Last time, I began a discussion of the safety of the devices that are becoming part of the Internet of things. Five principles aimed at better design of human–robot interactions are sensors for safety, context awareness, fast reaction time, robustness, and compliance (specifically, compliance motors). These provide a good starting point for exploring safety issues with big, industrial robots, but most future human–robot interactions and human–device interactions and interactions with the Internet of Things will not have anything to do with these large robots.
There is one exception—driverless cars and trucks. Fully automatic vehicles operating on our roads may be closer than we think. Depending upon legal concerns, I agree with Sergey Brin (Google) that some driverless vehicles are to be on the road regularly within five years. A driverless eighteen-wheeler makes my skin prickle, but I’m sure I’ll see the day when they’ll be rushing down interstates.
We already have cars that operate in driver-assist mode. These vehicles, which are designed to detect obstacles, warn about driver errors, parallel park, and so forth could be seen as large, interactive robots hidden in plain sight. The primary responsibility, however, is still with the driver, and there are many rules, regulations, and industrial practices in place to smoothly integrate those automobile innovations into our culture. In my lifetime, cars have been transformed for safety, with better breaking, crush zones, airbags, seatbelts, safety glass, and many other innovations. That history provides a strong foundation for development as cars and trucks evolve into full-fledged robots.
Oddly enough, cars may be more dangerous thanks to another connected “thing”, the smart phone. In fact, today’s cars have many distractions, with opportunities for phone calls, texting, viewing maps, blocking out exterior sounds, amping up music and talk radio, changing the dashboard display to monitor things like range and fuel consumption, tweaking seats and mirrors and airflow and heat—the list goes on. In fact, one of the great challenges for car designers today is providing enough electricity to power all the devices that ride along with us. (I know more than one person who has installed a second car battery.)
Unfortunately, our distracting devices have no context awareness. They don’t know whether you are parked or merging onto a highway. In the case of devices we carry into our cars, like the smart phones, they don’t know whether they are being held by a child in the back seat or the driver. I did see an article recently that points to an application for smart phones that will detect presence in a car and go automatically into hands-free mode (which research says is not necessarily better for safety in the case of audio conversations).
That may be a start towards applying some of the principles of human-robot interactions to devices other than robots. Sensing the presence of the vehicle, use by the driver, and traffic conditions could provide the information needed to understand context. If these existed, that still would not be the full story. A friend of mine who was a fighter pilot in Vietnam said the first thing he and his colleagues did once they got into the cockpit was turning off all the sensors because otherwise the alarms and other help they provided were distracting and more of a danger. Presumably, such systems are better 40 years later, but we will need to have devices that redirect, or cease working, or warn close enough to real-time so that people don’t get frustrated by their inability to react properly to changing conditions. They will also need to work in an imperfect world, one that will have many other active devices as essential parts of the environment.
One of the most intriguing possibilities in a world where devices are multiplying is collaboration, they have ready access to both data and processing power, and they communicate moment to moment with other devices. If common standards are in place, clusters of devices may create new and emergent capabilities. By analogy, data plus software applications already have led to mash ups that are produced informally by hackers and serve large numbers of people. Including physical components (sensors and actuators) that could physically or wirelessly snapped together like Legos would allow personalized benefits. Making this even more complex would be more general access to manufacturing capabilities so that some of the devices in the Internet of things would escape scrutiny, consideration, and regulation.
Both the huge rise in possible combinations of things and the development of untested devices will make safety more problematic. However, we may be saved from the worst unintended consequences if, as devices reach out to each other or are added into the Internet of things, background programs model possibilities rapidly and delay, block, or forbid some combinations. But, just as principles and standards in communications, cryptography, and transportation have needed extensive input and innovation to provide safety and coherence to networks and systems, the Internet of Things will need considerable input, discussion, creativity, and thoughtfulness to avoid peril and bring its potential benefits to individuals and society.