The latest in a rapid trend towards AI-powered tools, Predictmedix has announced the production of a mobile app that can detect alcohol and cannabis impairment using speech recognition and “multi-spectral imaging.”
AI-powered facial recognition is coming to the world of breathalyzers. Though the technology is still ill-defined, there are some causes for concern.
If that mishmash of convoluted buzzwords sounds concerning, then you’re not alone.
The new algorithm, which is being designed primarily for law enforcement and healthcare providers, can allegedly identify impaired individuals in under 30 seconds with an over 90% success rate. That’s leaps and bounds better than the leading cannabis breathalyzers and significantly better than the average alcohol breathalyzer.
Predictmedix previously designed an AI-powered detection tool for Covid-19 screening. (Photo: Predictmedix)
The speech recognition aspect of the app is not entirely new — in January, researchers at LA Trobe University invented a similar program called ADLAIA that can determine an individual’s intoxication level based only on a 12-second recording of their voice.
The facial recognition aspect is, however, truly unprecedented. It’s also not very well explained.
In their press release, Predictmedix gave only this explanation of the technology behind their program:
The critical element, multi-spectral imaging, is a form of heat-detecting infrared camera mostly used for military target tracking.
From what we can deduce, the algorithm identifies visible signs of intoxication (behaviors like eye drift) and combines that information with heat-sensitive imaging that can somehow deduce inebriation.
Predictmedix’s product is one of many facial recognition algorithms to hit the market over the past few years. Though many are designed for law enforcement, an increasing number are being applied with the intent of alcohol moderation.
The UK Home Office is currently trialing a program in supermarkets called Yoti, which uses facial recognition to detect customers’ ages when purchasing alcohol.
Though it has a high accuracy rate, the program routinely demonstrates higher margins of error for females over males and for darker skin tones over lighter ones, errors consistent with the biases frequently found within these technologies.
And that doesn’t even account for the litany of privacy concerns currently being voiced about the rise of facial recognition.
All of this is to say, the future of AI-powered technology can feel intimidating. Many of its faults and benefits have yet to be proven even as it’s wheeled out into the field for day-to-day use.
Until Predictmedix’s program makes its way to market, we’ll just have to wait and see what the implications might be.