Single Molecule Switch. Researchers at the Peking University of Beijing have created a switch that can be turned on and off by a single photon. This opens up the possibility of developing light-based computers and electronics. To make this work the researchers needed to create a switch using just one large molecule. The new switches begin with a carbon nanotube to which three methylene groups are inserted into the molecule, creating a switch that can be turned on and off again.
Until now researchers had not found a molecule that was stable and predictable. In earlier attempts of the technology a switch would turn ‘on’ but would not always turn off. Further, they needed to create a switch that lasted, since the switches created in earlier attempts began to quickly break down with use. The new switches function as desired and look to be good for at least a year, a big improvement.
Chips that Mimic the Brain. There are now two different chips that have hit the market that are introducing neural computing in a way that mimics the way the brain computes.
One chip comes from KnuEdge, founded by a former head of NASA. Their first chip (called “Knupath”) has 256 cores, or neuron-like brain cells on each chip, connected by a fabric that lets the chips communicate with each other rapidly. This chip is built using older 32 nanometer technology, but a newer and smaller chip is already under development. But even at the larger size the new chip is outperforming traditional chips by a factor of two to six times.
IBM also has released a neural chip it’s calling TrueNorth. The current chip contains 4,096 cores, each one representing 256 programmable ‘neurons’. In traditional terms that gives the chip the equivalent of 5.4 billion transistors.
Both chips have taken a different approach than traditional chips which use a von-Neumann architecture where the core processor and memory are separated by a buss. In most chips this architecture has been slowing down performance when the buss gets overloaded with traffic. The neural chips instead can simultaneously run a different algorithm in each core, instead of processing each algorithm in sequential order.
Both chips also use a fraction of the power required by traditional chips since they only power the parts of the chips that are being used at any one time. The chips seem to be best suited to an environment where the chips can learn from their experience. The ability of the chips to run simultaneous algorithms means that they can provide real-time feedback within the chip to the various processors. It’s not hard to imagine these chips being used to learn and control fiber networks and be able to tailor customer demand on the fly.
Improvements in WiFi. Researchers at MIT’s Computer Science and Artificial Intelligence Lab have developed a way to improve WiFi capabilities by a factor of three in crowded environments like convention centers or stadiums. They are calling the technology MegaMIMO 2.0.
The breakthrough comes from finding a way to coordinate the signals to users through multiple routers. WiFi signals in a real-world environment bounce off of objects and scatter easily, reducing efficiency. But by coordinating the signals to a given device like a cellphone through multiple routers the system can compensate for the interference and scattering by recreating a coherent understanding of the user signal.
While this has interesting application in crowded public environments, the real potential will be realized as we try to coordinate with multiple IoT sensors in an environment.