I AM standing in a cold north London workshop looking expectantly at a bizarre metal and plastic contraption. An acrid smell drifts from the machine as a length of plastic is drawn into a barrel at its centre and heated up. The molten plastic squirts from a nozzle onto a platform moving beneath it, drawing a pattern. The nozzle also moves up and down to build the design upwards like an expert cake icer.

Over the next few minutes, this 鈥淢akerBot鈥� will do something I can only dream of doing: it will create a spare part of itself as an insurance against future mishaps. Staring at the Heath Robinson-style kit before me, it is hard to believe that it 鈥� and a few hundred other devices 鈥� are paving the way to an era of desktop machines that can make just about anything, including copies of themselves.

It could be a revolutionary age. MakerBot is one of a range of desktop manufacturing plants being developed by researchers and hobbyists around the world. Their goal is to create a machine that is able to fix itself and, ultimately, to replicate.

To find out how close we are to that goal, I have come to the London Hackspace, a communal workshop where Russ Garrett, a software developer by day, keeps his MakerBot. Like 900 other enthusiasts, Garrett bought a mail-order kit from MakerBot Industries of New York for $750, and built the machine himself.

MakerBot and most of its kin are essentially a cut-price reinvention of the 3D printer. While professional machines still cost upwards of tens of thousands of dollars, a coalition of academics and tinkerers has created versions that do much the same thing for much less. Anyone with a few hundred dollars and some spare time can build their own 3D printer from a set of plans distributed free on the internet.

鈥淎nyone with a few hundred dollars and some spare time can build a 3D printer from a set of free plans鈥�

These machines can build any number of things, including everything from coat hooks to ipod docking stations. One MakerBot owner even made his engagement ring with it. But key to MakerBot鈥檚 popularity is its ability to make its own spares. Every kit contains a handful of parts made by other MakerBots, linking them in a mechanical family tree.

The MakerBot lineage is descended from RepRap (see 鈥淭he replicant鈥�) 鈥� the first machine designed to replicate parts of itself and brainchild of , a mechanical engineer at the University of Bath, UK. In 2006 he started the project with two goals: to create a 3D printer that anyone could make and use, and to make it capable of self-replicating. Most importantly, it would have an open-source design to encourage anyone to modify and improve it.

At the moment, RepRap can build about half of its own parts, including joints and casings. Some components, such as steel rods and microprocessors, are beyond its capabilities as yet. Still, Bowyer鈥檚 mechanical progeny reached a major milestone in November 2008, when Canadian Wade Bortz announced he had used his RepRap to create all the parts of a replica that it was possible to print 鈥� the first time this had been done 鈥渋n the wild鈥� outside Bowyer鈥檚 lab. It was sold online a few months later for a case of beer.

Bowyer鈥檚 first design, called Darwin, has since been replaced by Mendel, which is smaller and more reliable. 鈥淢endel can, if you discount nuts and bolts, print 50 per cent of the machine鈥檚 parts in under three days,鈥� says Bowyer. Mendel can make about the same proportion of its own parts as Darwin, but Mendel is a simpler, smaller and more reliable machine. It can also make much larger things than Darwin can.

Since then, tens of others have made mothers out of their machines, sometimes selling their offspring for hundreds of dollars to other enthusiasts keen to get a machine of their own. This has led to a veritable ecosystem of RepRap-type machines 鈥� an estimated 3000 exist 鈥� and while Bowyer is now focused mainly on making Mendel more robust and user-friendly, the RepRaps in the wild have begun evolving into different forms.

While exploring the , I come across one with the potential to be more self-replicating than any before, and it is provoking some excited comments. The poster, Frank Davies, based in Houston, Texas, is the proud owner of a RepRap ingeniously built using parts salvaged from a dot matrix printer and a Xerox photocopying machine, and he is now working on making his RepRap totally printable.

NASA engineer鈥檚 double life

Davies, by day a NASA engineer on the space shuttle programme, is effectively replacing RepRap鈥檚 skeleton with one of his own making. In place of the tracks along which the print nozzle glides are plastic concertina-like mechanisms called Sarrus linkages, originally used to ensure steam pistons moved in straight lines in an era when reliably straight rods weren鈥檛 available. Two perpendicular hinges connect the moving parts such that they can move only along the remaining, unrestricted axis.

鈥淎nother member of the community mentioned this linkage, so I Googled it and ran with the idea,鈥� says Davies. The result is a prototype whose platform moves in two directions . Davies is working on adding the third axis and print head to make a species of RepRap able to print an unprecedented amount of itself. 鈥淚f it goes well, it should be done in a few months,鈥� he says. 鈥淪ince I鈥檝e got a printer that can make arbitrarily shaped objects, it鈥檚 not hard to try new things out.鈥�

Other attempts to make more parts printable include replacing the machine鈥檚 rubber belts with printed rack-and-pinion gears. But there is a limit to how much of a machine can be made with plastic alone. What we need is a machine that can create parts made from a number of different materials.

The search leads me to Neil Gershenfeld, head of the Center for Bits and Atoms at the Massachusetts Institute of Technology. Gershenfeld is busy promoting FabLabs: rooms the size of a squash court packed with all the equipment necessary to take any design and make working products at a quality to match that of a professionally commissioned prototype. He has been spreading the word 鈥� so far there are FabLabs in Afghanistan, the Netherlands, Costa Rica, Ghana, the UK, Kenya, and the South Bronx in New York. In the process he has realised that what FabLabs really need is to be able to make themselves. 鈥淭he tools will have really succeeded when they can do that,鈥� he says.

In a year or two, FabLabs will simply be made inside existing ones, Gershenfeld promises. 鈥淲e鈥檒l still buy some components, like microcontrollers and stepper motors, but we鈥檒l make everything else.鈥�

A machine called MultiFab, created by recent MIT graduate Ilan Moyer, backs up that claim. Like RepRap, it is made from parts and materials costing just $400. It too can print plastic, but it can also wield milling and cutting toolheads to carve shapes in wax, cut vinyl, mill light plastic and wood, and carve out the conductive traces of custom circuit boards. The first thing the completed machine did was to carve out a circuit board to replace one of its own.

Moyer has experimented with using it to perform sequential operations 鈥� printing a structure, and then putting the finishing touches to its shape with a milling tool 鈥� and plans to add a laser cutting head. 鈥淓ventually I want to be able to put a FabLab in a briefcase,鈥� he says.

Still, ingenious as these machines are, they merely churn out piles of parts. What about assembly? A heap of plastic and metal is not a machine, just as you don鈥檛 have much in common with a pile of flesh and bones.

Greg Chirikjian, a roboticist at Johns Hopkins University in Baltimore, Maryland, agrees. 鈥淲hen a prototype only makes parts, the machine that made those parts wasn鈥檛 reproduced,鈥� he says. A true self-replicator must handle both fabrication and assembly. Chirikjian and his colleague Matt Moses are aiming to achieve this with a kind of Lego set that doesn鈥檛 need anyone to play with it.

The pair have already demonstrated key parts of such a system, using around 100 plastic blocks. Although it cannot yet fabricate these blocks itself, the machine is able to move in 3D to pick up and bind them into larger structures. Moses is currently working on having it make a complete replica of its own structure using Lego-like bricks, though the machine still relies on conventional motors 鈥� which have to be installed by hand 鈥� to drive its activity.

The blocks are simple rectangular plates with two threaded holes and four cone-shaped connectors, shot through with metal wiring so they can be used to create electronic circuits if need be. The machine uses a screwdriver-like tool to pick up, stack or secure blocks together. A line of basic blocks assembled together can act as a slide along which other components 鈥� made from groups of their siblings 鈥� can move driven by an attached motor.

The blocks are made from polyurethane, and the processes needed to make them could be mastered by a desktop machine, says Chirikjian. The machine could squirt a substance such as silicone over a spare part of itself to make a mould, before pumping in polyurethane to reproduce the original. The pair have shown that gaps left for wiring can be filled by a RepRap head modified to deposit a low-melting-point alloy. With the addition of a few magnets, Moses has now made a functional, if weak, motor.

This looks really promising: desktop machines capable of making complex electrical components should in theory be able to create copies of their own electronic brains. Hod Lipson, a robotics researcher at Cornell University in Ithaca, New York, says much more is possible. His students run a RepRap-like 3D printer project called Fab@Home, whose design boasts two syringes allowing it to squirt out two materials at a time. Researchers in Lipson鈥檚 lab are using a similar multi-barrelled approach to collapse printing and assembly into a single step, by printing fully functional modules that do not need to be assembled.

鈥淲e鈥檙e at the transition from printing parts to printing systems,鈥� says Lipson, 鈥渁nd working hard on printing a complete robot.鈥� The team will know when that milestone has been reached, he says, on the day the thing they print walks, or more likely crawls, from the printer under its own power.

So far the team has managed to print out working versions of the major components of a robot: electronic muscle actuators, relays, transistors, batteries and circuitry. Electrodes and wiring, for example, can be made using silver particles mixed into silicone, while the bulk of a battery can be made using a paste of zinc particles and liquid electrolyte. A robotic arm stands by to swap syringe cartridges as required.

Although the components are far inferior to off-the-shelf parts, Lipson is confident that his team need only find the right materials. His goal is to find a relatively small palette that can be mixed to create a wide range of components. 鈥淛ust as red, green and blue can be used to print full colour, this will be very powerful.鈥�

鈥淭he goal is to find a small palette of materials that can be mixed to create a much wider range鈥�

Lipson鈥檚 team still has a mountain to climb before it can print the sort of microchips that control his machines, and besides that, replicating machines have a fundamental limitation, he says: 鈥淭he thing you print can鈥檛 be as accurate as the thing you printed it with.鈥� Though the machines鈥� offspring may be perfectly functional for several generations, tiny errors will gradually accumulate so parts will periodically need to be replaced. And they will never be as precise as the originals.

This by no means sounds the death knell for self-replicating machines, says Bowyer. Every living thing can be seen as a self-replicating machine that relies heavily on components and assistance from others, he says, so why should machines be any different? Plants, mosquitoes and viruses are all accepted as self-replicating but rely heavily on other species to achieve it. It would be churlish to expect replicators to be any different.

A robot toddler is getting an extreme makeover to make it even more like a real child (see video above). Called iCub, it has been built to road-test theories about how flesh-and-blood children think, learn and develop.

Designed by a consortium of European universities, the iCub is modelled on a 2-year-old. But its hands are currently the size of an 8-year-old鈥檚 because it was too hard to make fully functional mechanical hands any smaller.

鈥淲e now have a hand that looks like that of an 18-month-old child,鈥� says , an engineer at the Italian Institute of Technology in Genoa. 鈥淚t鈥檚 got almost all the motion we have in the human hand.鈥�

The robot is also to get a new pair of legs. They are lighter and springier to make them robust enough to handle the bumps and scrapes that real children shrug off but that can all too easily damage a robot, whose gearbox and motor-packed joints can鈥檛 flex to reduce the force of an impact.

That design tweak will also make the robot safer for any humans around, something that engineers are increasingly thinking about as robots become advanced enough to mingle with us.

Gallery: Weightlifting ants and crafty crows

See the winners and runners-up in a competition held by the UK鈥檚 showcasing the best images of the latest research.

Gallery: Weightlifting ants and crafty crows

Innovation is our regular column that highlights emerging technological ideas and asks where they may lead.

Last week New 女生小视频 pondered the fragility of digital data stores over the very long term, in the event of a civilisation-wide calamity. But anyone worried about civilisation鈥檚 chances would do well to look to their own data stores first.

Most of us today are blithely heading for our own personal data disasters. We generate and store vast volumes of information, but few of us really look after it.

鈥淏enign neglect鈥� is how of Microsoft Research Silicon Valley in Mountain View, California, describes the way most people treat their personal archives of digital material. It鈥檚 a view formed by spending time with computer users to find out how much people value their accumulated data, how they try to protect it and whether they鈥檝e succeeded.

Infinite U-Store-It

Most people adopt what she dubs 鈥渢he infinite U-Store-It鈥� approach, accumulating data haphazardly on various computers, gadgets, removable disks and online services. 鈥淚f you鈥檝e ever looked inside a U-Store-It you鈥檒l realise why this is a bad idea,鈥� she says. 鈥淧eople don鈥檛 realise what they have, they just save everything and when they do clean up they don鈥檛 do it systematically.鈥�

When asked, people typically say they value their data a lot. But they lose it nonetheless, more from disorganisation than from a technological catastrophe such as a hard disk failure, Marshall has found. Data can fall prey to online services or ISPs closing accounts or changing their policies, logins being lost, or simply forgetting what and where we have in physical or virtual space.

Web services 鈥� 鈥渃loud鈥� computing 鈥� are becoming the home for much of our data: for example, people often store their photos on or business contacts on . Giving stewardship of our data to a third party in the cloud could be a way to keep it safe from both disaster and disorganisation.

Night-light storage

For example, computer scientists led by at the University of California, Santa Cruz, are developing hardware for storage services designed to look after data that you have yet to create.

Their plan, dubbed , is to use low-power storage 鈥渂ricks鈥� that can each make 1聽terabyte of data available instantly over the web while using just 2聽watts of power 鈥� roughly the same as a pair of computer speakers.

The bricks contain digital storage and processors to manage that store and coordinate with other bricks. They can be connected together to make as large a store as is necessary with very little effort, and are designed to prevent future obsolescence: they connect using standard network switches to allow today鈥檚 bricks, which are built around hard disks, to work smoothly with tomorrow鈥檚 flash-based bricks, or those containing storage formats as yet unknown.

Memory lane

But as well as developing cheaper, more cavernous digital U-Store-Its, we need help to explore, organise and rediscover forgotten, perhaps decades-old data.

Software developed at the library of Stanford University, California, to record stories of pioneers of early computing suggests how this might be done. The can recognise places, names and other organising concepts in a person鈥檚 digital 鈥減apers鈥�, such as emails, letters and research reports. It then creates a branching 鈥渕ind map鈥� linking items by people, places or ideas that they have in common, forming an interactive digest of person鈥檚 life.

Such a tool could be of use to any of us now that diverse, disorganised digital archives are becoming the norm.

Read previous Innovation columns: How far can you trust an AI assistant?, The relentless rise of the digital worker, What use is a smartbook?, The sinister powers of crowdsourcing, Making a map for everyone, by everyone, Where next for social networking?, The dizzying ambition of Wolfram Alpha, Can technology persuade us to stop trashing the planet?, You Facebook, you Tweet, now lifelog.