My Kingdom for a Tricorder

I've been thinking a lot about tricorders lately. If you weren't raised on Star Trek though, you'd be forgiven for not knowing what that is. In Star Trek Land, it's a common trope that a problem is presented: a sick patient, an alien power source, or a strange new world. In all cases, our heroes use a "tricorder" (a hand-held "scanning" device) to detect what they're looking for: a pathogen, fuel, or life signs. It's a convenient device to further the story and add some jargon to make things sound sci-fi: "I'm reading elevated levels of dilithium captain" etc.

For our present day however, the need for a tricorder is becoming more and more apparent. We're seeing massive advancements in data routing, warehousing, processing and machine learning, but very little on the collection of that data. Some of the most advanced ML outfits in the world are using limited data sets as their input: user-provided data, or simple computer vision are the most common sources. The result of course is that all of the power afforded us by these new technologies is limited by the kind of data we can feed them.

A lot of people have been saying that the Next Big Advance in technology will have to come in energy storage -- and they're probably right, but I think it's reasonable to say that following close behind will have to be sensor technology like portable, high-resolution infrared spectroscopy. Something that can identify the makeup of objects so we can make decisions around what to do with the thing we're scanning.

Imagine a world where you can determine, in a fraction of a second, what something is made of. Suddenly waste reclaimation can be automated: breaking down plastics, fabrics, circuit boards into fragments the size of a grain of sand to be reused, composted, or melted down without the need for human intervention. We can feed breath samples into a sensor, and combine the data collected with the billions of other samples to use machine learning to quickly and cheaply diagnose someone with a disease.

These are the two cases that've been bouncing around my head for ages, and there are bound to be more. We're only beginning to understand the potential for all of our new-found data-driven technologies, but what we do know is that they work best with large, high-quality data sets, and that their ability to give us answers to important questions depends on our ability to collect that data for them.

When the WHO declared Covid-19 a pandemic, I started to think about how we could automate, speed-up, and distribute testing and so far, my research has lead me to two interesting places:

• A Canadian company developing a portable NIR Spectometer
• A 2016 paper about using NIR spectroscopy to detect viral infections

So far as I can tell, neither source has considered combining their findings with ML, so I'm going to send some emails. Perhaps it's finally time for our own tricorder.