Why is the Amazon Echo incident important – a few takeaways from the Canvy Team

Update: According to Buzzfeed, the entire story is highly questionable, due to the fact that Amazon “said on Monday that its devices are not capable of calling 911”. Nevertheless, our point about smart devices potentially being responsible for people’s safety in the future is not invalid, even if the entire incident proves to be “embellished”.

***

What did the news report this week (on this subject)?

Who you gonna call? – Well, not the ghostbusters, since there was no ghost involved in the entire conundrum. And more to the point, the latest incident pointed out that the calling in emergency situations could go from a “who” to a “what”. The Amazon Echo (Alexa, in fact) system mistook a reply shouted during a domestic incident for an instruction – and called the sheriffs. This led to the de-escalation of the incident and generated quite a buzz in the online media – can smart devices safeguard humans?

Although perhaps a premature question, this surely falls upon a concept long-thought about in the R&D departments of many tech companies. Why premature? Due to the fact that the Amazon Echo’s intervention was accidental (nevertheless useful), and unexpected, it may be a long shot to see too much into it. A fortunate incident does not replace accurate and oriented tests, nor does it make a general rule.

What is beneath dispute, however, is the fact that this incident opened up to the wider public the idea of smart devices serving as a help in critical situations.

 

“Potentially life-saving” AI assistants

The image projected by such a sequence of words is truly appealing. Technology is appealing, in general, when placed in the context of helping humans, enhancing their abilities, helping them get through the mundane tasks quicker and better.

The downside of this image consists of, as always, the current ways we manage to fall short of this wonderful concept. Take cyber security alerts, for example. There is a notion related to having too many warnings, or warnings that are wrongfully triggered by not-so-dangerous security events – “alert fatigue”. The tool – a smart, real-time, alert generator cyber security system – when calibrated less than smart, ends up by defying its purpose. Important alerts end up by being ignored.

 

Smart devices for a safer life – but what are the requirements?

In a similar way, having smart devices safeguard people would be useful only if they are:

  • Optimally set up (with proper reaction thresholds, and the capability to integrate unexpected situations);
  • Accessible (while technology cannot break the cost barrier and remains inaccessible to a wide part of the population, no ubiquitous networks can exist, therefore the learning pool available for ML would remain limited, creating inaccurate patterns and sub-environments);
  • Endowed with the other part of the reaction chain – services specially tailored to pick up on the alerts and intervene;
  • Continually able to improve, with every new experience fed into the AI (wait – that is, in fact, the definition of Machine Learning – ML).
  • Cyber- secure: stories on how smart networks were hacked into serve as a warning, since this is the reality we live in.

Crossroads in technology

As wave breakers, the trials that transform tech concepts into prototypes and commercial versions often are a harsh reality check. We dream of AI, AGI, Machine Learning, digitizing everything – some of us even fear the day when machines will be able to design and produce other machines independently, or software would be capable of programming itself. Yet the reality represents most likely a hectic mix of concepts & limitations.

The road from theory to product, be it hardware or software, is a long, winding one. That is why devices and applications that can be trusted with human lives are extremely scarce. Making a completely autonomous, reliable prototype is extremely difficult and it takes a lot of trials.

Our app aims to be a friend to the people who employ it. Don’t take it this means developing such an app is easy – especially that we covered various elements of complexity. However, the challenges when developing a health or safety app are way higher.

What remains to be done is for us to keep looking forward, to aim for innovation, while threading carefully and detail-attentive towards each element that stands the test of reality.