March 10, 2008 at 8:00 PM ET
For years, nanotechnology has held out the hope of molecular-scale contraptions that can manufacture custom-made drugs or revolutionize the way computer chips work.
Now researchers in Japan say they have taken a big step toward that nano goal by creating the first molecular machine that can do parallel processing.
"The discovery could provide a way to control many molecular machines simultaneously, increase computer processing power, and perhaps keep Moore's Law alive," according to the Proceedings of the National Academy of Sciences, which published the researchers' paper online today.
The multitasking machine was coaxed to assemble itself on a surface of gold from 17 molecules of an organic compound called duroquinone. Sixteen of the molecules form a weakly bonded ring around the central molecule, which serves as the control unit for the machine.
Anirban Bandyopadhyay / ICYS
|Click for video: This graphic shows the structure|
for a parallel-processing molecular machine, Click
on the image to watch a video about the machine;
click here for a larger version of the graphic.
Using electrical pulses from the tip of a scanning tunneling microscope, the researchers could flip the control molecule to any one of four configurations, or states. Those flips, in turn, could change the states of the other 16 molecules - just as, say, knocking down one domino can simultaneously set off several chains of falling dominoes.
Researchers used the scanning tunneling microscope to make sure that the 16 molecules really did respond to the central control molecule as they hoped.
"We can use [the central molecule] like a space station to talk to spacecraft - or if you have seen the movie 'Fantastic Voyage,' it is similar to that," Anirban Bandyopadhyay of Japan's International Center for Young Scientists told me over the weekend. Bandyopadhyay and a colleague at the center, Somobrata Acharya, are the researchers behind the study unveiled today.
Bandyopadhyay said that the assembly was modeled on how glial cells work to pass along instructions among neurons in the nervous system. Such "one-to-many" communication is essential to the way the brain works, and computer scientists have said for decades that massively parallel processing could revolutionize the way machines think.
"The architecture looks almost like the neural network inside our brain," Bandyopadhyay told me. He said his findings showed that a single instruction given to the control unit was capable of generating more than 4 billion (416) possible outcomes.
Molecules and medicine
But wait ... there's more: If scientists can create assembles that can pass along instructions from one molecule to 16, then to 256, then to 4,096, and so on - pretty soon you could have nanofactories capable of churning out mega amounts of custom-designed molecules. That could open the way for medical therapies that have long been the subject of dreams (and nightmares).
"In the future, there will be no surgery for brain tumors," Bandyopadhyay said. "The blood [containing molecular assemblies] will be injected into the body, and will go to the targeted place."
Once nanochips containing the molecular assemblies reached a place where they sensed that a tumor was active, they would gear up the machinery and start producing molecules custom-made for small-scale chemotherapy. When the tumor was taken care of, the machines would shut themselves off. At least that's the theory.
Mark Ratner, a chemist at Northwestern University who specializes in nanotechnology, said the newly published research represented a significant step toward molecular-scale computers as well as molecular-scale medicine.
"People have been talking about both these things for a long time," Ratner told me. "People have even thought about putting these two things together. ... But this is quite pretty because [the researchers] actually use all of the constituents, and that's really neat."
Nano now? No ...
Ratner edited the paper for the Proceedings of the National Academy of Sciences, and he said the reviewers were particularly impressed to see how the molecules meshed themselves into a machine - as well as how an electrical input into one molecule could produce multiple responses. However, Ratner cautioned that the technology wasn't ready for practical application.
"Is it useful tomorrow? No," he said.
One of the biggest conceptual hurdles has to do with the input/output device: Although the assemblies themselves are at the molecular scale, the scanning tunneling microscope is a big piece of equipment. It wouldn't be practical to use those microscopes to read out the result of a nanocomputer, or harvest the chemicals produced by nanofactories.
Bandyopadhyay said other control methods would be developed for working devices - perhaps optical readers for the nanocomputers, or chemical triggers for the medical nanochips. Ratner said several companies, including an outfit called NanoInk, were working on technologies that might work.
In the meantime, Bandyopadhyay is working to ramp up his molecular machines from two-dimensional arrays to three-dimensional structures. "Within one and a half years we will have 1,024 machines connected," he told me.
Theoretically, the technology could allow for the development of a super-duper information processor contained in a sphere less than 2 inches in diameter, Bandyopadhyay said.
"That will contain the equal amount of components and connectivity that is required inside our brain," he told me.
Say that again?
After our initial talk, Bandyopadhyay sent me an e-mail going into more detail about the potential applications of his molecular machines. Here's the text, slightly edited to fine-tune the style and add Web links:
"The first application is mimicking the 1965 movie 'Fantastic Voyage.' Prior to our work, the prediction that in the future medical doctors will not have to go for surgery if a patient has brain tumor, or damaged lungs, heart or lever, was considered merely a dream. It was also predicted that, for any disease, the doctor would choose suitable molecular machines - but it was not known how they would be controlled after injection into the blood, though several machines already have been invented.
"Now our work provides a unique conceptual solution. What a doctor would have to do is attach a control program molecule at the center, similar to our kind of machine assembly. The specific molecular machines dedicated to do particular job would be attached around the ring. Then, finally, the complete assembly would have to be injected into the blood. The machine assembly would be capable of instructing the other machines docked with its ring members.
"The major query that arises is: Given that we study the system under a scanning tunneling microscope in an ultrahigh vacuum condition, what is the guarantee that it will work as a standalone system? Now, since to build a standalone system we require only a single contact to the central molecule that would be used to donate and extract electrons as required, we have already found suitable reversible redox active enzymes. Therefore, it could be programmed in such a way that a very particular environment would activate the central molecule in a remote environment.
"The second application is building a massively parallel supercomputer based on the working principle of our brain. The computer that we are going to build is based on the proposal of L. Chua and Roska’s work on cellular neural networks (CNN) in 1989, which is a combination of cellular automation and the neural network of our brain. In this concept, highly interconnected arrays of cells communicate with all their neighbors at a time, following a particular equation. In principle, these unconventional processors are astronomically powerful compared to existing processors.
"Several such processors have been proposed; however, the key to the system's realization - 'one-to-many communication' at a time - has not been realized yet. Therefore, this is a significant advance compared to the realization of single CNN cells. Also, please note that these processors will not use any logic gate. It will be purely visual computing, where patterns will replace the differential equations that have been used to express physical phenomena for the last 300 years of science.
"Until now, all these concepts were seen in mathematical models and particular CMOS chips that have been constructed based on CNN principles. But true realization of 'one to many communication' would make an important paradigm shift, as it opens the door to bottom-up parallel processing."
Will we see molecular machines doing useful work in our lifetime? If so, will this lead to the "singularity" that futurist/inventor Ray Kurzweil is predicting for the year 2045? And could that in turn set the stage for a real-life replay of "Terminator 3: Rise of the Machines"? Feel free to weigh in with your comments below.
Update for 12:11 a.m. ET March 13: Based on the comments, I changed the date for Kurzweil's prediction of the singularity to 2045. The year 2029 is when Kurzweil believes artificial intelligence will basically match human intelligence, as described here.