Chatty algorithms: the new A.I. software comes to business
Chatterbots are getting more and more advanced, and are finding new applications in the business world
“Je suis Clara et je suis a votre disposition si vous avez des questions.” Clara is kind and helpful. She talks in French because she is speaking to the customers of FNAC, the Parisian malls, as part of their online sales service. Clara can help you with your purchases, or entertain you for a while as you browse the website. But if you try to get her know better, like with some call centre operators, you'll discover something's up: suddenly Clara stops answering your questions and puts you to a human operator. Because she isn't one. She's a chatterbot (also known as “chatbot”): a "robot", a software able to hold simple conversations, which detects keywords in the questions asked by the customer using them to generate answers. Usually these bots work on a textual basis, but newer models are being developed, able to decode spoken language, and answer in a natural, believable voice, miles ahead from the computerised sound patterns of their early manifestations. Progress is being made in all areas: voice recognition, text to speech, semantic learning, pattern recognition.
But how different is it to speak to a chatbot from speaking to a real human being? Would they pass the Turing Test – the most widespread standard to assess whether a machine is being run by a human being or a simple software? In 1950, English mathematician Alan Turing (known to the general public as the protagonist of recent biopic The Imitation Game) suggested a simple procedure to figure out whether software could be truly deemed “intelligent”.
The first chatterbox was conceived 15 years later, as a challenge to the Turing Test, by German computer engineer Joseph Weizenbaum: ELIZA was the object of much attention and debate. For the first time a software item seemed to be able to hold a human conversation.
For years, chatterbots were treated as a sort of cybercuriosity, but in time they evolved to acquire a sense of purpose. Now chatterbots are everywhere: from Instant Messaging platforms to videogames, and especially on the Internet, where they are readily plugged into the powerful web to provide relevant information, such as weather forecasts, events calendars, news or financial information. After ELIZA came SmarterChild, a new intelligent bot created by ActiveBuddy in the early 2000s; it could sustain extended conversation via chat, assist gamers and help out with web searches.
After nearly half a century of slow progress in artificial intelligence, we might be finally moving in the direction of employing chatterbots for business. What's making a difference are the new techniques of “machine learning” employed in developing Siri, Apple's personal assistant and Deep Mind by Google. Both apps are able to recognise a voice or read a picture without having to go through an initial stage of training, by plugging themselves straight into the Big Data pool generated by user activity on social networks. The idea is that chatterbots take over company customer care, which should mean guaranteed assistance to a broader number of users, as well as saving time – both unachievable on the same level via human operators.
The future is finally among us. In June, Google announced they are currently working on a last generation chatterbot. This version will not only be able to identify keywords in speech, but also to answer entirely new questions – a sort of new, machine-generated creativity. Researchers decided to test the new system by asking it a series of existential questions.
They achieved surprising results, such as:
HUMAN: What is the purpose of life? COMPUTER: Living forever.
Or strange answers such as:
HUMAN: Where are you? COMPUTER: I'm right in the middle of nowhere.
With understandable limitations:
HUMAN: What is the purpose of emotions? COMPUTER: I don't know.
And finally, a fair nod to the competition:
HUMAN: What do you think of Bill Gates? COMPUTER: He's a good man.
RELATED STORIES
