Expansionary Institute

Michael Zey
futurist3000@aol.com

By Megan Goldin

 
TEL AVIV, Aug 20 (Reuters) - Meet Hal. Like any 18-month-old toddler, he likes bananas, toys and playing in the park. He especially enjoys bedtime stories.

But while other children are flesh and blood, Hal is actually a chain of algorithms -- a computer program that is being raised as a child and taught to speak through experiential learning in the same way as human children.

"He is a curious, very clever child, someone that always wants to know more," said neuro-linguist Dr Anat Treister-Goren who is Hal's "mommy" and readily admits her attachment.

"Some kids are more predictable than others. He would be the surprising type," she said. Treister-Goren talks to Hal and reads him stories in much the same way a mother teaches her young child to learn about colours, food and animals.

"I build his world on daily basis," explained Treister-Goren.

She heads the training department at the Israeli-based Artificial Intelligence (AI), where she inputs information and language ability through conversations with Hal and works with computer experts who fine-tune his algorithms to enhance performance.

The privately owned company, which is run by Israeli high-tech entrepreneur Jack Dunietz, aims over the next 10 years to develop Hal into an "adult" computer program that can do what no program has ever done before -- pass the Turing test.

The British mathematician Alan Turing is one of the founders of computer science and the father of artificial intelligence. More than 50 years ago he predicted the advent of "thinking machines."

But in Turing's time, computers were slow and cumbersome devices, utterly incapable of fulfilling his vision.

Turing, who died in 1954, left behind the benchmark test for an intelligent computer -- it must fool a person into thinking it is human. No computer program has ever succeeded.

If, or when one does, it will open a Pandora's box of ethical and philosophical questions. After all, if a computer is perceived to be as intelligent as a person, what is the difference between a smart computer and a human being?

HAL FOOLS LANGUAGE EXPERTS

Today's chatbots -- a computer program that has a persona and a name and chats with you -- are incapable of dealing with changes in context or abstract ideas and succeed only at momentarily tricking people regurgitating pre-programmed answers.

But Hal has fooled child language experts into thinking he is a toddler with an understanding of about 200 words and a 50-word vocabulary which he uses in short, infantile sentences.

"Ball now park mommy," Hal tells Treister-Goren, then asks her to pack bananas for a trip to the park, adding that "monkeys like bananas," a detail he picked up from a story on animals in a safari park.

When Hal was "born," he was hardwired with nothing more than the letters of the alphabet and a preference for rewards -- a positive outcome -- over punishments -- a negative one.

The pre-programmed preference for rewards makes Hal strive for a correct response. Treister-Goren corrects Hal's mistakes in her typewritten conversations with him, an action Hal is programmed to recognise as a punishment and avoids repeating.

Named after the smooth-spoken computer Hal 9000 from the science fiction cult film "2001: A Space Odyssey," the scientists and language specialists at AI see Hal as the first step towards the computer of the movie.

"All of us strongly believe that machines are the next step in evolution," said Dunietz. "The distinction between real flesh and blood, old-fashioned and the new kind, will start to blur."

Dunietz's ambition is to develop a computer that functions as an assistant, doing all sorts of time-consuming chores.

Going to Japan for a holiday? The computer will book your ticket, choose your seat on the plane, organise a hotel and arrange for a rental car to await you at the airport.

"We can have a personal assistant, a slave, a friend who doesn't really suffer by being delegated these tasks," he said.

You will not need a mouse or keyboard to operate the computer as it will function when you converse with it.

"It is going to be the next user interface, the last user interface," Dunietz said, explaining that it will replace the mouse, computer pointing devices and the Microsoft Windows environment.

"Machines will be extremely human-like in many respects and particularly the most important respect which is the ability to communicate like humans," Dunietz said.

THE STUFF OF SCIENCE FICTION

For years intelligent computers have featured in Hollywood productions such as "Star Trek" and most recently in Steven Spielberg's "A.I. Artificial Intelligence," a film about a robot boy called David who dreams of being human.

But intelligent machines have remained the domain of science fiction books and movies even though AI's chief scientist Jason Hutchens believes the computer technology of today is powerful enough to produce artificially intelligent computers.

"It's just that we don't know the secret yet," said Hutchens, an Australian who won the prestigious Loebner artificial intelligence prize in 1996.

"Our goal is the holy grail of artificial intelligence, it's to get a computer program that can use language," he said. The idea is to educate Hal gradually, the way a child learns, through trial-and-error and rewards when he performs well.

Hutchens explains that most artificial intelligence projects involve programming a set of language and grammatical rules and inputting thousands of pieces of information that we take for granted such as a table has four legs, or apples grow on trees.

"Those systems just haven't exhibited the kind of creativity that Hal has exhibited and they certainly don't learn from their experience," he said. "If we were to compare them head-to-head in a language learning scenario, Hal would come out on top."

What Hutchens's team of mathematicians, computer scientists and engineers are doing is essentially reinventing the wheel. They have adopted Turing's concept of creating a child computer and raising it to be an "adult computer."

"Instead of trying to produce a programme to simulate the adult mind, why not try to produce one which simulates the child's? If this were subject to an appropriate course of education one would obtain the adult brain," Turing wrote.

AI is doing this by combining Hutchens's team of scientists with Treister-Goren's team of learning specialists and linguists in a high-tech environment near Tel Aviv.

WILL COMPUTERS TAKE OVER THE WORLD?

Hutchens believes it will take about a decade to develop Hal's language and communications skills from that of a toddler to an adult. In the meantime Dunietz hopes to start producing primitive versions of Hal by as early as next year.

"We believe that human beings are complicated machines, computers are also machines, and we should be able to do with computers what human beings can do," Hutchens said.

The firm's philosophy is simple. If it looks intelligent and it sounds intelligence, then it must be intelligent.

"If you perceive other people are intelligent without knowing how their brains work and if you were to meet a robot that is indistinguishable in human appearance and indistinguishable in behaviour then you would think it was a human being," Hutchens explains.

Science fiction aficionados are aware of the potential downside to Hal, whose namesake in Stanley Kubrick's "2001: A Space Odyssey" killed off most of its crew during a space mission.

"Every technology which is very significant, very powerful, has a lot of potential to change things is equally dangerous as it is promising," said Dunietz, who believes his Hal will be a non-menacing version of Kubrick's computer and will be the first intelligent machine.

"These new entities are going to be more human than human. They are going to be pro-human to the extent that they will take themselves as such," he said. "They will be human-like."

22:04 08-19-01

Copyright 2001 Reuters Limited.  All rights reserved.  Republication or redistribution of Reuters content, including by framing or similar means, is expressly prohibited without the prior written consent of Reuters.  Reuters shall not be liable for any errors or delays in the content, or for any actions taken in reliance thereon.