TechnologyQuarterly

Open-source medical devices

Dealing with nuclear waste

Burt Rutan, a maverick in fl ight

June 2nd 2012

Robots onthe front-line

How far should they beallowed to go?

TQCOV-JUNE.2012.indd 1 22/05/2012 16:17

Monitor

1 Sloppy microchips, smarterirrigation, a new twist on robotarms, greener brick kilns,recycling washing enzymes,keeping oil pipes clear,peer­to­peer technical support,wooden batteries, �tness forgamers and the joys ofzoomable interfaces

Di�erence engine

8 Talking trashHow waste incinerators havecleaned up their act

Robots go to war

9 On the marchHow far should robots be allowedto go on the battle�eld?

Acoustic micro�uidics

11 A sound ideaManipulating droplets of �uidusing sound has many uses

Inside story

13 Hot stu�What should be done withhigh­level nuclear waste?

Open­source medical devices

15 When code can kill or cureAn open­source approach couldboost safety and innovation

Brain scan

17 Burt RutanA pro�le of the maverickaeronautical engineer

The Economist Technology Quarterly June 2nd 2012 Monitor 1

DECADES of �manna from heaven�.That is how Trevor Mudge, a comput­

er scientist at the University of Michigan,Ann Arbor, describes the technologicalimpact of the steady doubling, roughlyevery two years, of the number of transis­tors that can be crammed onto a siliconchip. It has increased the processing powerand storage capacity of computers whilereducing their size and energy consump­tion. The results are all around us. But thissteady doubling (known as Moore’s lawafter Gordon Moore, the engineer who�rst pointed it out in 1965) cannot go on forever, and nearly �ve decades later thelimits may �nally be within sight.

Transistors now measure as little as 22nanometres (billionths of a metre) inwidth, smaller than the wavelength oflight with which they are etched, and justa few tens of atoms across. As transistorsget smaller, keeping them cool and error­free becomes more di�cult. Lowering thevoltage at which transistors operate pro­duces less heat, but further reductions arenow di�cult because feebler voltagesresult in more frequent errors. The marchtowards ever­smaller and faster chips �isstarting to come unglued�, says Dr Mudge.That is why he and other computer scien­tists are taking a new approach: designingmicrochips that can tolerate errors in theiroperation. Such inexact or �sloppy� chips,as they are also known, can be smaller,

faster and more energy­e�cient.The important thing is to control where

the errors occur. Masahiro Fujita, who isdesigning sloppy chips at the Universityof Tokyo, notes that a single mistake at thebeginning of a sequence of instructionscan propagate through subsequent calcu­lations and completely mess up the be­haviour of a computer or robot. But, hesays, even relatively numerous mistakesare no big deal in other circumstances,such as when handling sound, images orvideo. Tiny sound distortions or slightlymiscoloured pixels will go unnoticed.

An international team of researchers atRice University in Houston, Texas, theSwiss Centre for Electronics and Micro­technology (CSEM) in Neuchâtel andNanyang Technological University inSingapore found that by reducing theoperating voltage, sloppy chips coulddeliver equivalent performance to ordin­ary chips using a quarter of the energy. Foraudio playback, the researchers found,sound quality was acceptable even witherror rates of 8%.

Hearing aids or mobile phones shouldlet people trade sound quality for batterylife, says Krishna Palem, the head of theproject. When the battery is running low,they could switch to a high­error modeand put up with the static. A prototypesloppy chip, developed with funding fromIntel, the world’s largest chipmaker, and

Oh, that’s near enough

Computing: Letting microchips make a few mistakes here and there couldmake them much faster and more energy­e�cient

On the coverRobots are playing anincreasingly important role inwarfare on land, at sea and inthe air. But as they becomesmarter,more capable andmore deadly, they are raisingdi�cult ethical quandariesabout the extent to whichthey should be able to maketheir own decisions, page 11

Contents

1

2 Monitor The Economist Technology Quarterly June 2nd 2012

2

1

America’s Department of Defence, is nowready for production. Dr Palem plans toestablish a start­up to sell such chips foruse in hearing aids with long battery lives.

Another trick is called �pruning�.Chips are wired so more power is deliv­ered to more important areas, while areasthat compute non­essential data (or areinfrequently used) are given less power orsimply removed altogether. Tests at CSEM

by Avinash Lingamneni found thatpruned circuits were twice as fast, con­sumed half as much energy and were halfthe size of conventional circuits.

Managing the probability of errors andlimiting where they occur can ensure thatthe errors do not cause any problems. Theresult of a mathematical calculation, forexample, need not always be calculatedprecisely�an accuracy of two or threedecimal places is often enough. Dr Palemo�ers the analogy of a person about tocross a big room. Rather than wasting timeand energy calculating the shortest path,it’s better just to start walking in roughlythe right direction.

Perfection is overratedThe microprocessor that powers a laptoptypically contains more than two billiontransistors, but it could run just �ne withone­eighth or even more of them produc­ing sloppy results, says Mr Lingamneni.The key is to ensure that sloppy circuitry isused only for certain tasks. It is not accept­able if one �le is attached to an e­mail butanother �le ends up being sent.

Can information that must be pro­cessed accurately be kept away from thefaster, sloppier circuits? Mr Lingamnenithinks it can. CSEM is developing an error­prone chip for audio­visual processing inmobile phones that dispatches di�erentprocessing tasks to the appropriate circuit­ry. The voltage is reduced in parts of thechip that �nish calculations faster or withgreater accuracy than necessary. Otherparts of the chip, where more importantcalculations are made, operate at highervoltage and with fewer errors.

Another approach to managing errors,

called �asymmetric reliability�, useserror­prone circuits for number crunchingto save power and run faster, says Sub­hasish Mitra, a computer scientist at Stan­ford University. Conventional circuits arethen used to spot and weed out unaccept­able errors. His work is funded by Ameri­ca’s Department of Defence and compa­nies including Bosch, Cisco, In�neon,Intel, Semiconductor Research Corpo­ration, Samsung and Texas Instruments.

With this technique, the mechanics ofcomputing become more akin to those ofhuman thought, says Joseph Bates ofSingular Computing, a �rm based inNewton, Massachusetts. In the �messyprocess� of brain activity, synapses trans­mit signals imperfectly between nervecells, he notes. Rather than squanderingresources in a futile attempt to eliminateminor errors, the brain �lters them out asprocessing moves up to higher mentalfunctions. Together with Charles RiverAssociates, a consultancy, Singular isworking on a $700,000 project for the US

Navy to design an �approximate comput­ing� video­processing chip to increase theability of battery­powered drones to trackpotential targets.

The technology behind sloppy micro­chips is now ready for widespread use,says Vivek De, Intel’s head of circuit re­search in Hillsboro, Oregon. But will com­panies that exploit sloppiness to boostperformance or battery life want to admitit? Buyers might think twice before buyinga device that operates with errors as amatter of course, concedes Hang ChangChieh, a professor at the National Univer­sity of Singapore.

Researchers have coined a variety ofeuphemisms to describe the technology,such as �inexact hardware�, �probabilisticcomputing�, �relaxed correctness� and�relaxed reliability�. In �ve years yoursmartphone or computer is unlikely tosport a sticker boasting that it containschips that sometimes get their sumswrong. Even so, such chips may be lurkinginside, increasing performance andextending battery life. 7

FROM the air, the Flint river basin insouth­western Georgia looks monoto­

nous. Appearances, however, deceive.What seems a Euclidian as well as a geo­graphical plain is actually a landscape oflow hills, shallow valleys, stream beds,drainage ditches and river banks. Nor isthe soil the same everywhere. Some issandy; some is rich in clay; some is a mix­ture of the two. These details matter if youare a farmer, which many residents are.

From the air, too, something else isapparent: a strange tendency for the ba­sin’s crops to grow in circular patches. Thisis caused by the way they are watered, forthe predominant system employs sprin­kler heads attached to hoses that danglefrom wheeled trusses which move in acircular pattern around a central tower.The whole arrangement is portable, andcan thus be transported from �eld to �eldin order to water a whole farm.

Centre­pivot irrigation, as this system isknown, is reasonably e�ective. But it is aone­size­�ts­all approach to distributingwater. Craig Kvien, of the University ofGeorgia, thinks he can do better. In 1999 DrKvien came up with the idea of varyingthe amount of irrigation water appliedacross a �eld, in light of a detailed exami­nation of that �eld’s characteristics. Incollaboration with FarmScan AG, anAustralian manufacturer of agriculturalequipment, he and his colleagues havedeveloped this approach into a techniquecalled variable­rate irrigation (VRI). Morethan 80 farms in Georgia now use it. Farm­ers in Australia, Germany, New Zealand,South Africa and Spain are taking aninterest, and two other �rms�Zimmaticand Valley Irrigation, both based in Ne­braska�have joined FarmScan AG inselling the kit.

VRI requires some e�ort to set up. Afarmer must produce a map of his landwith a resolution of less than a metre, todetermine its topography�particularlyany low­lying areas where water mightpool and higher spots that are prone torun­o�. Fallow areas, uncropped parts,watercourses, dirt tracks and wetlandsalso need to be fed into the system. Forfurther precision, a farmer can use soil­moisture probes to let him know howmuch water each bit of a �eld is using,since denser, clay­based soil requires less

Dribbles and bits

Agriculture: A new smart irrigationsystem, based on detailed mapping,can save water, cut costs and reducefertiliser run­o�

The Economist Technology Quarterly June 2nd 2012 Monitor 3

2

1

irrigation than looser, sandy soil.The farmer, or his agent, uploads all

this information into a piece of softwarewritten by one of the companies involved.This software uses the Global PositioningSystem, a network of satellites whosesignals provide precise location infor­mation, to monitor the position of eachsprinkler head as it turns on the pivot. Thesoftware then works out, on the �y, howmuch water should be emerging fromeach head at any given moment.

VRI systems are not cheap. Dependingon the size of the pivot and the complexityof the installation, it can cost $5,000­30,000 to smarten up a single irrigationsystem in this way. The payo�, though, isan average 15% reduction in water con­sumption, a reduction in fertiliser usebecause less is washed away by run­o�,and savings of $40­110 per hectare. 7

AROBOT’S hand is one of the mostimportant parts of its anatomy. In­

deed, it de�nes the machine’s capabilities.Industrial robots that do repetitive taskslike welding and cutting are usually single­handed. Those hands, moreover, tend tobe simple claws�as be�ts the simplicity oftheir job. Robots intended to work withand care for people, however, will needmore dexterity. They will have to performa wider range of actions than their indus­trial cousins, and do so more delicately.People, after all, can bruise in ways thatmetal panels cannot. As nature has al­ready concluded, two hands with �ngersare better suited to this range of tasks thanis a single claw. But hands with �ngersneed more internal machinery than claws,and are thus harder to design. To simplifythat task a group of researchers fromGermany and Italy, working as part of aEuropean robotics programme calledDEXMART, has borrowed an idea fromancient catapult­makers.

The most common way of moving arobot’s hands is with electric motors. Butpacking in the score or so of these, one foreach joint, that are needed to operate acomplicated robotic hand, and then add­ing the gears and other components re­quired to translate the motors’ move­ments into useful actions, means that ahand can quickly end up too large andunwieldy for domestic use. Some design­

ers of robots have tried to get around thisby adapting another of nature’s designs,and operating the joints of the �ngers andwrist using tendons pulled by motorslocated in the robot’s forearm, where thereis more space for them.

That is an improvement, but is still notideal. Which is where the DEXMART teamcomes in. Instead of having large motorsin the forearm to wind the tendons in andout, their design uses smaller ones to twistthe tendons. As anyone who came intocontact with Roman catapults 2,000 yearsago would have known, twisted �bres cantransmit hugely powerful force.

The group’s experimental hand doesnot throw boulders, but it can swiftlygrasp and lift a load weighing �ve kilo­grams using 24 small, high­speed motorsto twist a similar number of tendons. Eachof the hand’s �ve �ngers is operated byfour separate tendons (one for each of thethree joints in every �nger, and the otherto pull the �nger straight). The remainingtendons articulate the wrist: left and right,and up and down.

The hand is also �tted with sensors.Those on the motors measure the pullingforce. Those on the �ngers monitor thepressure being exerted. The �nger sensorsmeasure friction too, according to ChrisMay of Saarland University, in Germany,one of the team’s engineers. This meansthat if something which the hand is liftingstarts to slip, it can quickly tighten itsgrip�but only by the appropriate amount.It can thus pick up an egg without break­ing it. Equally, it can grasp a large bottle ofwater �rmly enough to pour from it.Whether it will be able to entertain itshuman companions by arm­wrestlingthem has yet to be determined. 7

A new twist onan old idea

Robotics: Researchers have taken anidea from ancient catapult­makersand used it to build a strong yetcompact robotic hand

Gently (but �rmly) does it

OUTSIDE a village called Mau, in UttarPradesh, half a dozen chimneys rise

from kilns into a colourless sky. Theseovens, six among the 100,000 which turnout the 200 billion bricks made each yearin India, are worked by dalits�membersof castes once regarded as untouchable.

India’s brick kilns are noxious sourcesof pollution, particularly soot, and work­ing them means a life that is always nasty,frequently brutish and often short. But ontop of this social evil is an environmentalone. The exhaust from the kilns mixeswith diesel emissions and other fumes toform a vast brown smog, known as anatmospheric brown cloud, which is up to3km thick and thousands of kilometreslong. Two of its main ingredients, thesmall carbon particles which the soot iscomposed of, and ozone, a triatomic formof oxygen, are important contributors tothe greenhouse e�ect, and thus to climatechange. Among other negative e�ects, thecloud is thought to be accelerating theretreat of Himalayan glaciers, which arefound at a similar altitude.

In theory, burning the coal that �res thekilns in a more e�cient and less pollutingway should save money for the kiln’sowners�an alignment of interests thatmight encourage the change to happen.Unfortunately, the main recommendedchange of design (at least, the changerecommended by the United NationsEnvironment Programme) is a ratherexpensive one: to switch from the tradi­tional style of kiln, known as a Bull’strench kiln, to a more modern designcalled a vertical­shaft kiln.

Vertical­shaft kilns cut soot emissionsby three­quarters, but they cost around10m rupees ($200,000) each and requiregood­quality clay, able to withstand rapidheating. That makes them too expensivefor most kiln owners. But recent researchconducted as a collaboration betweentwo Indian green­technology and consult­ing �rms, Greentech Knowledge Solutionsand Enzen Global Solutions, has suggestedsome more easily a�ordable changes thatcan be applied to existing kilns.

Greentech’s main suggestions are toincrease the number of air ducts in thekilns’ smokestacks and set the bricks to be�red in a zigzag pattern, rather than in thecurrent block arrangement. These two

Clever tricks forcleaner bricks

Environment: A few simple andrelatively cheap tweaks could greatlyreduce the environmental impact ofIndia’s brickmaking kilns

WHEN industrialists use enzymes tospeed up chemical reactions, they

generally take care to attach those en­zymes to solid surfaces and run the chemi­cals past them. Enzymes are expensive,and not to be thrown away lightly. Yetmillions of householders do precisely thatwhenever they wash their clothes. Lots ofwashing powders contain enzymes, butthese never get recycled. Instead, they arejust �ushed down the drain.

Chandra Pundir and Nidhi Chauhan,biochemists at Maharshi Dayanand uni­versity in Haryana, India, propose to dosomething about that. They see no reasonwhy washing enzymes should su�er thisignominious fate and, as they report inIndustrial and Engineering ChemistryResearch, they have found there is noreason why they should.

Their plan was to see if they could stickthe four enzymes used in washing pow­der�á­amylase, cellulase, protease andlipase�to PVC, a plastic that is cheap,chemically inert, wash­resistant, light­weight, easy to form into various shapesand nearly indestructible. To do so theytook a beaker made of PVC, �lled it with amixture of nitric and sulphuric acids, andleft it for six hours. Their purpose was tocleave the polymers on the beaker’s insidesurface into smaller molecules that havereactive chemical groups at their ends.

This done, they emptied out the acid,

washed the beaker with distilled waterand added a chemical called glutaralde­hyde. This primed the ends of small,cleaved polymers with aldehyde groups,to which the researchers expected each ofthe four enzymes to attach easily. Theythen poured a solution of all four en­zymes into the beaker and waited a day.The enzymes did, indeed, stick. Between40% and 70% of the enzyme molecules ofeach type were captured by the treatedPVC. Just as crucially, the process did notseem to a�ect their e�cacy.

The two researchers checked that byconducting the sort of test often seen inadverts for washing powder. They putwhite cotton cloths stained with starch,grass, egg or mustard oil into the beaker.They then washed the cloths with cheap,non­enzymatic detergents inside theenzyme­coated beaker and compared theresult with similar washing done in un­treated beakers using either the samenon­enzyme detergent, or an expensive,enzyme­containing detergent, or (as acontrol) plain water. After each wash theyanalysed the cloth for residual contentfrom the stain: albumin from egg; cellulosefrom grass; and so on.

The upshot was that the combinationof the cheap non­enzymatic detergent andthe immobilised enzymes in the beakergave results four or �ve times better thanthose provided by pricey enzymatic de­tergents in untreated beakers. Moreover,the enzyme­treated beakers continued towork for up to 200 washes.

The next stage is to devise a practicalway to make the enzymes meet theclothes. Dr Pundir’s and Dr Chauhan’s�rst attempt was a treated PVC scrubbingbrush. It worked well and, since mostIndian households wash their clothes byhand, should be rapidly deployable. Forthose with washing machines, though, adi�erent answer is needed: enzyme­coated plastic balls, perhaps. 7

Please rinse and return

Chemistry: It will soon be possible torecycle the enzymes used to washclothes, reducing waste and cuttingthe cost of doing the laundry

No need for enzymes to be �ushed clean away

4 Monitor The Economist Technology Quarterly June 2nd 2012

2 simple measures improve the circulationof air within a kiln, and thus the process ofcombustion. That, the company says,reduces a kiln’s emissions of soot by 60%.It also reduces its fuel consumption by 15%.The cost, around 1m rupees a kiln, canthen be recouped in three or four months.

Two further ideas in the �rms’ reportmight be a bit more di�cult to implement,but would also help. One is the wide­spread adoption of a technique common­ly used in central India, which involvesmixing coal dust with the clay the bricksare made from. This makes use of theotherwise useless dust, which burns, insitu, �ring the brick from within. It followsthat, because the coal is mixed with theclay, most of the resulting crud (70%, theresearchers found) is retained within thebrick once it has been �red.

The other proposal in the report is thatbrickmakers use machines which producehollow bricks. A hollow brick uses lessclay and thus requires less fuel to �re. It isalso a better product, because it providesmore insulation. Hollow­brick machinesare expensive, though, costing at least 20mrupees apiece.

Greentech reckons that if all of itsrecommendations were adopted, theywould save 5m of the 25m tonnes of coalwhich India’s brick kilns consume everyyear, and reduce the country’s annualemissions by the equivalent of 9m tonnesof carbon. Just as importantly, they wouldalso make the jobs of those who work thekilns rather less horrible. 7

The Economist Technology Quarterly June 2nd 2012 Monitor 5

1

COOL a mixture of natural gas andwater down to its freezing point and

the result is an icy material called meth­ane hydrate. Visionaries see this sub­stance, which is common in the depths ofthe ocean, as a novel source of fossil fuel.Oil companies, however, have a less rosyview of it, at least in the short term. Thereason is that one place in the oceandepths where methane hydrate forms alltoo frequently is inside pipes carrying oilfrom undersea wells to rigs at the surface.Like plaque in arteries, this accumulationrestricts the �ow of �uid and sometimesblocks things up completely.

One way to keep the oil �owing is toheat the pipe. That, though, is costly. An­other is to pump antifreeze, in the form ofmethanol, down it from time to time. Butmethanol is toxic, so this creates an envi­ronmental hazard. Kripa Varanasi and hiscolleagues at the Massachusetts Instituteof Technology, however, think they havefound a third way. As they describe inPhysical Chemistry Chemical Physics, theypropose to make the pipes’ interiors soslippery that methane hydrate will beunable to stick to them in the �rst place.

Dr Varanasi’s day job is developingwater­repelling compounds that can bepainted on to things like aeroplane wings,in order to stop them becoming coveredwith ice. It is not far from that to stoppinghydrate accumulating inside oil pipes.

To �nd a suitable substance with whichto coat the pipes, Dr Varanasi’s team built apiece of apparatus that looks a bit like anorgan. It is a series of glass tubes, eachopen at one end, that are �lled with hy­

drate­rich slush. (Because of methanehydrate’s propensity to catch �re, the teamused tetrahydrofuran hydrate, which hassimilar physical properties, instead.) Theopen end of each tube is pressed against asteel plate that has been coated with aseries of chemical mixtures that the re­searchers wish to test. Di�erent tubes abutdi�erent molecular combinations. Thewhole apparatus is then cooled to ­15°C fortwo hours, to mimic conditions in thedeep ocean, after which the tubes arebroken o� the steel plate one by one.

The crucial point is that the act ofbreaking is precisely controlled, and theforce required accurately measured whileit is going on. The amount of force neededshows how well stuck the tube was, andthus how well each chemical mixture onthe steel repells the hydrate.

Dr Varanasi and his colleagues tried tenpromising chemical cocktails. The winnerwas an 80/20 mixture of poly(ethyl meth­acrylate) and �uorodecyl polyhedraloligomeric silsesquioxane. So little forcewas needed to dislodge the tube that DrVaranasi thinks the mere movement of oilinside a pipe coated with it would beenough to keep that pipe clear.

Whether the process will work inpractice remains to be seen. Not least,experiments will now need to be carriedout with real methane hydrate, instead ofa substitute. But it looks promising. And ifit does work, it will be one of those in­cremental advances, important but oftenunnoticed and unreported, that help theworld, in both a literal and a �gurativesense, to run more smoothly. 7

Pipecleaner

Materials: Coating the insides of oil pipes in the deep ocean with slipperychemicals could prevent them from being blocked by icy deposits

FOR the past decade, technical supporthas been in the vanguard of global­

isation. With the costs of intercontinentalcommunication shrivelling to virtuallynothing, phone and online customerservices have migrated to wherever theycan be managed most e�ciently andcheaply. India blazed the trail, building a$5 billion outsourcing business on helpingWesterners solve high­tech problems.

Recently, the Philippines has taken overas the world’s call­centre hotspot, o�eringcomparable wage costs to India, with theadded bene�t�at least to North Americanears�of a Yankee drawl. But as 500,000Filipino customer­service representativesurge callers to have a nice day, they maywant to look over their shoulders.

Some of the biggest brands in software,consumer electronics and telecoms havefound a new workforce o�ering expertadvice at a fraction of the price of even thecheapest developing nation. They speakthe same language as their customers, andnot just in the purely linguistic sense�because it is the customers themselves.

�Unsourcing�, as the new trend hasbeen dubbed, involves companies settingup online communities to enable peer­to­peer support among users. Instead ofspeaking with a faceless person thou­sands of miles away, customers’ problemsare answered by unpaid individuals in thesame country who have bought and usedthe same products. This is done either indiscussion forums set up on the com­pany’s own website or on social networkslike Facebook and Twitter.

The savings can be considerable.Gartner, a market­research �rm, estimatesthat user communities can reduce supportcosts by as much as 50%. When TomTom, amaker of satellite­navigation systems,switched on social support, membershandled 20,000 cases in the �rst monthand saved the �rm around $150,000. BestBuy, an American gadget retailer, reckonsits 600,000 users save it $5m annually.

To motivate people to participate,Lithium, the software company that pro­vided TomTom’s and Best Buy’s systems,turns the whole thing into a game. Such�gami�cation�, increasingly ubiquitous inareas ranging from self­improvement toproject management, works by awardingpoints for a helpful answer, allowing

Outsourcing is so passé

Computing: �Unsourcing�, thereliance on contributions frominternet volunteers, may be thefuture of technical support

6 Monitor The Economist Technology Quarterly June 2nd 2012

2

THE main problem with both wind andsolar energy is not their cost (which is

falling satisfactorily with every passingyear) but their intermittency. Supplyingpower to the grid when the air is still orthe sun is below the horizon depends onstoring the surplus when the day is blus­tery and the sun is up. And, at the mo­ment, this is expensive.

Cheap and abundant materials formaking batteries, though, might changethat. Which is why a recent paper in Sci­ence, by Grzegorz Milczarek of PoznanUniversity of Technology, in Poland, andOlle Inganas of Linköping University, inSweden, may prove important. The tworesearchers propose making one of abattery’s three components, its cathode,out of the waste from paper mills.

A battery�any battery�consists of twoelectrodes (an anode and a cathode) andan electrolyte. Current, in the form ofpositive ions such as protons (the nuclei ofhydrogen atoms), �ows through the elec­trolyte from anode to cathode while abalancing current of electrons, which arenegatively charged, makes the same jour­ney via an external circuit, doing useful

work along the way. To recharge the bat­tery, the electrons are pushed in the oppo­site direction by (say) the current from asolar cell, thus drawing the ions inside thebattery back across the electrolyte.

Electrolytes are often made of simple,abundant (and therefore cheap) chemi­cals. The electrodes, however, are not.They usually require metals (lead, zinc,nickel or lithium, for example) whose costrenders so­called �grid scale� batteries,able to store huge amounts of power fordays or even weeks, prohibitively expen­sive. Cheaper electrodes would be a bigstep in the right direction and that is what,in the case of the cathode, Dr Milczarekand Dr Inganas hope they have made.

A good cathode material must becapable of receiving and storing charge, inthe form of positive ions and electrons, inlarge amounts. Lignin, one of the twomain components of wood, can be mod­i�ed to do just that. And lignin is cheap.Paper is made mainly of cellulose, theother component of wood, so the e�uentfrom paper mills, known as black orbrown liquor, is mainly water and lignin.

The reason Dr Milczarek and Dr In­ganas thought lignin molecules might besuitable for use as cathodes was that theyare rich in chemical groups called phenols,and phenols are easily turned into relatedgroups called quinones. It is these qui­nones that are the crucial components. Incombination with a second type of chemi­cal called a polypyrrole, they provide justthe sort of electron and proton receptors acathode requires. Polypyrroles are not ascheap as lignin, but compared with metalsthey are not expensive.

And so it proved. The two researchers’measurements suggest that the lignin­polypyrrole combination does, indeed,make an e�ective cathode, able to store alot of charge. These are early days, obvi­ously. But if someone could now come upwith an equally cheap anode, the age ofthe wooden battery�and with it the ageof cheap, reliable, always­on alternativeenergy�might yet dawn. 7

Woodenbatteries

Energy: Waste from paper mills couldbe put to use to make �grid scale�batteries that combine largecapacity with low cost

helpers to �level up�. This boosts theirstatus and often comes with a jokey hon­ori�c. Solve enough problems and youmight eventually become a �super fan�, inthe top 0.5% of responders.

Companies prize such fans especiallyhighly because they account for a dis­proportionate number of responses. Inthe case of Lenovo’s support network forits personal computers, a mere 30 superfans generated nearly half of the 1,200accepted solutions. A single super fan inthe Logitech network, code­named Kachi­Wachi, has posted over 45,000 responsesrelated to the company’s webcams.

There are limits to the scope of un­sourcing, however. Some things, such asbilling errors, have to be dealt with bysomeone with access to con�dentialcustomer data. Extending peer­to­peersupport to health, government and bank­ing could face legal and regulatory hur­dles. And Gartner warns that unsuccessfuldeployments risk a tidal wave of criticismon social networks. Even so, the prospectof cutting support costs seems likely toprove tempting for many companies. 7

GEEKS love games. But after spendinghours sitting in a seat programming or

building websites as part of their jobs,they often pass their idle hours ploppedinto a couch engrossed in a game or sittingon a commute, lobbing Angry Birds. It isnot a strategy for good health.

Dick Talens and his business partnerBrian Wang want to take advantage ofgeeks’ enthusiasm for games in order toimprove their �tness. Mr Talens is thetechnology chief of Fitocracy, which callsitself �the �tness social network to levelup in real life�. It is one of several sitesdevoted to �self­tracking� and the use of�gami�cation� to entice members toachieve goals. In his �rm’s case, you earnpoints by turning spare tyres into six­packs. Mr Talens, to judge by photos, hassucceeded. A chubby child, then a 100kgadult, he now weighs 80kg and competesin bodybuilding competitions.

Fitocracy awards points using a systemthat favours exercises that produce thebest results in the least amount of time.For instance, 50 body squats score 32points and �ve repetitions of a 60kg bar­bell squat gets you 57 points, while a farmore time­consuming one­hour walkearns 104 points. But the most interestingthing about Fitocracy is not the emphasison e�cient activities. Rather, it is that �thepeople who come to Fitocracy have neverpicked up a weight before and are franklygeeks like myself,� says Mr Talens.

The key is to give these novices modestand achievable starting points, such as awalk around the block. Gamers are usedto working through quests or levels ofincreasing di�culty. �You sort of trickpeople into doing it,� says Mr Talens. The�rm’s iPhone app allows users (if any canbe found without an Android smart­phone, which is de rigueur among pro­grammers) to update workout records onthe �y, too, winning points and badges.

Fitocracy is not alone. Nike’s newFuelBand, a wristband that measures thewearer’s activity level and awards pointsthat count towards a daily total, alsoseems to be aimed at game­loving techiesas much as sporty types. But Fitocracy isunabashedly geek­centric. Mr Talens wasonce sent a translation of a friendly gymtaunt comparing weight­lifting prowess,�Do you even lift?� It was in Klingon. 7

Fitness for geeksand gamers

Technology and health: A start­uptries to make �tness more appealingto game­loving techies by borrowingfeatures from role­playing games

The Economist Technology Quarterly June 2nd 2012 Monitor 7

ON JANUARY 16th 2003, 82 secondsafter the space shuttle Columbia lifted

o�, a piece of foam insulation weighingless than a kilogram broke o� its fuel tankand hit the left wing. Bosses at America’sspace agency, NASA, were largely reas­sured by a subsequent presentation deliv­ered by the �debris assessment� team.They did not request that a military spysatellite photograph the wing, in orbit,before re­entry. But its thermal protectionwas, in fact, badly damaged. As it re­en­tered the atmosphere 13 days later, theshuttle, and its crew of seven, burned up.

In a report eight months later, theColumbia Accident Investigation Boardcast considerable blame on the debris­assessment team’s presentation to higher­ups during the shuttle’s �ight, preparedusing PowerPoint software. Informationhad been poorly condensed onto 28 slides.On one cluttered slide the words �signif­icant� and �signi�cantly� were used �vetimes with a range of meanings from�detectable in a perhaps irrelevant calibra­tion case study� to �an amount of damageso that everyone dies�, noted EdwardTufte, a consultant to the investigators. Itwas easy to see how managers could haveviewed the slides without grasping thelevel of danger, the board concluded.

Rarely, of course, does a botched slidepresentation have such tragic conse­quences. But as Dr Tufte, a statistician atYale University, argues in his book �TheCognitive Style of PowerPoint: PitchingOut Corrupts Within�, cramming infor­mation onto slides all too often breedsconfusing jargon, alphabet soup, andverb­less phrases assembled into bogus

bullet­point hierarchies.A new approach, then, would be great­

ly welcomed by many presenters andaudiences alike�and �deep zooming�software may provide it. Zoomable userinterfaces (ZUIs), as they are known, arearriving on the coat­tails of touch­screengadgets such as the iPhone that havepopularised zooming to magnify graphics.

With ZUIs (pronounced zoo­ees), infor­mation need not be chopped up to �t onuniformly sized slides. Instead, text, im­ages and even video sit on a single, lim­itless surface and can be viewed at what­ever size makes most sense�up close fordetails, or zoomed out for the big picture.The presentation software designed byPrezi, a �rm based in Budapest, Hungary, isbased on this kind of �in�nite canvas�, asits founder, Peter Halacsy, calls it. Forexample, a naturalist delivering a pre­sentation on gira�e habitats can tucktables on, say, the nutritional qualities offoliage into the leaves of di�erent treespecies seen in satellite imagery of a sa­vanna. The data could be left hidden for atalk to schoolchildren, or zoomed in onand revealed for an audience of scientists.

Before giving a talk, a presenter canpick waypoints on the canvas to be visitedin sequence by pressing a button, withsmooth pans, zooms and rotations fromone to the next. But Prezi can also freepresenters from the predetermined se­quence epitomised by a �deck� of slides. Aspeaker can ��y� over the canvas, usingspatial memory to access informationwhen it is needed. Prezi has more than10m users and signs up another 1m or soeach month. It has become more promi­

nent after being used by some speakers atTED, a conference series adored by techies.(TED is an investor in Prezi.)

Forthcoming software for timelinepresentations, dubbed ChronoZoom,o�ers another zoom­based approach.Events are described or represented alonga timeline using text, images, and video.Zoom in so that a recent 24­hour section ofthe timeline �ts on a laptop screen, and atthis scale, the timeline stretches about 17billion kilometres to the left. Zooming outreveals the past decade, or century�or theentire sweep of 13.7 billion years of cosmictime. Zooming in again can show contem­poraneous developments in ancientBabylon and Mycenaean Greece, say, orhow matter behaved in the �rst fractionsof a second after the Big Bang. All this canprovide a �gut understanding� of histori­cal context and causation, says WalterAlvarez, the project’s co­founder.

Roland Saekow, the project’s leader,says that when demoing ChronoZoom helikes �to listen for the gasp� of amazement.About 30 people are developing Chrono­Zoom at Microsoft Research, MoscowState University and the University ofCalifornia, Berkeley, where Mr Alvarezand Mr Saekow work. A �curated� versionedited by professional historians, geol­ogists and cosmologists is in the works (atest version is already up). In a year or so afree version will be made available for useas a presentation tool.

Ready for your close­upThe zoom­based approach can transformmulti­page websites into a single broadsurface that simultaneously displays allcontent. Instead of clicking and waiting fora new page to appear, a visitor can zoomdirectly to areas of interest. On the HardRock Café website, a page built usingMicrosoft’s Silverlight software shows1,610 memorabilia items. By using thescroll wheel to zoom, details of each onecan be expanded to �ll the entire screen.

Software that zooms deep into movingimagery may be next. America’s Depart­ment of Energy is developing software todrill into scienti�c animations of particlebehaviour in nuclear reactions. CalledVisIt, its zooming range is equivalent tozipping from a view of the Milky Way to agrain of sand, says Becky Springmeyer,who is working on the project at the Law­rence Livermore National Laboratory.

Will deep zooming catch on? MarkChangizi, an evolutionary neurobiologistand author of �The Vision Revolution�,notes that the human visual system has tocope with zooming when moving, say,through a dense grove of trees. Today’szooming software operates on a di�erentscale, but as touch­screen devices prolifer­ate, zooming has become a popular wayto manipulate maps and photos�so per­haps it will catch on in other areas, too. 7

Prophets of zoom

Computing: New �deep zooming� software not only amazes�it o�ers a newway to display and navigate through large amounts of information

and as the rubbish piles up, o�cials are re­considering the sensitive issue of incinera­tion. Modern incinerators capture the en­ergy from solid waste as well as theemissions from the combustion. SuchWTE plants burn rubbish at temperatureshigh enough (over 850°C) to break the mo­lecular bonds in dioxins and other toxicchemicals and thus render them harmless.The �ue gases are cooled in heat exchang­ers, producing steam to drive electricity­generating turbines.

The gases are then passed to a cleaningsystem that �lters �ne particles from the

�ow and scrubs the gas to remove sulphurdioxide, acids and heavy metals. Next, the�ue gases pass through a catalytic conver­ter, where the nitrogen oxides are chemi­cally reduced using ammonia or urea. Fi­nally, any volatile heavy metals remainingin the �ue gas are absorbed by activated­carbon powder. The volume of the ash leftafter combustion amounts to around 5% ofthe waste ingested. The ash at the bottomof the combustion chamber is either bu­ried in municipal land�lls or recycled asaggregate for the construction industry.The �y ash that rises up the �ue needs fur­ther processing to remove any toxic parti­cles that might be clinging to its surface.The result is a remarkably clean and e�­cient process for disposing of rubbish.

Because much municipal rubbish (eg,paper, cardboard, wood, cloth and foodscraps) has a biological origin, the electric­

TOWARDS the end of �Toy Story 3�,Buzz Lightyear, Sheri� Woody and the

other toys �nd themselves heading intothe maw of a moving­grate garbage incin­erator. In real life, if the plant had beenbuilt before 1989, burning plastic toys in itwould produce a nasty dose of dioxinsand furans�toxic emissions from combus­tion in the presence of chlorine�alongwith heavy metals and some dubious or­ganic compounds. Until then, few peoplewere aware that such chemicals presenteda serious health hazard, capable of upset­ting the immune system, damaging the liv­er and causing cancer. Municipal incinera­tors were among the worst o�enders.

The industry subsequently spent bil­lions of dollars retro�tting incineratorswith activated­carbon injectors and parti­cle traps to capture the dioxins, furans andvolatile metals like cadmium and mercury.Thanks to new rules, the emission of suchtoxic chemicals from waste processing hasbeen reduced a thousandfold. The totaloutput of dioxins and furans from all theincinerators in America is now less thanten grams a year, according to the Environ­mental Protection Agency (EPA). Peopleburning rubbish in their backyards, bycontrast, may produce 50 times as much.

Even so, municipal incinerators�espe­cially the new waste­to­energy (WTE)plants that use rubbish as a fuel to generateelectricity and heat for local distribution�continue to have an image problem. Inmany countries people generally prefertheir waste to be composted (provided, ofcourse, the land�lls are nowhere near theirown backyards). But without costlyplumbing, land�lls produce copious quan­tities of methane, a powerful greenhousegas that does more than 20 times the dam­age to the environment as comparableemissions of carbon dioxide.

At some of the larger municipal land­�lls, the methane produced by anaerobicdecomposition is captured and used togenerate electricity. The mountain of rub­bish at the Puente Hills Land�ll in Los An­geles, the largest of the 1,900 municipalland�lls in America, is over 500­feet high�taller than most of the skyscrapers in thecity’s downtown area. With 60 years’worth of decomposing rubbish, PuenteHills produces enough methane to gener­ate electricity for 70,000 homes.

But land�ll space is increasingly scarce,

ity and heat produced by WTE plants isconsidered renewable energy. And for ev­ery tonne of municipal waste that avoidsbeing buried in a land�ll and is burned in­stead in a WTE plant, the amount of meth­ane entering the atmosphere is reduced bythe equivalent of almost a tonne of carbondioxide, calculates the EPA.

Nimbyism aside, the main objections toWTE incineration are that waste should berecycled rather than burned, and that bet­ter WTE technologies are waiting in thewings. Recycling requires the rubbish to besorted into recyclable metals, plastics, glassand paper, with the biological residue thenbroken down using anaerobic microorgan­isms into biogas and compost. The biogas(primarily methane, carbon dioxide andhydrogen sulphide) can then be used togenerate electricity, or even cleaned andcompressed to make fuel for cars.

That biological step might eventuallybe replaced with some form of heat­treat­ment, such as high­pressure superheatedsteam in an autoclave, or gasi�cation usinga plasma arc that vaporises the waste inthe absence of oxygen, so that few of thenoxious products of combustion are pro­duced. Either way, the result is a biofuel forgenerating electricity. Such technologiescould one day prove more attractive thantoday’s waste­to­energy incineration.

In his book �Garbology: Our Dirty LoveA�air With Trash�, the Pulitzer prize­win­ning author Edward Humes notes that oth­er wealthy countries with high living stan­dards have rejected the disposableproducts that make up much of America’srubbish. According to the OECD, the aver­age person creates 3.3lb (1.5kg) of rubbish aday in France, 2.7lb in Canada and no morethan 2.3lb in Japan. By the OECD’s reckon­ing, the average American produces 4.5lb aday, and more recent accounting puts the�gure at over 7lb a day, less than a quarterof which is recycled.

Why does America produce so muchmore rubbish? The di�erence is that in Eu­rope and Japan it is manufacturers, ratherthan consumers, that are held responsible(via taxes on packaging waste) for the costof processing the packaging used to wraptheir goods. This gives them an incentive touse less of it. By contrast, in America, thecost of cleaning up the mess is dumped atthe consumer’s door. That, more than any­thing, is what needs to change. 7

Talking trash

Energy from waste: Incinerators that use rubbish as a fuel to generate electricity and heat continue to have an imageproblem. That is unfair, because the technology has advanced considerably and has cleaned up its act

8 Di�erence engine The Economist Technology Quarterly June 2nd 2012

The Economist Technology Quarterly June 2nd 2012 Robots go to war 9

IN THE early afternoon of August 18th2008, a reconnaissance unit of about 100

French paratroopers, accompanied by asmall number of Afghan and Americansoldiers, was ambushed by a similarlysized Taliban force in the Uzbin Valley, notfar from Kabul. Ten French soldiers werekilled in �ghting that continued into thenight�France’s biggest loss since it sent sol­diers to Afghanistan in 2002. But it mighthave been avoided had the unit had a sin­gle aerial­robot scout, says Gérard de Bois­boissel, a specialist on military robots atthe French army’s Saint­Cyr military acad­emy. That assessment, shared by many, ledto a retooling of France’s armed forces. To­day drones, also called unmanned aerialvehicles (UAVs), routinely accompanyeven small French units.

More broadly, �ghting forces and intelli­gence services worldwide are equippingthemselves with all manner of robots thatoperate on land and sea, and in the air. Theconduct of war is being transformed�andlargely, it seems, to the West’s advantage.But knotty ethical quandaries are croppingup as the mechanical guts, electronic sen­sors and digital brains of robots continueto improve. Some fear that robots, whichare ingeniously mobile and can collect andprocess huge quantities of data, make it tooeasy to launch attacks. Others worrywhether robots can be trusted to maketheir own decisions while in combat.

Dragon�ies, �eas and dogsMilitary robots come in an astonishingrange of shapes and sizes. DelFly, a dra­gon�y­shaped surveillance drone built atthe Delft University of Technology in theNetherlands, weighs less than a gold wed­ding ring, camera included. At the otherend of the scale is America’s biggest andfastest drone, the $15m Avenger, the �rst ofwhich recently began testing in Afghani­stan. It uses a jet engine to carry up to 2.7tonnes of bombs, sensors and other typesof payload at more than 740kph (460mph).

On the ground, robots range from truck­

sized to tiny. TerraMax, a robotics kit madeby Oshkosh Defense, based in Wisconsin,turns military lorries or armoured vehiclesinto remotely controlled or autonomousmachines. And smaller robotic beastiesare hopping, crawling and running into ac­tion, as three models built by Boston Dy­namics, a spin­out from the MassachusettsInstitute of Technology (MIT), illustrate.

By jabbing the ground with a gas­pow­ered piston, the Sand Flea can leap througha window, or onto a roof nine metres up.Gyro­stabilisers provide smooth in­air�lming and landings. The 5kg robot thenrolls along on wheels until another hop isneeded�to jump up some stairs, perhaps,or to a rooftop across the street. Another ro­bot, RiSE, resembles a giant cockroach anduses six legs, tipped with short, Velcro­likespikes, to climb coarse walls. Biggest of allis the LS3 (pictured), a four­legged dog­likerobot that uses computer vision to trot be­hind a human over rough terrain carryingmore than 180kg of supplies. The �rm saysit could be deployed within three years.

Demand for land robots, also known asunmanned ground vehicles (UGVs), beganto pick up a decade ago after American­ledforces knocked the Taliban from power inAfghanistan. Soldiers hunting Osama binLaden and his al­Qaeda �ghters in the Hin­du Kush were keen to send robot scoutsinto caves �rst. Remote­controlled groundrobots then proved enormously helpful inthe discovery and removal of makeshiftroadside bombs in Afghanistan, Iraq, andelsewhere. Visiongain, a research �rm,reckons a total of $689m will be spent onground robots this year. The ten biggestbuyers in descending order are America,followed by Israel, a distant second, and

Britain, Germany, China, South Korea, Sin­gapore, Australia, France and Canada.

Robots’ capabilities have steadily im­proved. Upload a mugshot into an SUGV, abriefcase­sized robot than runs on caterpil­lar tracks, and it can identify a man walk­ing in a crowd and follow him. Its maker,iRobot, another MIT spin­out, is bestknown for its robot vacuum cleaners. Itslatest military robot, FirstLook, is a smallerdevice that also runs on tracks. Equippedwith four cameras, it is designed to bethrown through windows or over walls.

Another throwable reconnaissance ro­bot, the Scout XT Throwbot made by Re­con Robotics, based in Edina, Minnesota,was one of the stars of the Ground Robot­ics Capabilities conference held in San Die­go in March. Shaped like a two­headedhammer with wheels on each head, theScout XT has the heft of a grenade and canbe thrown through glass windows. Wheelspikes provide traction on steep or rockysurfaces. In February the US Army ordered1,100 Scout XTs for $13.9m. Another ver­sion, being developed with the US Navy,can be taken to a ship inside a small aquat­ic robot, and will use magnetic wheels toclimb up the hull and onto the deck, saysAlan Bignall, Recon’s boss.

Even more exotic designs are in devel­opment. DARPA, the research arm ofAmerica’s Department of Defence, is fund­ing the development of small, soft robotsthat move like jerky slithering blobs. EATR,another DARPA project, is a foraging robotthat gathers leaves and wood for fuel andthen burns it to generate electricity. Re­searchers at Italy’s Sant’Anna School ofAdvanced Studies, in Pisa, have designed asnakelike aquatic robot. And a small heli­

March ofthe robots

Robotics: From reconnaissance tobomb­defusal to launching attacks,military robots are on the march,raising knotty ethical quandaries

1

10 Robots go to war The Economist Technology Quarterly June 2nd 2012

2 copter drone called the Pelican, designedby German and American companies,could remain aloft for weeks, powered byenergy from a ground­based laser.

All this technology may not always pro­vide a meaningful advantage. This year theUS Marine Corps will start testing BostonDynamics’s four­legged beast of burden,the LS3. Its elaborate design keeps itupright even on rocky ground, and it isvery di�cult to knock over. But its petrolengine makes it as loud as a lawnmower.The Taliban have a much stealthier system,notes a former French army lieutenant.Their mules quietly eat grass.

A slippery slope to war?A larger worry is that countries with high­performance military robots may be moreinclined to launch attacks. Robots protectsoldiers and improve their odds of success.Using drones sidesteps the tricky politicsof putting boots on foreign soil. In the pasteight years drone strikes by America’s Cen­tral Intelligence Agency (CIA) have killedmore than 2,400 people in Pakistan, in­cluding 479 civilians, according to the Bu­reau for Investigative Journalism in Lon­don. Technological progress appears tohave contributed to an increase in the fre­quency of strikes. In 2005 CIA dronesstruck targets in Pakistan three times; lastyear there were 76 strikes there. Do armedrobots make killing too easy?

Not necessarily. When Mary Cum­mings, a former US Navy pilot, stopped �y­ing F­18 �ghter jets in 1997, there were novideo links between cockpits and com­mand centres, and even radio contact waspatchy at times. As a result, pilots oftenmade their own calls on whether or not tostrike. Today’s drones, blimps, unmannedboats and reconnaissance robots collectand transmit so much data, she says, thatWestern countries now practise �warfareby committee�. Government lawyers andothers in operation rooms monitor videofeeds from robots to call o� strikes that areillegal or would �look bad on CNN�, saysMs Cummings, who is now a robotics re­searcher at MIT. And unlike pilots at thescene, these remote observers are una�ect­ed by the physical toil of �ying a jet or the

adrenalin rush of combat.In March Britain’s Royal Artillery began

buying robotic missiles designed byMBDA, a French company. The Fire Shad­ow is a �loitering munition� capable oftravelling 100km, more than twice themaximum range of a traditional artilleryshell. It can circle in the sky for hours, usingsensors to track even a moving target. A hu­man operator, viewing a video feed, thenissues an instruction to attack, �y else­where to �nd a better target, or abort themission by destroying itself. But bypassingthe human operator to automate attackswould be, technologically, in the �realm offeasibility�, an MBDA spokesman says.

Could the �man in the loop� be re­moved from robotic weapons? The IsraelDefence Forces have installed �combatproven� robot machineguns along thecountry’s borders. When sensors detect anintruder, the barrel pivots to follow him. Ahuman soldier, watching the scene re­motely via a �bre­optic link, decideswhether or not to issue a warning (througha loudspeaker) or press the �re button. Therobot sentry, the Samson Remote WeaponStation, could function without human in­tervention, says David Ishai of Rafael, its Is­raeli manufacturer, based in Haifa. But, hesays, switching to automatic mode wouldbe a bad idea�and illegal to boot.

Traditional rules of engagement stipu­late that a human must decide if a weaponis to be �red. But this restriction is startingto come under pressure. Already, defenceplanners are considering whether a droneaircraft should be able to �re a weaponbased on its own analysis. In 2009 the au­thors of a US Air Force report suggestedthat humans will increasingly operate not�in the loop� but �on the loop�, monitoringarmed robots rather than fully controllingthem. Better arti�cial intelligence willeventually allow robots to �make lethalcombat decisions�, they wrote, providedlegal and ethical issues can be resolved.

A report on the matter issued by Brit­ain’s Ministry of Defence last year arguedthat if a drone’s control system takes ap­propriate account of the law on armedcon�icts (basically military necessity,humanity, proportionality and the ability

to distinguish between military targets andcivilians), then an autonomous strikecould meet legal norms. Testing and certi­fying such a system would be di�cult. Butthe authors concluded that �as technologymaturesðpolicymakers will need to beaware of the potential legal issues and takeadvice at a very early stage of any new sys­tem’s procurement cycle.�

Pressure will grow for armies to auto­mate their robots if only so machines canshoot before being shot, says Jürgen Alt­mann of the Technical University of Dort­mund, in Germany, and a founder of theInternational Committee for Robot ArmsControl, an advocacy group. Some robotweapons already operate without humanoperators to save precious seconds. An in­coming anti­ship missile detected even adozen miles away can be safely shot downonly by a robot, says Frank Biemans, headof sensing technologies for the Goalkeeperautomatic ship­defence cannons made byThales Nederland.

Admittedly, that involves a machine de­stroying another machine. But as humanoperators struggle to assimilate the infor­mation collected by robotic sensors, deci­sion­making by robots seems likely to in­crease. This might be a good thing, saysRonald Arkin, a roboticist at the GeorgiaInstitute of Technology, who is developing�ethics software� for armed robots. Bycrunching data from drone sensors andmilitary databases, it might be possible topredict, for example, that a strike from amissile could damage a nearby religiousbuilding. Clever software might be used tocall o� attacks as well as initiate them.

In the air, on land and at sea, military ro­bots are proliferating. But the revolution inmilitary robotics does have an Achillesheel, notes Emmanuel Go� of the Frenchair­force academy in Salon­de­Provence.As robots become more autonomous,identifying a human to hold accountablefor a bloody blunder will become very dif­�cult, he says. Should it be the robot’s pro­grammer, designer, manufacturer, humanoverseer or his superiors? It is hard to say.The backlash from a deadly and well­pub­licised mistake may be the only thing thatcan halt the rapid march of the robots. 7

Robots on the front line: iRobot’s SUGV; the Recon Robotics Scout XT Throwbot; and Rafael’s Samson Remote Weapon Station

The Economist Technology Quarterly June 2nd 2012 Acoustic micro�uidics 11

AT FIRST glance it appears to be a minus­cule marble spinning around its verti­

cal axis. Look closer, however, and you seea stationary spherical membrane of �uid,just 3 microns across. It is the stu� insidethe droplet that is rotating. This self­con­tained centrifuge has been created byblasting a droplet of �uid with sound.

It doesn’t just look cool. The manipula­tion of drops of liquid using sound, knownas acoustic micro�uidics, is spawning agrowing number of applications. Jona­than Cooper, of the University of Glasgow,thinks the technique could do away withbulky laboratory equipment, with all its�nicky moving parts. This, Dr Cooper sug­gests, could radically cut the cost of diag­nosing and treating diseases, or detectingharmful pathogens in hospitals, watersources or food­processing facilities.

Sounds, it turns out, are ideally suitedfor handling unruly �uids. By carefullycontrolling the way sound makes the tran­sition from a solid to a �uid medium it ispossible to pump, mix or heat up a �uid.With enough power, droplets can even bemade to vaporise or squirt spectacularlyo� the solid surface, like the ink in an inkjetprinter, but without the nozzle.

Sonic screwdrivingAll this is done by harnessing surfaceacoustic waves (SAWs). When these in­tense oscillations, travelling through a sol­id medium, come into contact with a dropof �uid lying on its surface, the drop ab­sorbs some of their energy. This energytends to dissipate rapidly, but when SAWsreach a frequency of around 10MHz in wa­ter­based �uids, they give rise to an unusu­al phenomenon called acoustic streaming.As the name suggests, instead of spreadingthroughout the �uid as humdrum com­pression waves, the SAWs cause some sec­tions of the liquid to �ow but not others,creating steady micro­currents.

By varying the angle and direction ofSAWs relative to the droplet, the internalstreams can be made to �ow in just aboutany direction. Aim the wave at the centreof the droplet, but at an angle, and the

forces generated will cause the droplet toroll. Sideswipe it, and you will get a whirl­pool which can be used to stir the contents.Drive the stream faster and the oppositehappens: liquid separates, as it would in acentrifuge. Vaporisation is achieved byramping up the SAWs’ energy, and jettingoccurs when a droplet is bombarded bywaves from all directions, and then has no­where to go but up.

The idea of getting SAWs to do usefulwork is not entirely new. They have longbeen employed in devices like mobilephones to �lter electric signals (which canbe done by converting electrical into me­chanical energy and back again). Then, in

1999, Achim Wixforth, a physicist at theUniversity of Augsburg, in Germany, dis­covered that when small quantities of liq­uid were placed in their path, they causedit to stir. By placing interleaved comb­likeelectrodes on the surface of lithium nio­bate, a material that is piezoelectric (and soexpands or contracts as a current is passedthrough it), Dr Wixforth was able to �ne­tune the waves to do his bidding.

He quickly realised that the techniquemight come in handy when mixing tinyquantities of di�erent liquids�in DNA mi­croarrays, for example. These are tiny glasschips dotted with thousands of micro­scopic wells where DNA fragments reactwith other chemicals in order to measurethe activity of genes or identify genetic va­riations. Using SAWs to mix the reagentswould speed up analysis.

This is much harder than it sounds. Thewells’ minute size makes mechanical stir­ring impractical. Simple vibrations do notdo the trick, while heat risks damaging thesample. Typically, scientists have no choicebut to await patiently for the di�usion toproceed at its natural pace, which can be

What a sound idea

Medical technology: Using sound waves to manipulate tiny droplets of �uidpromises to revolutionise medical diagnostics and improve drug delivery

1

12 Acoustic micro�uidics The Economist Technology Quarterly June 2nd 2012

2 painfully slow. Using SAWs, Dr Wixforthwas able to nudge them along. In 2000 acompany called Advalytix was spun out ofhis research (which, through a series of ac­quisitions, ended up as part of Danaher, abig maker of electronic sensors).

Besides mixing, Dr Wixforth alsoshowed how SAWs could channel �uidsas part of more complex tasks on so­calledmicro�uidic lab­on­a­chip devices. Oneexample is the polymerase chain reaction(PCR), a standard technique which em­ploys enzymes to churn out millions ofcopies of a DNA sequence and involves cy­cles of gentle heating and cooling. It can beachieved in half the time when SAWs areused to mix the �uids and guide themaround the device.

Dr Cooper and Julien Reboud, alsofrom the University of Glasgow, have nowdeveloped a way to carry out the entirePCR process�mixing, heating and cooling,as well as analysis�using a single piezo­electric source, without the need for sepa­rate heating elements or detection equip­ment. Better yet, �uid no longer needs tomove from one reaction chamber to thenext. Instead, the droplet is �rst spun untilthe cell membranes rupture in a processcalled lysis, releasing the DNA inside.Then, the waves’ frequency is tweakedand the sample blasted head­on in order toheat it. Cooling occurs through unaideddissipation, though this too is quicker withsmaller droplets, which lose energy morerapidly than bigger ones. Repeating the cy­cle drives PCR. Finally, SAWs are also usedto measure the change in density of the �u­id caused by the newly replicated DNA.

At £10 ($16) a pop, piezoelectric actua­tors are not costly. Still, using one acousticsource means that the devices can be madecheaper still. On the face of it, though, rely­ing on a single source limits the sorts ofwave that can be produced, and thereforethe types of task they can perform.

Dr Cooper and Dr Reboud found a clev­er way round this. They use phononic lat­tices�arrays of bumps and dimples em­bossed on a substrate�to alter the way inwhich SAWs reverberate. Moreover, sincethe resulting interference pattern also de­pends on the waves’ frequency, the samearray can concentrate SAWs in di�erentways at di�erent frequencies, working theliquid sample in di�erent ways.

The researchers goal is to build a PCR

system where a blood sample is placed ona disposable array made out of a cheapmaterial, such as paper, and inserted into ahand­held device equipped with the pi­ezoelectric actuator and other �xed ele­

ments. It also would make it easier to keepimpurities at bay, as there will be fewercomponents to be kept sterile. This wouldbe a boon for diagnosticians in developingcountries, where squeaky clean, well­equipped labs are a rarity. So far Dr Cooperand Dr Reboud have demonstrated thatsound­aided PCR can be used to detectPlasmodium falciparum, the parasitewhich causes malaria, in a blood sample.They are working on similar tests for tuber­culosis and sleeping sickness.

Sound is not just good at detecting dis­eases. It can also be used to help treat them.

Leslie Yeo and James Friend, who jointlyhead the Micro/Nanophysics ResearchLaboratory at RMIT University in Mel­bourne, are devising ways to use SAWs toimprove inhalers, which would allowdrugs and vaccines to be delivered to thelungs with much greater e�ciency.

Normal inhalers produce a wide rangeof droplet sizes. As a result much of thedrug ends up being wasted because manydroplets are too large to reach deep lung tis­sue, explains Dr Yeo. With SAWs it is possi­ble to tailor droplet size precisely, to as littleas a few microns�small enough to pene­trate all the way inside the lungs.

Another advantage of using SAWs isthat they can perform extremely delicatemanipulations. This can be crucial when

dealing with fragile biomolecules such asproteins, peptides, stem cells and DNA

molecules which are easily denatured ordamaged when vaporised. Having suc­cessfully tested an SAW asthma inhaler onsheep, Dr Yeo and Dr Friend are exploringa similar approach for vaccines. This ap­proach, like that of Dr Cooper and Dr Re­boud, is particularly well­suited for the de­veloping world. Even where injectablevaccines are available, inhalants are oftenmore desirable. Poor countries most inneed of vaccines tend to lack adequatesanitation. Needles are often the biggesttransmitters of disease, Dr Yeo says.

Dr Cooper points to other possibilitieso�ered by SAW devices’ low power re­quirements. Many parts of Africa, for in­stance, do not enjoy reliable access to elec­tricity. Dr Cooper is therefore looking athow mobile phones can be used to powerthese portable devices and, possibly, evendisplay test results. Since there are an esti­mated 500m mobile phones in Africa, andbecause devices plugged into mobileswould need no other equipment, manymore people could act as health workers,either for others in the same remote vil­lage, say, or just for their own family. With$100,000 from the Bill and Melinda GatesFoundation Dr Cooper is designing aphone­powered SAW device to diagnosemalaria. So far he has succeeded in using afully charged mobile to run ten tests.

You ain’t heard nothing yetFor all the recent developments, Dr Cooperpoints out, the �eld is still in its infancy.New ideas are cropping up all the time.Some researchers are putting SAWs towork in cell sorting, where a short acousticpulse can nudge di�erent kinds of cellsdown di�erent channels, and cell separa­tion, where droplets containing variouskinds of cell are spun to divide the cellsinto groups, as in a centrifuge.

For his part, Dr Wixforth, the �eld’s pio­neer, has recently come up with a way tosimulate a heartbeat, and the way it pumps�uid around a model of the circulatorysystem, using SAWs. He and his colleagueshope this will shed light on the way bloodand the molecules within it�for instanceglycoproteins, which are found in cellmembranes and play a role in interactionsbetween cells�behave as they coursethrough the human body. This, in turn,might help explain what causes some ofthem to malfunction, as they do in a rangeof diseases. After more than a decade, DrWixforth’s discovery continues to produceunexpected reverberations. 7

�Surface acoustic waves could be used to deliverdrugs and vaccines to the lungs more e�ciently.�

THE accident at the Fukushimanuclear power­station in Japan

last year, the worst since the explo­sion at Chernobyl in 1986, has ledmany countries to reconsider theircommitment to nuclear power. Ithas also drawn new attention to theenduring problem of dealing withnuclear waste. Around 270,000tonnes of high­level waste, mostlyspent fuel, are in temporary storagearound the world. Another 10,000tonnes of waste are added each year,according to the World Nuclear Asso­ciation, an industry body.

Should it be buried in undersea�ssures, stored underground or ma­nipulated to make it less harmful? Theabandonment in 2009 of a plan for ahuge storage site shielded by 300 me­tres of volcanic rock in Yucca Moun­tain, Nevada, gave the matter new ur­gency in America even beforeFukushima. The presidentially man­dated Blue Ribbon Commission onAmerica’s Nuclear Future, which de­livered its �nal report in January 2012,emphasised the �urgent� need for afresh waste­disposal strategy. Mean­while, new rules adopted by the EuropeanUnion last year require member countriesto draw up long­term plans for dealingwith their nuclear waste by 2015.

Dealing with nuclear waste does notjust mean preventing it from doing harmtoday, whether as the result of accident,malicious intent or natural disaster. It alsomeans ensuring that waste does not poi­son future generations�people who dis­cover nuclear waste in the far future maynot realise what it is�and doing so in a waythat is acceptable to today’s taxpayers.There are essentially three approaches:bide your time and wait for technology toimprove; bury the waste somewhere se­cure; or reprocess it into new fuel, greatlyreducing the amount of waste left over.

The nastiest power­station waste, high­level spent fuel, is usually cooled in poolsof water for at least �ve years and thenpacked away in thick concrete­and­steel�dry casks�, each the size of a small truck.Gradually weakened by radioactive decayand the resulting heat, a cask can hold backthe radiation for a few hundred years, per­haps even a few thousand. But a dry caskwould certainly not last 200,000 years,which is roughly how long it takes for thelongest­lived forms of plutonium to decayto harmlessness.

Storage and handling technology is stillimproving, however. SKB, a Swedish com­

The Economist Technology Quarterly June 2nd 2012 Inside story 13

STRO, it reduces radiation by�locking it up safely in minerals�,says Gavin Atkins, a spokesmanfor the government agency. Amer­ica’s Department of Energy plansto use the process to dispose of ra­dioactive powder in Idaho leftover from the extraction, decadesago, of uranium from spent fuel.

Future scientists will designeven better materials and process­es, which has led some to suggestthat the best thing to do with nuc­lear­waste materials is simply tostore them above ground and waitfor technology to improve. But thatwould push disposal costs on futuregenerations. America’s Blue RibbonCommission noted in a report thatthere is �no ethical basis� for doingthat. Moreover, something could gowrong in the meantime, says MickApted, a contingency­scenario ana­lyst for Intera, an engineering consul­tancy based in Denver, Colorado. AsFukushima showed, unburiedwaste is vulnerable to natural disas­

ters. Another risk is political instabil­ity: there could be a war or a revolution.

And instead of advancing, the expertiseneeded to handle nuclear waste might beneglected and lost. The upshot, Mr Aptedsays, is that the waste must be buriedsooner rather than later.

Journey to the centre of the EarthBut where? Forget about dropping nuclearwaste into a volcano: magma is nowherenear hot enough to melt it. Firing wasteinto the nuclear �reball of the sun, mean­while, would be astonishingly expensive,and a launch accident could spread radio­active material over a wide area. An ideasuggested early in the nuclear age was tobury waste in Antarctica’s ice sheets. Heat­ed by radioactive decay, the canisterswould melt the ice beneath them andslowly sink. But this plan was abandonedin the 1970s. Back then a handful of coun­tries, including France, dumped canisterscontaining small amounts of low­level nu­clear waste into the ocean. This conformedto �common sense at the time� and was es­sentially harmless, says Gérard Ouzou­nian, an o�cial at ANDRA, the agencycharged with disposing of France’s waste.

Such dumping was banned in 1993. Butit might make sense to bury waste deep inthe seabed. Using modern equipment forundersea oil­and­gas drilling, canisterscould be buried far from volcanic areas inholes plugged with seabed clay. If buried

Hot stu�

Nuclear waste: Researchers aredevising new ways to deal with thebyproducts of nuclear power. But it isnot just a technological problem

pany that stores nuclear waste, has devel­oped a new type of durable canistersheathed in two centimetres of copper, ametal that has survived as jewellery formillennia. The canisters are welded withan innovative �friction stir� technique thatsoftens surfaces to be joined withoutweakening them by melting. A nickel ma­terial called Alloy 22 also shows promise. Itcontains chromium to resist corrosion, andmolybdenum and tungsten for strength.

The Australian Nuclear Science andTechnology Organisation has devised aprocess using heat and high pressure tomake synthetic rock out of radioactivewaste and minerals including calcium, tita­nium and zirconium. Called SynrocAN­ 1

concrete­lined, zinc­coated drums. In 2005they began lining the top, bottom andsides of certain chambers with a �ve­centi­metre layer of gravel, held in place withwire mesh sprayed with concrete. Eachchamber’s gravel lining creates �a path ofleast resistance� to divert any wateraround it. The project’s leader, MiroslavKucerka of the Radioactive Waste Reposi­tory Authority, says it is a world �rst.

Finland and Sweden have also built un­derground laboratories to test geologic for­mations and packaging technologies. Theresults look good, says Claes Thegerström,the head of SKB, which operates the Swed­ish facility, the Aspo Hard Rock Laboratory.Repositories built at depths of half a kilo­metre or more in suitable rock would evensurvive the glaciers of a future ice age asthey remodel Europe’s landscape, he says.

Yet so far not a single permanent reposi­tory has been approved. Construction willbe expensive. But managing public opin­ion may pose a bigger problem. As an in­dustry joke puts it, the social science hasbecome more di�cult than the physicalscience. O�cials in France and Swedenreckon that building permits will be grant­ed within a few years. But hostile publicopinion can derail even advanced plans�more than $9 billion had been spent onAmerica’s Yucca Mountain repositorywhen it was cancelled. People living nearthe Carlsbad site for weapons­relatedwaste were, in contrast, consulted andcompensated. Sweden’s SKB chose the sitefor its proposed repository after a long con­sultation and several referendums.

A boring alternativeAmerica’s Blue Ribbon Commission con­cluded that a �new, consent­based ap­proach� was needed. It also called for re­search into an alternative form ofunderground burial�packing waste intoholes drilled several kilometres deep. Re­search into borehole disposal, as it isknown, is now taking o�, says FergusGibb, a geochemist at the University ofShe�eld in England.

As an extra precaution, he says, canis­ters could be designed so the hot wastemelts the adjacent rock. Once cooledand resolidi�ed, in as little as a few de­cades, the rock would form a tightseal. A workshop held last October atSandia National Laboratories in Al­buquerque, New Mexico, conclud­ed that borehole disposal wouldbe cheaper, more �exible and fast­er to implement than repositorydisposal. But the protection pro­

in a subduction fault, the movement ofone tectonic plate sliding below anotherwould push the canisters down into theEarth’s mantle at a rate of about ten centi­metres a year. America cut o� funding forseabed­disposal research in 1986, and oth­er countries followed suit. But some peo­ple have continued to develop the idea.

Permanent RadWaste Solutions, basedin California, has patented a steel, copperand lead container cleverly shaped towithstand pressure in a subduction fault. Itresembles a bowling pin the size of a car.As it is drawn into the subduction fault andthe pressure increases, it is closed up evermore tightly by pressure­compensatingseals, says designer Dean Engelhardt, a for­mer nuclear­weapons engineer at GeneralDynamics. The idea of seabed burial, how­ever, has failed to catch on, not least be­cause it would require renegotiation of in­ternational environmental laws.

Underground burial is regarded as asimpler and more attractive alternative.Salt deposits excavated 650 metres be­neath the desert near Carlsbad, New Mexi­co, already hold canisters with enough ra­dioactive waste from America’s weaponsprogrammes to �ll more than 28 Olympicswimming pools. Burying spent fuel istrickier. Unlike weapons waste, power­sta­tion waste generates lots of heat whichmight, in the long term, destabilise geologi­cal formations. No country has yet built apermanent burial site for spent fuel. Butthe technology for building, monitoringand eventually sealing repositories hasbeen worked out, says Charles McCombieof the European Repository DevelopmentOrganisation, a 14­country working groupbased in Baden, Switzerland.

To �nd out how a repository carved outof argillite, a sedimentary rock, wouldhold up, French researchers built an under­ground laboratory 490 metres down nearthe northeastern town of Bure. Heat andradioactivity testing show that a reposi­tory could be built in about ten years andthen safely �lled and monitored for at leastanother hundred before being sealed, saysMr Ouzounian.

Any permanent repository would bebuilt in a dry geological medium. Even so,water could seep in through cracks createdby an earthquake, say, or a shifting landmass, allowing radioactive material to getout. To prevent that, Czech engineersworking with a German consultancycalled DBE Technology have built a �hy­draulic cage� to keep water out of a formerlimestone mine near the Czech town ofLitomerice that holds radioactive waste in

vided by rock­melting would probablyhave to be abandoned as a concession topublic opinion, says Dr Gibb. Tell peoplethat nuclear waste will be buried in con­centrations hot enough to melt rock, hesays, and some will worry that it might ex­plode like a volcano.

Some say the stu� should not be per­manently buried at all. �Spent fuel, in gen­eral, is not waste,� says Andrey Bukhovt­sev of Rosatom, Russia’s state nucleargiant, re�ecting a widespread view in thatcountry. Reprocessed waste yields usableuranium fuel, along with �an alchemist’sdream� of other elements useful in indus­try and medicine, says Sherrell Greene,who retired in late 2011 as head of nucleartechnology at the United States Depart­ment of Energy’s Oak Ridge National Lab­oratory in Tennessee, and now works as aconsultant. France and Japan are leaders inreprocessing. It is expensive, but reduceswaste and provides a secure fuel supply.

The trouble is that reprocessing alsoseparates out pure plutonium, which canbe used to make nuclear weapons. For thatreason, President Jimmy Carter halted re­processing in America in 1977 to slow thespread of bombmaking expertise. Someother countries followed suit.

In 2009 scientists at the Oak Ridge Na­tional Laboratory, including Dr Greene,demonstrated a new reprocessing tech­nique called UREX that does not producepure plutonium. But it will take a decade todevise the chemical and mechanical stepsto commercialise their method. And somewaste will always be left. Rather than justburying the problem, it may be time tostart burying the waste. 7

�So far not a single permanent repository fornuclear waste has been approved.�

2

14 Inside story The Economist Technology Quarterly June 2nd 2012

The Economist Technology Quarterly June 2nd 2012 Open­source medical devices 15

SMART pumps deliver drugs perfectlydosed for individual patients. Easy­to­

use de�brillators can bring heart­attackvictims back from the brink of death. Pace­makers and arti�cial hearts keep peoplealive by ensuring that blood is pumpedsmoothly around their bodies. Medical de­vices are a wonder of the modern age.

As these devices have become more ca­pable, however, they have also becomemore complex. More than half the medicaldevices sold in America (the world’s larg­est health­care market) rely on software,and often lots of it. The software in a pace­maker may require over 80,000 lines ofcode, a drug­infusion pump 170,000 linesand an MRI (magnetic­resonance imaging)scanner more than 7m lines.

This growing reliance on softwarecauses problems that are familiar to any­one who has ever used a computer: bugs,crashes and vulnerability to digital attacks.Researchers at the University of Patras inGreece found that one in three of all soft­ware­based medical devices sold in Amer­ica between 1999 and 2005 had been re­called for software failures. Kevin Fu, acomputer science professor at the Univer­sity of Massachusetts, calculates that such

recalls have a�ected over 1.5m individualdevices since 2002. In April researchers atMcAfee, a computer­security �rm, saidthey had found a way to get an implantedinsulin pump to deliver 45 days’ worth ofinsulin in one go. And in 2008 Dr Fu andhis colleagues published a paper detailingthe remote, wireless reprogramming of animplantable de�brillator.

When software in a medical devicemalfunctions, the consequences can be farmore serious than just having to rebootyour PC. During the 1980s a bug in the soft­ware of Therac­25 radiotherapy machinescaused massive overdoses of radiation tobe delivered to several patients, killing atleast �ve. America’s Food and Drug Ad­ministration (FDA) has linked problemswith drug­infusion pumps to nearly20,000 serious injuries and over 700deaths between 2005 and 2009. Softwareerrors were the most frequently cited pro­blem. If buggy code causes a pump to in­terpret a single keystroke multiple times,for example, it could deliver an overdose.

In addition to accidental malfunctions,wireless and networked medical devicesare also vulnerable to attacks by malicioushackers. In the 2008 paper Dr Fu and hiscolleagues showed how an implantablecardioverter de�brillator could be remote­ly reprogrammed either to withhold thera­py when it is needed or to deliver unneces­sary shocks. Dr Fu says that when it comesto testing their software, device manufac­turers lack the safety culture found in other

high­risk industries such as avionics, andare failing to keep up with the latest ad­vances in software engineering. Insup Lee,professor of computer science at the Uni­versity of Pennsylvania, agrees. �Manymanufacturers do not have the expertise orthe willingness to utilise new tools beingdeveloped in computer science,� he says.

Just how bad it is, though, no oneknows for sure. The software used in thevast majority of medical devices is closedand proprietary. This prevents commercialrivals from copying each other’s code orchecking for potential patent infringe­ments. It also makes it harder for securityresearchers to expose �aws. The FDA,which could demand to see the sourcecode for every device it approves, does notroutinely do so, but instead leaves it tomanufacturers to validate their own soft­ware. Two years ago it o�ered free �staticanalysis� software testing to infusion­pump manufacturers in the hope of reduc­ing injuries and deaths. But no manufac­turer has yet taken the FDA up on its o�er.

Open to scrutinyFrustrated by the lack of co­operation frommanufacturers, some academics nowwant to reinvent the medical­device indus­try from the ground up, using open­sourcetechniques. In open­source systems, thesource code is freely shared and can beviewed and modi�ed by anyone whowants to see how it works or build an im­proved version of it. Exposing a design to

When code can kill or cure

Medical technology: Applying the�open source� model to the design ofmedical devices promises to increasesafety and spur innovation

1

16 Open­source medical devices The Economist Technology Quarterly June 2nd 2012

2 many hands and eyes, the theory goes, re­sults in safer products. This seems to be thecase for desktop software, where bugs andsecurity �aws in open­source applicationsare typically �xed much more quickly thanthose in commercial programs.

The Generic Infusion Pump project, ajoint e�ort between the University ofPennsylvania and the FDA, is taking thesetroublesome devices back to basics. The re­searchers began not by building a deviceor writing code but by imagining every­thing that could possibly go wrong with adrug­infusion pump. Manufacturers wereasked to help, and several did so, includingvTitan, a start­up based in America and In­dia. �For a new manufacturer, it’s a greathead start,� says Peri Kasthuri, vTitan’s co­founder. By working together on an open­source platform, manufacturers can buildsafer products for everyone, while still re­taining the ability to add extra features todi�erentiate themselves from their rivals.

Mathematical models of existing andnew pump designs were tested against thepossible risks, and the best­performingmodels were used to generate code, whichwas installed on a second­hand infusionpump bought online for $20. �My dream�,says Dave Arney, a researcher on the pro­ject, �is that a hospital will eventually beable to print out an infusion pump using arapid prototyping machine, downloadopen­source software to it and have a de­vice running within hours.�

Equally ambitious is the Open SourceMedical Device initiative at the Universityof Wisconsin­Madison. Two medical phys­icists, Rock Mackie and Surendra Prajapati,are designing a machine to combine radio­therapy with high resolution computed to­mography (CT) and positron­emission to­mography (PET) scanning. Their aim is tosupply, at zero cost, everything necessaryto build the device from scratch, includinghardware speci�cations, source code, as­sembly instructions, suggested parts�andeven recommendations on where to buythem and how much to pay. The machineshould cost about a quarter as much as acommercial scanner, making it attractivein the developing world, says Dr Prajapati.�Existing devices are expensive both tobuy and maintain,� he says, whereas theopen­source model is more sustainable. �Ifyou can build it yourself, you can �x ityourself when something breaks.�

Open­source devices are also to befound literally at the cutting edge of medi­cal science. An open­source surgical robotcalled Raven, designed at the University ofWashington in Seattle, provides an a�ord­

able platform for researchers around theworld to experiment with new techniquesand technologies for robotic surgery.

All these open­source systems addressvery di�erent problems in medical sci­ence, but they have one thing in common:all are currently prohibited for use on livehuman patients. To be used in a clinical set­ting, open­source devices must �rst un­dergo the same expensive and lengthyFDA approval processes as any other med­ical device. FDA regulations do not yet re­quire software to be analysed for bugs, butthey do insist on a rigorous paper trail de­tailing its development. This is not alwaysa good �t with the collaborative and ofteninformal nature of open­source coding.

The high cost of navigating the regula­tory regime has forced some not­for­pro�t,open­source projects to alter their businessmodels. �In the 1990s we developed an ex­cellent radiation­therapy treatment­plan­ning system and tried to give it away to oth­er clinics,� says Dr Mackie. �But when wewere told by the FDA that we should getour software approved, the hospital wasn’twilling to fund it.� He formed a spin­o��rm speci�cally to get FDA approval. Ittook four years and cost millions of dol­lars. The software was subsequently soldas a traditional, closed­source product.

Others are skirting America’s regula­tory system altogether. The Raven surgicalrobot is intended for research use on ani­mals and cadavers, while the Open SourceMedical Device scanner will be largeenough only to accommodate rats and rab­bits. However, says Dr Mackie, there is

nothing to stop anyone taking the designand putting it through a regulatory processin another country. �It may even happenthat the device will be used on humans inparts of the world where strict regulationdoes not exist,� he says. �We would hopethat if it is used in such a way, it will be wellenough designed not to hurt anybody.�

Changing the rulesThe FDA is gradually embracing openness.The Medical Device Plug­and­Play Interop­erability Program, a $10m initiative fundedby the National Institutes of Health withthe support of the FDA, is working to setopen standards for interconnecting de­vices from di�erent manufacturers. Thiswould mean that, say, a blood­pressurecu� could instruct a drug pump to stop de­livering medication if it sensed that a pa­tient was su�ering an adverse reaction.

More intriguing still is the Medical De­vice Co­ordination Framework being de­veloped by John Hatcli� at Kansas StateUniversity. Its aim is to build an open­source hardware platform including ele­ments common to many medical devices,such as displays, buttons, processors andnetwork interfaces, and the software torun them. By connecting di�erent sensorsor actuators, this generic core could thenbe made into dozens of di�erent medicaldevices, with the relevant functions pro­grammed as downloadable �apps�.

Eventually, medical devices mightevolve into collections of specialised (andpossibly proprietary) accessories, with theprimary computing and safety featuresmanaged by an open­source hub. The FDA

is working with Dr Hatcli� to develop pro­cesses for creating and validating safety­critical medical apps.

In the meantime, there are moves afootto improve the overall security and reli­ability of software in medical devices.America’s National Institute of Standardsand Technology has just recommendedthat a single agency, probably the FDA,should be responsible for approving andtracking cybersecurity in medical devices,and the FDA is re­evaluating its ability tocope with the growing use of software.

Such changes cannot happen too soon.�When a plane falls out of the sky, peoplenotice,� says Dr Fu. �But when one or twopeople are hurt by a medical device, oreven if hundreds are hurt in di�erent partsof the country, nobody notices.� Withmore complex devices, more active hack­ers and more inquisitive patients, openingup the hidden heart of medical technologymakes a great deal of sense. 7

�By working together on an open­source platform,manufacturers can make safer products�

The Economist Technology Quarterly June 2nd 2012 Brain scan 17

�SPACE travel is the only technologythat is more dangerous and more

expensive now than it was in its �rst year,�says Burt Rutan, an aerospace engineerand advocate of private space�ight. �Fiftyyears after Yuri Gagarin, the space shuttleended up being more dangerous and moreexpensive to �y than those �rst throw­away rockets, even though large portionsof it were reusable. It’s absurd.�

As NASA, America’s space agency,comes to terms with the shuttle’s retire­ment, Mr Rutan is at the vanguard of amovement to reassert America’s domi­nance in space using commercial space­craft, and in the process open it up to thegeneral public. In 2004 a reusable space­plane he designed called SpaceShipOnecompleted the �rst manned private space­�ights, winning the $10m Ansari X prize.Scaled Composites, the aerospace com­pany Mr Rutan founded, is now devel­oping a larger SpaceShipTwo craft forVirgin Galactic, a space­tourism company.And Mr Rutan recently unveiled his bold­est design yet: what would be the largestaircraft ever, dubbed Stratolaunch, whichwould carry and launch a rocket capableof reaching orbit.

He predicts that within a dozen yearsprivate spacelines will have �own morethan 100,000 people outside the atmo­sphere. �Every one of those people willhave paid for their seat�and they won’t betold what to do by an employer or thegovernment,� he says. �This volume willcreate opportunities for people to come upwith reasons for �ying in space other thanit’s a fun, roller­coaster ride. We didn’tknow the importance of home computersbefore the internet. We had them mostlyfor fun, then the internet came along andwas enabled by all the PCs out there.�

Mr Rutan believes that �rms investingin private space­travel might see returnsrivalling those of today’s internet giants.�Once the research is done, the directoperating costs of �ying a routine Space­ShipTwo �ight will be a small fraction ofthe ticket price,� he says. �Every elementof this new industry will be extremelypro�table.� In the future a planned Space­ShipThree might, for example, open upsuborbital intercontinental travel, reduc­ing the travel time from London to Sydney

to just a couple of hours.Whether or not the analogy between

space�ight and the internet is apt, the twoare linked in one sense: much of the mon­ey for today’s private space companies hascome from starry­eyed computer entrepre­neurs. Paul Allen, co­founder of Microsoft,bankrolled both SpaceShipOne and Strato­launch. Elon Musk, who made his fortunefrom PayPal, went on to form SpaceX, the�rst private company to reach orbit andanother partner in the Stratolaunch pro­ject. Amazon’s founder, Je� Bezos, hasBlue Origin, and Je� Greason, a formerIntel manager, has XCOR Aerospace, bothstart­ups dedicated to increasing publicaccess to space with low­cost launches.

As the private sector steps up, Mr Rutansees the state’s role in space graduallyshrinking. In March, he notes happily,America’s Congress extended until at least2015 a restriction preventing the FederalAviation Administration, which regulatesaviation, from enacting safety regulationsfor commercial space travel. �If spaceshipsare regulated like commercial airliners, itwill probably never happen,� he says.

Small is beautifulNot surprisingly, Mr Rutan is no fan ofNASA, with its high costs and huge size. Hedisapproves of its unmanned programmebecause exploration with robotic vehicleshas come to be seen as an alternative tosending humans. NASA’s managementculture, he says, meant it was unable tosolve the space shuttle’s safety problemsand cost overruns. The sprawling shuttleprogramme (�a dismal failure�, accordingto Mr Rutan) involved up to 12,000 work­ers and eventually cost over $200 billion.By contrast, Scaled Composites developedSpaceShipOne, including its rocket motor,electronics, control systems, test facilitiesand a simulator using no more than 65people, and for a modest $25m.

Although SpaceShipOne carried justone astronaut, had no room for cargo andachieved only suborbital �ight (the shuttlecould reach low Earth orbit), Mr Rutan’sachievement was remarkable. It willprobably be the last time that a new space­craft is essentially the vision of a singleperson. His design conformed closely topublic expectations of how a spaceshipshould look: graceful space­age curves,elegant wings and futuristic portholes. �Idesigned the shape, the systems, the land­ing gear and the con�guration,� he says.�Knowing that doing things by committeealways takes a long time, I made decisionson the spot rather than analyse and study

A maverick in �ight

Burt Rutan, a pioneering andunconventional aerospace engineer,has made a career of doing whatother people say is impossible

1

18 Brain scan The Economist Technology Quarterly June 2nd 2012

and get other opinions.�SpaceShipOne relied on two key tech­

nologies. The �rst was the use of a jet­powered carrier aircraft, called WhiteKnight, to launch from high in the atmo­sphere. This meant that SpaceShipOnecould take o� from a normal airstrip rath­er than an expensive launch pad, requiredmuch less fuel than if it had started atground level, and could glide to safety ifanything went wrong. The second in­novation was a new way to re­enter theatmosphere, rotating the entire tail up­wards to provide self­correcting �feath­ered� drag, like a falling shuttlecock. �I hadsome of the best aerodynamicists comeout of academia to tell me the featheringwouldn’t work,� Mr Rutan recalls. �But Ithrive on things that are risky, knowingthat if they’re not risky, there’s no chanceof having a breakthrough.�

That might make Mr Rutan sound likesomething of a daredevil. Yet of the 46planes and one spacecraft he has designedover four decades, none of his piloted testaircraft have ever crashed. (Three ScaledComposites workers were killed in aground­level explosion in 2007, however,while working on Virgin Galactic’s Space­ShipTwo). Mr Rutan also had a close shavein 1986, when the wings of his high­endur­ance Voyager aeroplane, piloted by hisbrother Dick, scraped badly on the run­way during take­o�. But Voyager made italoft successfully and, nine days later,became the �rst aircraft to circumnavigatethe globe without stopping or refuelling.

Voyager is credited for demonstratingthe strength and reliability of lightweightcomposite materials in aviation, pavingthe way for modern �plastic planes� suchas the largely carbon �bre­reinforcedplastic Boeing 787. �We used the bestmaterials we could a�ord at the time,�says Mr Rutan. �I felt that if I could pullthat o�, we would have the credibility of acompany that was technically sophisticat­ed. I didn’t think of it as something thatwould lead to all­composite airliners.�

Having built model aircraft as a child,Mr Rutan studied aeronautical engineer­ing at California Polytechnic State Univer­sity. He then spent many years creatingpropeller aircraft and small jets, seatingbetween two and eight passengers, for thegeneral aviation market. He quicklygained a reputation for designing elegant,quirky aircraft sporting canards (smallforward wings), twin booms and angledwinglets. In an industry that traditionallyvalues function over form, Mr Rutan’splanes deftly combined the two. �I choose

attractive designs over ugly ones all thetime,� he says. �To my mind, things thatare e�cient tend to be beautiful, like long,slender, smooth wings. And it’s usuallyjusti�ed by them being lighter or easier tobuild, or having better performance.�

Some of his more outlandish projectshave included drones for the Pentagon,40­metre wind­turbine blades, a �mastwing� for the 1988 America’s Cup­winningyacht, a composite body for GM’s 1991Ultralite concept car, and an odd, asym­metrical plane called the Boomerang. Butmost ambitious of all is Stratolaunch.Funded by Mr Allen, it will recycle twosecond­hand Boeing 747 jets into a hugetwin­bodied aircraft, weighing over 500tonnes and with a record­breaking 116­metre wingspan. It will carry and launch amulti­stage SpaceX rocket that, it is hoped,will eventually carry humans into orbit.

Since his retirement, Mr Rutan is nolonger responsible for this monsterplane’s e�cacy. However, he still works asa consultant at Scaled Composites, wherehe recently delivered what could be hismost telling criticism of the giant aircraft.�I generally don’t go to meetings becausebeing candid on things I don’t like is likethrowing a monkey wrench into the ma­chinery,� he says. �But when I saw thelatest Stratolaunch drawings, I beat themup some, because it was just soðugly.�

One of Mr Rutan’s last designs beforeretiring from Scaled Composites last yearwas for that science­�ction staple, a �yingcar. The BiPod is a sleek, twin­bodiedhybrid­electric vehicle powered by twomotorcycle engines, two electric motorsand a bank of lithium­ion batteries. Eachboom houses a single­seater cockpit, oneequipped with a joystick for �ying, theother with a steering wheel for driving. Toconvert from air to ground transportation,the wings are removed by hand andstowed between the pods, making theBiPod thin enough to squeeze into a one­car garage. Although a BiPod prototype�ew the day before Mr Rutan retired, theproject is now on hold at Scaled Compos­ites, awaiting the money to develop itfurther and take it into production.

Mr Rutan no longer has the resourcesof a spacecraft manufacturer at his dis­posal, but he continues to design planes.His latest idea is for an aircraft that is part­seaplane, part­wingship, able to skime�ciently from lake to lake near his retire­ment home in Idaho. �I learned aboutwingships on a trip to Russia for the Penta­gon in 1993,� he says. �They direct airunder the wing so at low speeds they have

almost no drag. I’ll probably start buildingsomething in a month or so, just for fun.�

Thanks to the ultra­e�cient Voyager,the hybrid­electric BiPod and the fact thathe is in the process of building a 31­acrephotovoltaic solar farm in the Californiadesert, Mr Rutan is occasionally mistakenfor an environmentalist. It is an accusationhe is quick to deny. �I drove an electric carfor seven years because of its advancedtechnology, not because I have any con­cerns about energy resources,� he says. �Ihave none at all. And when environ­mentalists say that global warming isdangerous, unprecedented and that we’llhave a tipping point for atmosphericcarbon dioxide, it’s just nonsense.�

Doing his own thingHe also believes that he has solved theKennedy assassination and that ancientEgyptians built pyramids and monumentsby casting solid stone in moulds. As mightbe expected, Mr Rutan is used to scepti­cism. �At various times over 20 years, I didpreliminary designs for aircraft like theStratolaunch. For that whole time I wasencouraging us to do something thatalmost everyone else felt you could notdo,� he says. �But you never run intobreakthroughs when you say, ‘You can’tdo that.’ You run into them when you’vefound something that doesn’t make senseand you �nd a way to make it work. Abreakthrough always starts with non­sense.� It is his willingness to follow hisown instincts towards unorthodox sol­utions that had made Mr Rutan such asuccessful innovator. He doesn’t care whatother people think, and has always justgone his own way regardless. 7

�SpaceShipOne will probably be the last time that anew spacecraft is the vision of a single person.�

O�er to readersReprints of Technology Quarterly are availablefrom the Rights and Syndication Department.A minimum order of �ve copies is required.

Corporate o�erCustomisation options on corporate orders of100 or more are available. Please contact us todiscuss your requirements.For more information on how to order specialreports, reprints or any queries you may haveplease contact:

The Rights and Syndication DepartmentThe Economist26 Red Lion SquareLondon WC1R 4HQTel +44 (0)20 7576 8148Fax +44 (0)20 7576 8492e­mail: rights@economist.comwww.economist.com.rights

2