Intel develops a computer chip to smell hazardous chemicals

Intel

Intel develops a computer chip to smell hazardous chemicals

Intel develops a computer chip to smell hazardous chemicals; the computer motherboard maker has designed a chip that can smell various chemicals present in the air.

The chip is designed with a neural algorithm for the rapid online learning and identification of odorant samples under noise. The chip is based on the architecture of the mammalian olfactory bulb and implemented on the Intel Loihi neuromorphic system.


Intel has done or at least what they say they've done is successfully reproduce a part of a human brain and put it onto a computer chip, and specifically the part of our brain that is responsible for smelling and for understanding what it is that we are smelling so.

The technology company develops a chip that can smell without a nose.


That was obvious at first to understand how the human brain smells. So what happens when you smell is that you stimulate up to 450 different types of olfactory senses that sit inside your nose, and they get stimulated by airborne molecules. And so certain of these senses will be stimulated by cells and molecules they'll send a signal to the brain. And then in the brain, this transforms into an electrical pulse within a web of interconnected neurons. And the brain reads those electrical pulses and understands what it is. That you're smelling what the brain does.
Even if you're coming across some new information it can still sort of identify what it is that you're smelling, and ultimately our human brain can in this way use these electrical pulses and read those electrical pulses patterns. It can identify between trillions of different smells.


With the help of Intel’s Loihi platform, the chip uses machine learning algorithms to identify scents of hazardous chemicals that present in the air.


At the initial stage, the chip can identify 10 hazardous chemicals with different scents after just one single exposure. The human brain can differentiate between trillions of different scents that are not something that the computer chip contains just yet.

To start with Intel's researchers, the chip can differentiate between 10 different smells at this stage. To do that they took a wind tunnel. They put 72 chemical sensors in the tunnel, and then they circulated those 10 different smells that they picked that included methane, acetone, etc. And they circulated those scents through the tunnel and observed how the chemical sensors were reacting. Obviously, the different smells stimulated different chemical reactors sensors and they recorded that. They made it into a data set, and that's the data set they fed the chip. So, the chip based on those reactions on those patterns of reaction is able to identify between the ten different senses that the researchers have been working with.


Intel's team created an algorithm that just like the brain can read those patterns of electrical pulses can read patterns, and transform them into the different well. Transform them into identifying different smells.


Nabil Imam, a senior research scientist in Intel Lab’s neuromorphic computing group, said in a statement,

“My next step is to generalize this approach to a wider range of problems-from sensory scene analysis (understanding the relationships between objects you observe) to abstract problems like planning and decision-making. Understanding how the brain’s neural circuits solve these complex computational problems will provide important clues for designing efficient and robust machine intelligence.”

So, the technology company sort of created an artificial network of neural and artificial neural networks. Based on the neural networks it is capable of identifying certain patterns and associating them to different smells.


The breakthrough is ground-breaking for two reasons. First, is that this can lead to a number of advancements in the implementation of AI. Secondly, the replication of electrical impulses in our body on a chip level opens up endless possibilities in the future of AI.


The applications of this type of technology and this type of algorithm could be wide-ranging and they're already looking at potentially, for example, understanding the way we make decisions on transforming that into algorithms, and then putting that into chips just like they did with this technology.


It’ll be interesting to see how this chip is implemented in the near future.





Source