Slack-jawed over the possibilities
Slack-jawed over the possibilities. The BBC seemed to think so.

Artificial Intelligence and the Internet in 1997: The Future Was Already Knocking

November 1997 was an interesting moment in the evolution of computers and communications. The Internet was moving out of its infancy and beginning to establish itself as something considerably more important than an electronic encyclopedia or a convenient way to exchange messages. It was becoming a commercial marketplace, a gathering place for communities, and a new environment for human interaction. At the same time, researchers were pursuing a question that had fascinated computer scientists for decades: Could a machine carry on a conversation convincingly enough to be mistaken for a human being?

These were among the subjects explored in the November 12, 1997 edition of the BBC World Service program Soundbites. The discussion reflected a period when the possibilities of computer technology seemed almost limitless, even though much of what would eventually become commonplace was still experimental.

The Internet of 1997 was a very different place from the one we know today. Access was frequently through dial-up connections, using telephone lines and modems that announced their arrival with a familiar sequence of electronic squeals and clicks. The World Wide Web was expanding rapidly, but much of its content remained relatively static. Email, bulletin boards, newsgroups, online discussion forums and chat rooms provided the principal means of social interaction. America Online and CompuServe were introducing millions of people to the experience of communicating with strangers and exchanging information across geographical boundaries.

There were already indications that this electronic environment could become a powerful social and commercial force. Businesses were establishing websites, experimenting with online advertising and discovering that products and services could be marketed to customers without the traditional limitations of geography. Amazon and eBay were among the companies helping establish the foundations of electronic commerce. What had once been a network primarily associated with universities, government institutions and computer enthusiasts was becoming a commercial proposition.

The social dimension was developing alongside the commercial one. In 1997, SixDegrees.com appeared, offering users the ability to create personal profiles and establish connections with other people. It was an early example of what would eventually become social networking. The terminology and technology of today’s social media did not yet exist, but the underlying idea was already taking shape: computers could become instruments not simply for obtaining information, but for creating and maintaining relationships.

But perhaps the most intriguing part of the 1997 discussion concerned artificial intelligence and the possibility that computers might eventually become capable of conducting meaningful, human-sounding conversations.

The subject had received considerable attention following the remarkable success of IBM’s Deep Blue computer, which defeated world chess champion Garry Kasparov in May 1997. Although chess-playing computers had been around for years, Deep Blue demonstrated that machines could perform at an extraordinarily sophisticated level in a contest traditionally regarded as a measure of human intellectual ability. The question now was whether computers could move beyond calculation and demonstrate something resembling human understanding.

One of the most interesting experiments in that direction was the annual Loebner Prize, established by Hugh Loebner in 1990. The competition was based on the principles of the Turing Test, proposed by British mathematician Alan Turing in 1950. Its premise was relatively straightforward: if a person communicated with a computer through a written conversation, could that person distinguish the machine from a human participant?

The 1997 competition produced an especially interesting result. The winning entry was CONVERSE, a conversational artificial intelligence program developed under the direction of David Levy of London’s Intelligent Research Ltd., working with a team that included researchers from the University of Sheffield. The program presented a simulated human personality named Catherine, a 34-year-old British journalist living in New York.

Catherine was judged the most human-like computer in the competition, earning Levy and his team the $2,000 prize and bronze medal.

What made Catherine particularly interesting was her ability to introduce contemporary subjects into a conversation. Rather than simply waiting for questions and supplying prepared answers, she could initiate discussions, make observations and attempt to steer the exchange toward subjects with which she was familiar.

One memorable example involved President Bill Clinton and a recent White House gathering involving comedian Ellen DeGeneres and a lesbian couple. Catherine brought up the news herself, asking judges whether they had seen the television coverage. This gave the impression of a conversational personality rather than a machine simply responding to a series of programmed questions.

CONVERSE drew upon approximately 80 conversational topics, dictionaries, information databases and programming designed to interpret questions and responses. Its Catherine personality was intended to give the system a recognizable identity and allow it to maintain the appearance of someone with interests, opinions and a background.

There were limitations, of course. Catherine could become repetitive and sometimes struggled when conversations moved unexpectedly into unfamiliar territory. She did not possess anything resembling human consciousness. Nevertheless, CONVERSE represented a meaningful advance in conversational programming and offered an early glimpse of what artificial intelligence might eventually accomplish.

Looking back, the significance of the November 1997 Soundbites discussion is that it captured several developments that would eventually converge into the digital world we inhabit today. Artificial intelligence was beginning to explore human conversation, the Internet was becoming a commercial enterprise, and online communities were laying the groundwork for social networking.

In 1997, these developments were still largely separate experiments. Nobody could have known precisely how quickly they would evolve or how completely they would transform everyday life.

But the essential questions were already being asked: Could computers understand us? Could people conduct their social lives through electronic networks? Could commerce operate successfully without a physical storefront? And might machines eventually become companions rather than merely instruments?

Nearly three decades later, those questions have become central to our everyday experience. The remarkable thing about revisiting 1997 is discovering just how much of the future was already visible to those who were paying attention.

Here is that episode of Soundbites from The BBC World Service of November 12, 1997.

Donate now at Fractured Atlas! Buy Me A Coffee