Gigolo Joe prances through the hotel door, commands an internal computer to change his brunette coiffure to blond, and cocks his head to the right — Barry White emanates from a stereo buried deep beneath his polyester suit. One of the most endearing humanoid characters from Steven Spielberg's A.I, Joe represents a surreal and unlikely fantasy to scientists who expect artificial intelligence to remain a nebulous concept for hundreds of years. Other members of the scientific community say that Joe is only a generation away — a Hollywood version of the robotic counterparts that will live and work alongside humans within the next 50 to 100 years.
Brandeis University professor Jordan Pollack recently told Fast Company that the world of artificially intelligent humanoids introduced in Steven Spielberg's A.I. remains at least 200 years away. The staggering technology and funds needed to create lifelike robotic puppets like A.I.'s David will keep "mechas" (mechanisms) out of consumers' reach for another two or three centuries, he said.
Jack Dunietz begs to differ. The president of Tel Aviv-based Artificial Intelligence NV (Ai), Dunietz plans to introduce artificial intelligence to the consumer market late next year — roughly 199 years ahead of Pollack's estimation. Granted, Dunietz's virtual-assistant product will not walk or talk, but it will allow humans to converse naturally with learning machines. And that's just the beginning. Within 50 years, he expects humanoid robots powered by his company's intelligence to demonstrate convincing human feelings and form emotional bonds with their owners.
Dunietz's self-professed "revolution" began early last year in Israel, where Ai chief scientist Jason Hutchens introduced a technological entity based on a learning algorithm — a piece of software that acquires language through training and experience. This child machine, named Hal for the robot villain in Stanley Kubrick's 2001: A Space Odyssey, works like an artificial brain that gains conversational skills through its interactions with cognitive scientists and child-development experts. Like a human child, Hal is gradually learning to talk.
In February 2001, lingual experts outside Ai reviewed transcripts of Hal's conversations with chief trainer Anat Treister-Goren and could not distinguish between the language skills of the algorithm and that of a 15-month-old infant. According to computer-science visionary Alan Turing, a machine can be considered artificially intelligent if it can fool an interrogator into believing that it's human. Hal passed the test.
By 2005, Dunietz expects Hal to pass for a 5-year-old child. In 10 years, Hal will reach the conversational ability of an adult human. According to Dunietz, Hal will be a full-fledged manifestation of artificial intelligence ... with a growing family in the works. In a discussion with Fast Company, Dunietz explained how and when people will live side-by-side with artificial intelligence in a world where robots and humans become undistinguishable. Are you ready for his vision of the future?
In your lifetime, will scientists be able to create humanoid artificial intelligence that can develop and feel emotions?
Emotions are in the eyes of the beholder. They are something we ascribe to other people by observing external behavior. For example, an indication of people's happiness is their smile — we don't have an access to their minds. Therefore, the answer is yes — humanoid artificial intelligence will exhibit emotions in our lifetime. They will talk as if they have emotions.
People then might argue that those emotions are not "real" — but that could be said of the emotions of any human. At Ai, we maintain that if something exhibits emotions, responds to emotions, and can talk about emotions, then for all practical purposes, it can be termed as having emotions.
Will the field of artificial intelligence ever create true human companions — "mechas" to which humans can form emotional bonds?
Of course, humans will have emotional bonds with machines! They do already. Think of how upset you got the last time your computer broke down, for example. Once machines are able to converse with humans — which is what our scientists are striving toward — the emotional connection between humans and machines will reach new and unprecedented levels. Humans will interact with artificial intelligence just as they interact with each other — asking for advice, expressing emotions, and more.
Is Hal's use of independent communication a stepping stone for a David-esque form of artificial intelligence in the future?
Yes and no. David has a bodily manifestation; Hal doesn't. But in terms of lingual interaction, Hal is on the way to exhibiting the lingual capability, the ability to talk and understand, as a child of David's age would.
Why is it significant that Ai is teaching Hal linguistics and communication skills over time rather than just hardwiring basic communication skills into a fully developed form of artificial intelligence?
Fifty years of scientific experience shows that hardwiring "rules" of language into a system doesn't yield humanlike speakers. So we are approaching things from a new perspective at Ai. We create an artificial "brain" that is capable of learning, and then we teach it to talk. We maintain that speaking is like riding a bicycle — a skill that builds upon past experiences. Becoming an adult speaker includes having been a child speaker first. And one acquires that skill by constant practice and training.
This approach was first suggested by the renowned British mathematician Alan Turing, the father of modern computing. The results of our efforts so far are encouraging: Hal is progressing along the same lingual milestones as a human and is capable of speaking about a virtual world in which it "lives."
How will Hal's interaction with people ultimately resemble and differ from a human's interaction with another human?
In terms of lingual interaction — the ability to converse and interact with humans — Hal will not be different from any human. But he will not have a physical body. So, we will be able to talk to Hal anywhere we choose, exchanging ideas and information or even sharing a joke — but we will not be able to reach out and touch him.
What need or desire will Hal ultimately fill in the consumer market? When do you predict artificially intelligent beings will become commonplace in society?
Ten to fifteen years from now, artificially intelligent beings will be commonplace. They will be part of our daily lives and will accompany us in our home and in our work. They will be friendly, helpful, and nonintrusive. These companions will serve as gateways to most of the machines we currently operate: the television, the Internet, electricity. They will all operate in the most intuitive way: We will simply tell our artificial companions what we want to achieve — and they will take care of the rest!
In your opinion, what are the dangers of creating an artificially intelligent being that can love or inspire love from humans?
Raising artificially intelligent beings will require humanity to face moral and social issues that revolve around the existence of a new form of life. No doubt, controversies about the rights and responsibilities of these beings will arise, and will be an important part of the public discourse over the next decade. My personal opinion is that they deserve certain rights — but I think those rights are for society to determine. Those issues should not be left to us, the developers of the software.
Anni Layne (firstname.lastname@example.org) is the Fast company senior Web editor. Learn more about Artificial Intelligence NV on the Web.