AFTER THE TURING TESTBy
Artificial intelligence (AI) is an acknowledged inevitability. As Pamela McCorduck explains in her book Machines Who Think, “AI is an epochal scientific, technological, and social—human—event. We’ve developed a new mind to live side by side with ours.” Now the question becomes how smart our machines can become.
Siri and Cortana can be useful, even friendly, and IBM’s Watson supercomputer has graduated from being Jeopardy champion to learning to do actual medical diagnoses. As McCorduck promised, “If we handle it wisely, [AI] can bring immense benefits, from the global to the personal.” And as computers are learning to drive our cars, AI is becoming more familiar, even expected.
How far we will take this experiment depends on researchers’ ideas of what AI is and why we are developing it. One scientist told McCorduck that AI is about our “looking out at the world,” that it is “foundational” in our search to know. Others, McCorduck reports, connect the development of this external intelligence to our very evolution. “We’re a mere step in the evolution of intelligence in the universe, attractive even in our imperfections, but hardly the last word.”
THE TURING TEST
Sixty-six years ago, mathematician Alan Turing came up with a mile-mark test for machine intelligence. If a person is unable to decide whether they are talking to a computer or another person 30% of the time during a series of five-minute keyboard exchanges, the machine has joined us as recognizably “human.”
Recently, there have been a number of claims of success for the computers in Turing Test competitions, including a program called Eugene Goostman, which purported to be a 13-year-old Ukrainian boy. On June 7, 2014, Eugene persuaded 33% of the judges at the Royal Academy in London that they were indeed conversing with a human being.
There are those who still dispute the chatterbot’s performance, but many have moved on to another, loftier expectation. The question for these researchers now is whether robotic intelligence can achieve conscious awareness.
Reporting on Quartz Daily Brief (qz.com), Olivia Goldhill outlines what it will take to create a conscious robot according to Ryota Kanai, a cognitive neuroscientist at the University of Sussex and University College London.
Kanai posits three requirements. “To build a conscious robot, you first need to program it with a model of the world so that it can recognize changes in the environment, and secondly create a program that links action with sensation, which will give the robot an internal representation of itself.” Perception and recognition of objects have been slow, like the progress mapped for machine speech-recognition, but progress seems to be accelerating.
Goldhill continues: “Thirdly, [Kanai] wants the robot to be able to generate hypotheses about the world, which he believes will essentially act like imagination.” Such a robot would need a model in place to provide an understanding of what it sees going on outside of itself and the sensory input to create an image of itself in order to “internally simulate the outcome of its possible actions without actually enacting them.” At this point, Kanai believes the robot will be capable of conscious choices about its own actions.
Of course, the level of consciousness will be very rudimentary, about at the level of maybe an ant, according to Kanai, but work has begun on at least the theoretical part of this project, if not the actual programming.
But there are those who have a different definition of consciousness, one that’s beyond just an awareness of one’s existence and surroundings. Goldhill also reports on Paul Verschure, a research professor at the Catalan Institute of Advanced Research in Catalonia, who believes that consciousness is information plus an awareness of and ability to interpret the behavior of others. In other words, it involves a social dimension where the robot would not only see itself but also understand that it must interact with others as well.
There are companion robots for sale today that will interact with your children. They learn the kids’ names and even the sound of the voices of the other members of the family. They’ll help with homework and read the kids a story every night at bedtime. Perhaps the most famous is the four-foot-tall Pepper, the “emotional robot” from SoftBank Robotics. Plans to begin development of robots that will assist the elderly in their homes, including monitoring and relaying health data to others, are also underway. Even a most rudimentary consciousness would transform these robots as “new mind[s] liv[ing] side by side with ours.”