Imagine you are in a car driving at 90 miles per hour. You're behind the steering wheel, but you are not paying attention. You even doze off for a moment... and that's perfectly fine...
Your car drives itself. This so-called level 5 autonomous driving is powered by Machine Learning and requires no human interaction. Your vehicle is able to steer, accelerate, brake, and even monitor road conditions like traffic jams.
You are free to let go of the steering wheel.
But will you?
And will the car manufacturer advise you to do it?
Only if it works flawlessly for every situation.
Neural networks and edge-cases
Though car technology is capable of many great things, we do not have full-self-driving cars yet. This has to do with the amount of so-called ‘edge-cases’. Whilst driving on the motorway with proper maps and known road conditions might work, the neural networks that power level 5 autonomous vehicles are not capable enough yet to handle all exceptions. Level 5 might be feasible at some point, but no one knows exactly when. And the industry is becoming increasingly reluctant to set a timeframe.
Letting go of the steering wheel: chatbots
There is a clear analogy with the use of chatbots in the customer service industry. Chatbots might work for ordering pizza (the highway), but not for modifying your retirement plan (low-resolution maps). But whilst it is abundantly clear for most people you do not let go of the steering wheel until you are absolutely sure it is working, enterprises deploy chatbot technology without any oversight as if we have reached level 5 autonomy years ago.
Skin in the game
This has to do with what’s called 'skin in the game' (thematized by Nassim Nicholas Taleb): The direct result of using an immature autopilot can lead to the death of the driver, with a huge backlash to the manufacturer, whereas the vendor or consultant selling chatbot technology suffers no direct consequence. Nevertheless, unsatisfied customers do pay the price, and losing customers is equivalent to death from a commercial perspective.
No wonder the automotive industry has gone down a different path. They give people all sorts of assistants while driving, such as lane-keeping, adaptive cruise control, and emergency braking. At some point, they might be capable of combining this technology with Machine Learning and deliver autonomous driving. Along the way, they add value to customers: enjoying a more relaxed and safer drive.
Augmented customer service
Our perspective at Deepdesk is the same. We already add value by augmenting agents in contact centers. With Machine Learning we make their work easier and more efficient. We believe building from here is a good way to achieve autonomy at some point in the future. By choosing to assist agents as our starting point, we circumvent using customers as the proverbial crash test dummies, while building towards full autonomy - whenever that may be...