PSA: Please Don't Watch Movies In Your Tesla - Forbes

This past spring, we saw headline after headline instilling fear in consumers that our private household conversations were being listened to by Big Tech employees. Here are just a few:

• "Amazon Admits Listening To Alexa Conversations"

• "Apple contractors 'regularly hear confidential details' on Siri recordings"

• "Google workers are eavesdropping on your private conversations via its smart speakers"

It seemed like no voice-activated assistant was immune.

But many of us who work in tech were unsurprised. That’s not to say we took the feelings of having our privacy violated lightly. Many of us unplugged our Alexas and threw them out the window. But those of us who work with the kinds of machine learning and automation technologies that underpin these voice assistants are acutely aware of their limitations. Investors love automation – after all, that’s what gets companies their 90%+ SaaS margins – but us technologists know that automation can only take you so far, that eventually, the Roomba will get stuck in the proverbial closet.

But does that mean we throw out automation altogether? Of course not. There are so many efficiency gains to be reaped via automated tools, as long as one understands their limitations. The key is using computers to do what computers are good at, and employing humans to do what humans are good at.

Thinking about automation this way is not new. Some of the most pioneering work in understanding the balance between humans and machines goes back to a groundbreaking paper written in 1960 by J. C. R. Licklider entitled “Man-Computer Symbiosis.”

In this seminal computer science work, Licklider describes the distinct competencies of humans and computers. Essentially, humans are good at judgment, creativity, brainstorming and can handle “very-low-probability situations” (what we might call “edge cases” these days). Computers, on the other hand, are basically good at counting things, albeit very, very quickly. As such, computers “will carry out the routinizable, clerical operations that fill the intervals between decisions,” while humans will ultimately be in charge of judgments, interpretations and making said decisions.

Presciently, Licklider concludes that “the computer will serve as a statistical-inference, decision-theory, or game-theory machine to make elementary evaluations of suggested courses of action whenever there is enough basis to support a formal statistical analysis. Finally, it will do as much diagnosis, pattern-matching, and relevance-recognizing as it profitably can, but it will accept a clearly secondary status in those areas.”

It's obvious that Licklider’s conclusion – that computers will dominate whenever there is a robust statistical model and ample data, but always playing second fiddle to human judgment -- is more relevant today than ever before. We see this tension between automation and human decision making play out every day, whether in my own field of information security, where scalability perpetually collides with the creativity of cyberattackers, or when someone considers watching a movie while their Tesla is on autopilot.

Computers operate well whenever there is enough data and a robust model to efficiently calculate an answer – think of this as the equivalent of asking Alexa a very common and easily recognizable question. But they will always be secondary to humans when handling the edge cases or less common scenarios or decisions, such as asking Alexa a difficult-to-answer question while you have a mouth full of peanut butter. That’s when those (admittedly creepy) humans must take over the listening function and figure out what you just said.

So how do we technologists reap the benefits of automation while staying true to Licklider's guidance? And what's the value of automation if we still need humans in the loop? It's simple: Automation still covers all the repeatable ground, whether that’s handling the common scenarios or sifting through enormous piles of data for computable signals. Even with the inherent limitations of computers and automation, they still help us achieve massive scale and handle unimaginably large quantities of information. That Roomba still vacuums most of the floor. We just need to keep an eye on it and help it along when it wanders into the closet.

In our modern world, where there is so much data humans could not possibly keep up, we need automation just as much as automation needs us. Licklider’s “symbiosis” is more alive today than ever before.

In conclusion, please don’t watch movies in your Tesla.

Let's block ads! (Why?)