March 14, 2017
‘Siri, What’s Your Opinion of AI?’
How can we trust the AI in a driverless vehicle when the relatively simple interactions we experience today are botched so badly?
Astounding leaps forward have been made in two key technologies in the last couple years: voice recognition and artificial intelligence. Virtually everyone is becoming used to conversing with Siri, Alexa and others, backed by powerful voice recognition engines to convert speech to text. At the same time, AI is at the heart of breakthroughs in driverless vehicles, robotics, IoT devices and other emerging technologies. The promise of these technologies to make our world safer, improve our lives and address chronic societal problems is no longer the province of science fiction writers. But a dichotomy must be resolved: the gap between the current performance of voice and AI technologies and the levels required to achieve the benefits of driverless vehicles and other technologies.
Practically everyone can relate to frustrating incidents in using voice- and AI-based systems for rudimentary tasks. My personal experience with these systems often produces results that are comical, sometimes offensive, and many times just downright misleading. The Bluetooth in my car cannot even correctly interpret names when I want it to dial individuals with simple, single-syllable first and last names. And we have all seen crazy text conversations posted on Facebook.
See also: Don’t Be Distracted by Driverless Cars
On the AI front, the use of virtual assistants is becoming more common. My recent experience with one was a trigger for writing this blog. Communicating with a virtual assistant to address a technical email issue, I was sent into endless loops and misunderstandings, always to be followed with, “Did I answer your question?” My pleas to PLEASE LET ME TALK TO A LIVE HUMAN were answered with nonsense responses, or “Please rate your satisfaction with the answer.” You can guess my response. Similarly, if you have ever used a virtual assistant to try to schedule meetings, you have most likely come away completely flummoxed and ended up emailing or calling your colleague directly to schedule it. These types of interactions begin to give you a deeper appreciation for the nuances of human communication.
My point in describing the tech shortcomings is that we are headed into a future where we will rely extensively on them. The AI behind decisions that affect your life will not all occur when you are barreling down the highway at 65 mph. But much of our transportation, healthcare, entertainment and education and many aspects of our daily lives will rely on the recommendations and decisions made by AI-based systems. And we will control the world around us largely via voice commands.
Both voice and AI technologies are improving rapidly, but the question becomes – how can we be confident that the AI in a driverless vehicle will make the right life and death decisions in milliseconds when the relatively simple interactions we experience today are botched so badly?
Will the robot companion of my elderly father make a fatal mistake on medicine dosage? The accuracy and success of these technologies when used in the future for driverless vehicles, smart homes and the world at large is essential.
The implications for the insurance industry are significant. The progress of these technologies bears close monitoring so that the regulatory environment for, availability of and usage of new capabilities in the connected world do not actually make the world riskier and less safe.
See also: Who Is Leading in Driverless Cars?
For now, we must live with (or suffer through) the current state of the tech. Just for fun, I asked Siri, “What’s your opinion of driverless vehicles?” My question was interpreted as “What’s your opinion of dry rose vehicles?” Need I say more?