section on synthetic biology.  Here is my commentary, "The Changing Economics of DNA Synthesis".  The PDF version appears to be available for free, at least for the moment.

My interview in the Fall 2009 issue of GBN:bulletin is now available from the Global Business Network.  The issue also contains a short interview with Stewart Brand about his new book -- it is well worth reading.

Revisiting Mood Hacking with Scents

Following on my post last spring about mood hacking, October brought more hints that behavior can be explicitly modified using scents.  A variety of news outlets picked up on a press release from BYU describing a forthcoming paper in Psychological Science that demonstrates, "that clean scents not only motivate clean behavior, but also promote virtuous behavior by increasing the tendency to reciprocate trust and to offer charitable help."  Here I am quoting from a pre-print, entitled "The Smell of Virtue", cached at the University of Toronto.  The paper describes two experiments in which citrus-scented window cleaner appeared to alter behavior.  I have to say that I found the references to Proust, saints, sinners god, and cleanliness (all that in 4 pages!) to be distractions from the main ideas, not to mention the data.

Here is the ScienceDaily reporting, and here is Time's take.

(Not everyone is happy with the methodology described in the paper, the conclusions, and the way it was written.)

What makes this interesting (to me) is that the researchers don't necessarily imply a direct biological mechanism.  The induced behavior may simply be the result of a learned association.  That is, there is no suggestion that anything about the scent that serves to flip a biological switch that leads to different behavior.  Rather the lead author, Katie Liljenquist of BYU, and her colleagues had  previously demonstrated a link between transgression and a desire for cleanliness (see "Washing Away Your Sins: Threatened Morality and Physical Cleansing", Science, 313(5792), 2006).  "Out, damned spots!" and all that.

The citrus scent may simply something that Prof. Liljenquist's test subjects (probably undergraduates at US universities) have learned to associate with cleanliness.  Would students at Asian universities have the same response to the same scent?  I suppose one way to quickly address this question is to see what sort of scents Asians prefer in their window cleaners.  Here is my point: even though there may be no innate molecular pathway exploited in this "behavior reprogramming", it may still be possible to exploit culturally defined (or perhaps "contextually constructed") neural pathways (from the receptors to the brain) for the purpose of mood hacking.

I am not particularly excited about the possibility of having my own mood hacked without my knowledge.  That this might be accomplished even in the absence of genetically identifiable response pathway should give one pause.  Any molecular pathway responsible for this effect (should it prove reproducible and engineerable) is unlikely to be well understood for many years to come.  But if the results from the citrus-scent study are to be believed, then it is already possible to manipulate behavior using scents, even though we have little idea how to defend against it other than by using more scents.  Perfume warfare.  Lovely.

Can't wait until the iGEM undergraduates get a hold of this.  They have already built bugs that smell like bananas and mint.  When will they start trying to influence the judges' decisions directly using synthetic scent pathways?