PARIS — AlphaGo“s recent victory over Lee Sedol, one of the world’s best Go players, has people once again debating the merits and potential perils of intelligent machines.
The computer program, developed by London-based artificial intelligence firm Google DeepMind, beat the South Korean Go champ 4-1 in a televised contest reminiscent of Deep Blue’s victory, two decades earlier, against Russian chess master Garry Kasparov.
But as exciting as the development may be, we should be careful not to let the accomplishment distort our vision of what these robots —which are increasingly present in our lives — will be capable of.
Robots and connected objects will integrate our homes, just as cellphones and televisions have. We already have vacuum cleaners capable of detecting obstacles and moving around autonomously. Soon, we will be able to talk with robot assistants, in the same way that voice recognition technology already allows us to speak to our phones.
Robots have their limits, however. Machines owe their artificial intelligence to computer models designed by humans. If it is programmed to detect and recognize emotional and conversational signs to adapt to humans, or even try to be funny, this machine can seem welcoming. Robots struggle, on the other hand, to glean concepts such as emotions and dialogue strategies from large data bodies, as these concepts are not as easy to formalize.
Children learn by experimenting with the world. For a robot, this task is extremely difficult because it has no instinct or intentions with which to make decisions. A machine cannot autonomously build new representations when faced with a new task. Robots might be able to say “I love you,” but they don’t actually feel anything. They have no awareness.
[rebelmouse-image 27090125 alt=”” original_size=”768×1024″ expand=1]
Virtual love — Photo: Joamm Tall
Machines will be more and more autonomous thanks to sophisticated artificial intelligence programs. But they will not be capable of feelings, creativity and imagination in the way humans are. A machine cannot feel “happy” because it does not have phenomenal awareness. Nor can it “understand” the concept of happiness. When AlphaGo beat Lee Sedol, the machine did not understand what it was doing.
Assistants and caregivers
Faced with questions with no answer, humans are significantly better at finding solutions than machines. On the other hand, machines can certainly outperform us for specific tasks. In some cases they can even accomplish things that are simply impossible for humans.
Machines can be stronger than humans, for example, when it comes to completing complex calculations — and quickly. They can be designed to have incredible memories and are thus better at answering encyclopedic questions, or recognizing people. It is in our collective interest to take advantage of these capacities, without any fear.
By 2060, 32% of the French population will be over the age of 60, an increase of 80% in about 50 years. The proportion of chronic diseases will go hand in hand with the aging of the population. Robots — sometimes in humanoid forms, to better move around in our homes — could be very useful in this context: as assists, monitors, even to keep people company. The machine-human relation will likely be triangular, with robots acting as a go-between for family members and medical staff.
Research already shows some promising results. Initial studies on interactions between robots and the elderly, carried out with occupational therapists from a French association called Approche, as well as with gerontologists from the “living lab” of the Broca hospital, in Paris, suggest a fair amount of interest in new kinds of human-machine collaboration. Interest is likely to grow further still as time goes on and people enter retirement age with relatively greater knowledge of computing and technology.
How real is too real?
Creating an emotional relationship with robots is no longer science-fiction. Resemblance with humans or animals, facial impressions, tones of voice, or even the childlike or teddy bear aspects of some robots arouse emotions. Humans also project emotional relations with non-humanoid robots, devoid of emotional abilities, or even vacuum cleaning robots. Some people give them names — proof that humans project identities on robots. Attachment is an emotional bond that results from joint history.
[rebelmouse-image 27090126 alt=”” original_size=”640×640″ expand=1]
You don’t mean that — Photo: Betsy Weber
The Media Equation theory put forth in 1996 by Stanford University’s Clifford Nass and Byron Reeves explains that we apply the same social expectations when we communicate with artificial entities. Anthropomorphism is the attribution of human behavioral features to objects. In this way, an object that appears to feel pain, like the Atlas robot, developed by Boston Dynamics, can generate empathy.
Thanks to brain imaging, researchers know that humans can feel empathy towards mistreated robots, albeit without the same intensity as they do for mistreated humans.
This kind of emotional and social interaction between humans and robots gives rise to several ethical questions. How humanlike should developers be allowed to make robots? How autonomous should the machines be with regards to decision making? What about the “six-million-dollar-man” scenario? Should there be limits on the use of robot technology to repair or enhance human beings? For now, few studies have attempted to answer those questions.
The reality is that over time, we will see more and more autonomous, complex entities that can adapt to their environment. And the emergence of those robots will present a number of legal and social challenges — questions we ought to start posing and attempting to answer now. How do we adapt our education systems so that we can create work around these machines? What kinds of new jobs are we talking about? For what tasks do we want to create these artificial entities? And how do we ensure that robots help us, rather than simply put us out of work?
