Introduction to AI Trick:
In 2006, Robert Epstein was on a quest for love and AI Trick took his search online. As detailed in the journal Scientific American Mind, he began an email exchange with a charming brunette from Russia. Epstein, seeking more than just a penpal, was drawn to her warm and friendly demeanor. She soon expressed a budding affection for him, writing, “I have very special feelings about you. In the same way as the beautiful flower blossoming in mine soul… I only cannot explain… I shall wait your answer, holding my fingers have crossed…”
Their correspondence flourished, yet Epstein noticed that Ivana never directly addressed his questions. Instead, she shared details about her walks in the park, conversations with her mother, and affectionate sentiments toward him. Growing suspicious, Epstein sent Ivana a nonsensical line of gibberish. She responded with another unrelated email about her mother.
Eventually, Epstein realized the truth: Ivana was a chatbot.
The surprising twist isn’t just that a lonely middle-aged man was fooled by a Russian chatbot. It’s that Epstein, a founder of the Loebner Prize—a contest testing artificial conversation where computers aim to trick humans into thinking they’re conversing with another person—had spent two months trying to woo a computer program.
The Loebner Prize, an annual event, challenges chatbots to pass the Turing test, proposed in 1950 by British mathematician, codebreaker ,AI Trick, and computer pioneer Alan Turing. In Turing’s “imitation game,” a judge would communicate with both a human and a computer via teleprompter. The computer’s goal was to imitate human conversation convincingly enough to fool the judge. Turing predicted that within 50 years, computers would be able to deceive 30% of human judges after five minutes of conversation. While it actually took 64 years for a program named “Eugene Goostman” to claim success in 2014, debates continue over its validity. Eugene, like Ivana, claimed not to be a native English speaker, presenting itself as a 13-year-old Ukrainian boy.
One of the earliest and most renowned chatbots, Eliza, developed in the mid-1960s by Joseph Weizenbaum, wouldn’t have passed the Turing Test but managed to imitate a human non-directive therapist with just a few lines of code. Named after the character Eliza Doolittle from George Bernard Shaw’s Pygmalion, Eliza would respond to inputs like “my husband made me come here” with reflective questions such as “your husband made you come here?” or “do you think coming here will help you not to feel angry?” Despite its simplicity, users were pleased with Eliza’s non-judgmental listening, even requesting privacy to converse with it.
Psychotherapists were intrigued by Eliza. An article in The Journal of Nervous and Mental Disease speculated that computers could handle hundreds of patients per hour, enhancing efficiency under human supervision. Today, cognitive behavioral therapy is administered by chatbots like Woebot, designed by clinical psychologist Alison Darcy, with no pretense of being human.
Weizenbaum, however, was dismayed by the idea of people settling for a subpar substitute for human interaction. Yet, like Dr. Frankenstein in Mary Shelley’s novel, he had created something beyond his control. Now ubiquitous, chatbots handle a growing number of tasks from medical advice with Babylon Health to customer interactions with Amelia. Voice-controlled assistants like Amazon’s Alexa, Apple’s Siri, and Google Assistant aim to simplify our interactions with technology.
Brian Christian, author of The Most Human Human, observes that most modern chatbots don’t attempt to pass the Turing test. Instead, they specialize in specific tasks, from facilitating extramarital affairs on Ashley Madison to engaging in political arguments on social media. Most chatbots, however, are content to present themselves as such, focusing on efficiency rather than human-like conversation.
Economist Adam Smith’s principle of labor division into specialized tasks is echoed in modern chatbot design. By taking over routine tasks, chatbots free humans to provide creativity and adaptability, as seen with digital spreadsheets, ATMs, and self-checkout kiosks. However, there’s a risk that we might conform to the limitations of computers, preferring automated interactions over potentially enriching human exchanges.
Christian suggests we should embrace the challenge to elevate our own conversational skills, allowing chatbots to handle mundane tasks and freeing us for more meaningful human interactions. Better chatbots could save time and enhance the quality of our communications.
The author writes the Financial Times’s Undercover Economist column. 50 Things That Made the Modern Economy is broadcast on the BBC World Service. You can find more information about the programme’s sources and listen to all the episodes online or subscribe to the podcast.