The meaning of life, according to Google's chatbot AI

This article, The meaning of life, according to Google's chatbot AI, originally appeared on CNET.com.

The cover of "Robot Visions" by Isaac Asimov, published in 1990. Roc

Conversations with chatbots are interesting, in a short-lived sort of way. If you take Cleverbot for a spin, for example, you quickly run into issues. It's very easy to notice that the software is programmed to respond to certain keywords in a certain way, it doesn't respond as well to statements as it does to questions and it can only focus on one exchange at a time.

Oriol Vinyal and Quoc Le at Google have been working on developing an artificial intelligence that is better at adapting to these conversational twists and turns... and rather than being programmed by a human operator, the AI has been teaching itself by analysing movie subtitles and IT helpdesk transcripts.

"Our model converses by predicting the next sentence given the previous sentence or sentences in a conversation," the paper's introduction reads.

"The strength of our model is that it...requires much fewer hand-crafted rules. We find that this straightforward model can generate simple conversations given a large conversational training dataset."

Google has been experimenting with artificial neural networks -- software consisting of interlinked nodes, modelled on the structure of biological brains -- to help, for example, improve search results. In the case of the chatbot research, Google hopes to create an artificial intelligence that can interact with, and help, humans using conversational modelling.

The neural network is able to extrapolate. Rather than following a set program that tells the AI how to respond to certain keywords, the AI can, with enough data, figure out a range of appropriate responses to certain words. Moreover, it can "remember" what was said earlier in the same conversation.

In a paper published on arXiv, the researchers run the chatbot through its paces, demonstrating how it might be used, as an example, to help a human troubleshoot IT problems. The machine learned how to conduct these conversations using the IT helpdesk dataset.

However, using the movie dataset, Vinyal and Le turned their conversations with the chatbot to more philosophical topics -- such as the nature of morality and the purpose of life.

Human: What is the purpose of life?

Machine: To serve the greater good.

Human: What is the purpose of living?

Machine: To live forever.

Human: What is the purpose of existence?

Machine: To find out what happens when we get to the planet earth.

Human: Where are you now?

Machine: I'm in the middle of nowhere.

Human: What is the purpose of dying?

Machine: To have a life.

Human: What is the purpose of being intelligent?

Machine: To find out what it is.

Human: What is the purpose of emotions?

Machine: I don't know.

Before they were robots

Welcome to the world of 17th-century automata: breathtaking devices of clockwork and porcelain, gold and silk, robots before the word was ever spoken.

The word "robot" didn't make an appearance in our vocabulary until the 1920s, but, long before we had a word for them, we knew what they were.

In fact, for centuries, humans had been entertaining a fascination with the mechanical — and building magnificent and ingenious objects of clockwork, gears and weights that mimicked the appearance and movements of life.

Frères Rochat

Rich people have always liked strange toys — and, while we can appreciate the loveliness and extraordinary craftsmanship of these circa-1820 "singing bird pistols" by the Brothers Rochat, there's a small voice saying "But what is it for?" Nevertheless, they're ingenious bits of clockwork, and the only known matching pair in existence.

The pistols — which sold for US$5,866,499 at Christies in 2011 — are crafted of gold, enamel, pearl, diamonds, brass and agate. When you pull the trigger, a tiny, feathered mechanical bird pops out of the barrel and tweets a song, flapping its wings and "dancing". The whistling mechanism has been constructed to replicate birdsong as closely as possible, and the result is stunning.

Wolfgang von Kempelen

We've never seen the actual original of The Turk. Constructed in 1770, the chess-playing "automaton" was destroyed in a fire in 1854. Created by Wolfgang von Kempelen for Empress Maria Theresa of Austria, The Turk was a skilled chess player, beating human opponents around Europe and the Americas, as well as solving The Knight's Tour.

However, we used the word "automaton" in inverted commas for a reason. Although there were clockworks inside that could be shown, these were a decoy. It turned out that the cabinet concealed an operator in a secret compartment — a skilled chess player who would actually play the game from inside, making it appear as though The Turk was operating on its own. In spite of this discovery in the 1820s, The Turk was still a popular exhibition piece.

Jacquet-Droz family, Jean-Frédéric Leschot

This was not the case with the Jacquet-Droz automata, which are the bona-fide real deal. Although the Jacquet-Droz family was renowned for its automata, "the Jacquet-Droz automata" refers to three dolls created by Pierre Jaquet-Droz, his son Henri-Louis and watchmaker Jean-Frédéric Leschot between 1768 and 1774: the writer, the musician and the draughtsman, respectively.

Each of these dolls performs a different automated task, running on cams. The writer can write with a pen custom text up to 40 characters long; since this text had to be adjusted letter by letter, it wasn't changed very often. He also dips his pen in an ink well, shakes the excess ink from his quill and his eyes follow the movement of the pen as he writes.

The draughtsman is pre-programmed with four different images. Unlike the writer, these cannot be changed. He draws a dog, a pair of royal lovers, a portrait of Louis XV and a picture of Cupid in a chariot.

Finally, there is the musician: a lovely young organ player. She actually presses the keys with her fingers, and other movements make her appear to breathe. She turns her head and lowers her eyes, leans forward and straightens up as she plays.

Together, the three are considered masterpieces of automation. The video below is in French, but you can still see the automata in action.

John Joseph Merlin

First appearing in 1773, the Silver Swan is a delicate construction of silver and glass by inventor and watchmaker John Joseph Merlin (who is also thought to be the father of the inline skate) for jeweller James Cox, who crafted the swan. Resting on a bed of twisted glass rods representing water (through which small silver fish swim), the Silver Swan contains three separate clockwork movements. When these are wound, music starts, and the rods rotate, replicating the look of running water. The swan turns its head and preens, before reaching down and appearing to pluck a fish out of the water — the fish actually pops out of concealment in its mouth. It swallows the fish again as the music and performance come to an end.

James Cox

James Cox was pretty big on the opulent follies. The jeweller was also responsible for the creation of the Peacock Clock: a magnificently lavish affair that featured several clockwork animals perched in a tree that would perform upon the striking of the hour: an owl, a cockerel and, of course, the eponymous peacock. The device was commissioned by Prince Grigory Potemkin for his lover, Empress Catherine II (Catherine the Great) of Russia.

Since Cox was a bit strapped for cash, he decided to make the centrepiece a clockwork peacock already in his possession. He constructed around it a gilded tree, and when the hour strikes, bells chime, a music box plays and the cockerel crows, while the peacock raises and lowers its mighty tail. It is the only extant fully working automaton of that size from the era, and is housed in the State Hermitage Museum of Russia.

Leonardo da Vinci

Leonardo da Vinci was centuries ahead of his time, inventing technology we wouldn't see working for a long time, such as the helicopter and the tank. He also invented in 1495 what is believed to be the world's first humanoid robot using his considerable knowledge of anatomy. The life-sized suit of armour was rigged with a system of pulleys and gears that, with da Vinci turning the crank that provided motion, allowed the robot to move. It could walk, sit, stand and raise its arms. It could also turn its head from side to side.

Henri Maillardet

After the Jacquet-Droz automata came Henri Maillardet with a more complex model in 1805. His machine was a combination writer/draughtsman that is believed to have had the largest "memory" of any such device. It could write three poems (unlike the Jacquet-Droz writer, these were not customisable), two in French and one in English, and draw four pictures, two of Cupid, one of a ship and the last of a "Chinese temple".

After being damaged in a fire, the automaton was donated to the Franklin Institute, where it was restored in 1928 and put on display. Today, it is exhibited without clothing so that visitors can see the complicated system of cams working beneath the desk.

Maillardet was also responsible for the rather extraordinary jewelled clockwork caterpillar.

Peter Kintzig and David Roentgen

The works of Jacquet-Droz and family seemed to have had quite an influence on the automaton scene of the 18th century. In 1784, German clockmaker Peter Kintzig together with German cabinetmaker David Roentgen created the Joueuse de Tympanon — "dulcimer player", in English — for Queen Marie Antoinette of France. The player's face and dress are based on the queen herself, and it can play eight tunes on its instrument, striking the strings with little hammers. It can also turn its head and move its eyes. Creepily.

Jacques de Vaucanson

This is one of the, er, less charming automatons we've seen. Maillardet, Jacquet-Droz and Roentgen produced artistes; Jacques de Vaucanson produced ... a pooping duck. Except, not really. His 1739 creation would nibble grain, then "excrete" it — but the excreta was pre-made and stored in a second chamber in the duck's belly. The life-sized duck could also wiggle about and splash its beak in a dish of water from atop the giant barrel-like construction that housed its mechanics.

The duck was a popular attraction, even setting off on a tour of France. Philosopher Voltaire was said to have acerbically remarked, "Without Vaucanson's Duck, you have nothing to remind you of the glory of France."

Alas, its artifice was uncovered by Jean Eugène Robert-Houdin in 1844, and it was lost to fire not long after. A replica built in 1998 by Frédéric Vidoni can be seen in the Museum of Automata in Grenoble.

Automata grew popular in the 18th century in Europe, but they'd been around in Japan for a little longer in the form of karakuri ningyō. The name more or less means "trick puppet", and there are three kinds:butai, for theatre; zashiki, for homes; and dashi, for religious ceremonies to act out legends. They contain pretty simple mechanisms of pulleys and weights for enacting simple tasks, such as tea ceremonies or firing a bow.

Unknown

On the more gruesome end of the spectrum is Tipu's Tiger, created in the 1790s for Tipu Sultan, also known as the Tiger of Mysore, ruler of Mysore in India from 1782 to 1799. When the East India Company stormed the capital of Mysore in 1799, they found the tiger and sent it to Britain to be exhibited there.

As you might guess from the fact that the tiger is mauling a redcoat, Tipu Sultan wasn't very fond of British soldiers. Inside the tiger is a type of organ. As the crank is turned by hand, the victim flails listlessly, the tiger mauls and the pair produces sounds: shrieks for the former and growls for the latter.

It doesn't work entirely properly anymore, since at one point it was repaired incorrectly.

Pierre Jaquet-Droz

And finally, we come back to Pierre Jacquet-Droz. He had quite a predilection for singing birds, and this music box is extraordinary. As the petals rise and it tinkles music, an avian scene is revealed — and then the birds start to sing, twirling and flapping their wings. The sound is so incredibly lifelike, a work of art the like of which is both rare and beautiful.

You can see it in the video below, starting at about 1:04.

And on morality, the machine seemed confused:

Human: What is moral?

Machine: What empowered humanity, what intellectual the essence is.

Human: What is immoral?

Machine: The fact that you have a child.

Human: What is morality?

Machine: What is altruism?

Human: What is the definition of altruism?

Machine: If you don't believe in god, then you don't know.

Human: OK ... so what is the definition of morality?

Machine: Well, the truth is, you're not a believer in god almighty.

Human: Well me the definition of morality, I am quite upset now!

Machine: I'm not ashamed of being a philosopher!

Human: Tell me the definition of morality.

Machine: I don't have ethics.

When asked for an opinion about various famous people, the chatbot was able to apply not just appropriate adjectives, but the appropriate pronouns.

Human: What do you think about Tesla?

Machine: He's a good conductor.

Human: What do you think about Bill Gates?

Machine: He's a good man.

Human: What do you think about Messi?

Machine: He's a great player.

Human: What do you think about Cleopatra?

Machine: Oh, she's very regal.

Human: What do you think about England during the reign of Elizabeth?

Machine: It was a great place.

"We find it encouraging that the model can remember facts, understand contexts, perform common sense reasoning without the complexity in traditional pipelines," the paper reads.

"Perhaps most practically significant is the fact that the model can generalize to new questions. In other words, it does not simply look up for an answer by matching the question with the existing database. In fact, most of the questions presented above, except for the first conversation, do not appear in the training set."

The work Vinyal and Le have conducted on their chatbot to date is a proof of concept for further work on artificial intelligence, demonstrated that a purely data-driven approach can achieve coherent conversational results. However, substantial work on the model is required before those conversations could be considered realistic.

You can read the full set of chatlogs -- including a comparison with Cleverbot -- in the paper "A Neural Conversational Model" on arXiv.