At Nexperia, we’re always looking to the future for innovations in digital logic. Everything from AI leaving being implemented into personal devices using the logic gates they have already, soft logic gates that use fluid instead of electricity, to the beginnings of tiny genetic computers. 2017 should have even more interesting developments! Let’s have a look at some of these research projects.
Machine learning has traditionally been bound to supercomputers due to the massive amounts of processing power required. Machine learning traditionally had to be run in serial because many calculations would be dependent on the results of previous calculations. Basically, computers would have to calculate one equation at a time, which requires a lot of processing power to work through in a timely manner. Although remote data centers can do these calculations quickly, they require internet connectivity.
The Allen Institute for Artificial Intelligence came up with a workaround that they call XNOR.ai. This was accomplished by essentially simplifying and rounding the mathematical operations that computers have to perform. XNOR.ai, named after the XNOR gate, uses the logic gates that are already in processors to carry out simpler operations that can be handled at the transistor level. The challenge at hand for the Allen Institute’s team was to frame complex mathematical operations into binary.
This was accomplished by giving the program an allowed margin of error. XNOR.ai sacrifices accuracy for efficiency and accessibility. It should be noted, however, that XNOR.ai isn’t the only program taking this approach, there are also programs like Scorchai. However, XNOR.ai’s open source approach makes it intriguing, and you can download the software on Github. It will be interesting to see the applications people come up with! You can see XNOR.ai in action in the video below.
Soft robotics is a new field, but it’s already drawing a lot of interest due to its potential applications. Traditionally, robots have been confined to metal bodies, making them not ideal for tasks that require a delicate touch. To get around this, researchers at Harvard and Caltech have developed the first autonomous, entirely soft robot nicknamed “The Octobot.” Just like an actual octopus, it has no hard components, making it a good fit for those delicate tasks.
The Octobot is made possible by microfluidic logic gates, which work like electronic logic gates, but transfer liquid fuel instead of electricity. So far, the team has managed to make AND and OR logic gates with the new technique. The technology has a long way to go before it’s seen in any consumer applications, but the potential should draw interest for years to come! You can find Caltech’s paper here, and you can watch Octobot in action in the video below.
Duke University’s Analog DNA Computers
It may sound like science fiction, but the first logic gates with synthetic molecules were actually designed in the early 1990’s. Recently, the technology has taken some interesting leaps. Unlike digital circuits that use currents and voltages to represent inputs and outputs, DNA computers utilize concentrations of different DNA strands as signals. One big hang up for this technology was that DNA was often thought of as digital data, similar to the binary codes used by computers. The goal of DNA computing was to make tiny devices that could fit in the bloodstream and diagnose dangerous chemicals linked to diseases. However, the furthest the researchers got with digital DNA computing was making simple functions like calculating square roots and playing tic-tac-toe.
Professor John Reif and grad student Tianqi Song making DNA computers in test tubes. Courtesy of Duke University
Researchers at Duke University worked around this by taking an analog approach. Since devices that can fit in the bloodstream are so small, they could only get basic on and off states. Unfortunately, diagnosing atypical chemicals linked to diseases like cancer required more complex functionality. The team at Duke thinks that their analog approach will allow DNA circuits to calculate more complex functions like logarithms and exponentials. The circuits accomplish this by adding, subtracting, and multiplying as molecules form and break bonds. This technology is still very much in its infancy, so designating them as hype is a fair assessment. However, the potential benefits make this technology intriguing nonetheless.
What Will 2017 Bring?
It’s interesting to see that logic gates are going beyond the realm of electronics. Whether logic gates are applied to new fields that nobody has thought of yet, or advances in software allow us to get even more use out of the digital logic gates we have now, it should be a fun ride!