Before Google’s lifelike Duplex A.I., these chatbots paved the way
How sad do you have to be to have a conversation with a computer? Here in 2018, that’s not even a question any more. At Google I/O, the world’s biggest search company debuted its new Google Duplex tool: an A.I. that’s capable of making phone calls and doing things like booking you in for haircuts or securing you a table at the hottest restaurant in town. Best of all? Thanks to its hyper realistic synthetic voice — right down to the realistic, human-like “uh-huhs” and pauses — the person on the other end of the line probably won’t even realize they’re talking to a bot.
It might seem super futuristic, but Duplex is really just the latest development in the history of chatbots: A.I. systems designed to replicate human interactions. These bots have been around for decades — they just haven’t always been as sophisticated as the one Google recently showed off. Here’s a quick recap of the 10 most significant milestones we passed to reach this point:
The Turing Test
Alan M Turing and colleagues working on the Ferranti Mark I Computer, 1951 Science & Society Picture Library/Getty Images
Most people had never seen a computer in 1950. Those who had seen one typically viewed them as gigantic calculating machines that took up entire rooms and weighed more than a fully-stocked light aircraft. Back then, the idea that you might one day interact with a computer like you would a person took a pretty big speculative leap. One person willing to make that leap was the pioneering computer scientist Alan Turing.
In his 1950 paper, “Computing Machinery and Intelligence,” Turing examined the question of whether or not a machine could be capable of thinking. He proposed a hypothesis to test this: an imitation game in which a computer must try and fool a human into thinking that he or she was communicating with another person.
Turing never built an actual chatbot to test his theories, but the Turing Test has inspired many subsequent attempts to build chatbots. Most famously, it’s the basis for the annual Loebner Prize.
In the mid-1960s, an MIT computer scientist named Joseph Weizenbaum created ELIZA, a so-called computer psychotherapist. Conversations with ELIZA were carried out via text, with users typing a sentence and ELIZA then reflecting it back to them in language which either supported or questioned their statement.
Despite the fact that ELIZA didn’t actually “understand” the topics it discussed, Weizenbaum was surprised by how readily his students poured their hearts out to the computer — sharing everything from their own personal heartbreak stories to fears of academic failure.
HAL remains arguably the world’s referenced chatbot, despite only existing in the realms of science fiction. Memorably stealing the show in Stanley Kubrick’s 1968 movie 2001: A Space Odyssey, HAL is the all-knowing A.I. who controls the operations of the spacecraft Discovery One.
Unlike the text-based ELIZA, HAL is capable of interacting with users via voice recognition and natural language processing. Described as “foolproof and incapable of error,” HAL nonetheless goes wrong and, in the process, spawns one of computing’s most memorable slogans: “I’m sorry, Dave, I’m afraid I can’t do that.”
Created in 1972 by psychiatrist Kenneth Colby at Stanford University, PARRY was an attempt to model a person — in this case, a person with paranoid schizophrenia — in machine form. PARRY was a more complex chatbot than the earlier ELIZA, with a more pronounced personality and more complex conversational abilities.
At a conference in 1972, PARRY and ELIZA were connected over ARPANET, an early iteration of the internet, where they carried out the world’s first chatbot-to-chatbot conversation.
Rollo Carpenter’s Jabberwocky is an example of a smarter second generation of chatbot. Designed in 1981 with the goal of simulating “natural human chat in an interesting, entertaining and humorous manner,” Jabberwocky didn’t have many practical applications, but was nonetheless a source of great entertainment.
Particularly impressive was its ability to learn new responses and context based on its conversations, rather than relying on a static database of answers.
A bit like HAL 9000 — minus the whole murderous streak — “Knowledge Navigator” was an influential vision of computing’s future, created by Apple in 1987. Illustrated through a series of slickly-produced video vignettes, Knowledge Navigator was a proposed chatbot-based system in the form of an onscreen butler software agent.
As Apple imagined it, users would interact with the Knowledge Navigator through spoken, real world commands. Despite possessing its own personality, it was primarily intended to be used for information retrieval and executing commands.
Apple eventually delivered a form of Knowledge Navigator with Siri (which we’ll discuss later.) However, this tantalizing glimpse forward has since inspired many people outside of the Cupertino company who are interested in building productivity-focused chatbots.
Yes, we hate to include it on this list, but Clippy was likely the first chatbot that many people reading this article ever used. A cheerful dancing paperclip character who appeared in Microsoft Office 1997, Clippy popped up on-screen whenever users set out to carry out a task like writing a letter.
Clippy had the potential to be useful, but rarely managed to be. It was a limited example of the kind of A.I. agent many of us rely on today, but frustratingly poorly executed.
Seattle-based illustrator Kevan J. Atteberry, who designed the character, still has a message on his website crediting him with creating “probably one of the most annoying characters in history!” Clippy died on his way back to his home planet in 2003.
Watson is a question-answering artificial intelligence-based computer system developed by IBM. Named after IBM’s first CEO, Thomas J. Watson, IBM Watson famously won the game show Jeopardy! against former champions Brad Rutter and Ken Jennings.
After the game, Jennings held up a sign reading: “I for one welcome our new robot overlords.” Today, Watson is used for a variety of non-game show applications, ranging from drug discovery to generating recipes for cooking.
According to IBM, the project’s goal is to make computers “interact in natural human terms across a range of applications and processes, understanding the questions that humans ask and providing answers that humans can understand and justify.”
Originally started as a five-year, 500-person DARPA project, the more familiar form of Siri eventually landed on Apple’s iPhone 4s, debuting just one day before company co-founder Steve Jobs passed away.As you’re likely aware, Siri can interact using spoken commands, and is able to perform a variety of useful functions — from looking up local weather information to booking a table at a restaurant.
Even if it’s rapidly surpassed by other smarter A.I. assistants, Siri deserves credit for finally bringing the technology into the mainstream.
Amazon Echo. Google Home. Apple’s HomePod. If you don’t already own one of these talkative smart speakers, you may well soon find yourself in the minority. Aided by breakthroughs in A.I. fields like deep learning, today’s smart speakers finally live up to the promise of functional chatbots.
They can understand what you’re saying, make you laugh with a quick joke, and help you perform a million-and-one other useful functions. And, as the latest Google IO event showed, they’re getting smarter all the time.
- Apple’s latest hire just might make Siri seriously smarter
- Everything you need to know about Google Assistant
- A.I. and Google News: The push to broaden perspectives
- Here’s everything we saw at the action-packed Google I/O 2018 keynote
- Speak to me, robot: Startling Google Duplex demo raises endless concerns