Loving the robot: How thinking computers can make us better people

As computers get smarter, what role is left for humans? Author Brian Christian explains how intelligent robots make us better...

If computers can beat the Turing Test, what does that mean for the future role of humans? Photo: Shutterstock

It might not be as dramatic as a last-ditch battle against relentless killer robots, but it might be just as decisive for humanity - could we be only five minutes of idle chit-chat away from domination by machines?

For decades, computers have been chipping away at the range of skills that can only be mastered by humans. From performing calculations to beating grand masters at chess, composing poetry and music, talents previously considered 'humans only' have been replicated by computers.

Indeed, so great is the list of machines' achievements that only the biggest challenge remains to them: can a computer really be said to think?

It is this question that the Turing Test attempts to answer - with the unlikely tool of small talk.

The test was first proposed by mathematician Alan Turing in the 1950 paper Computing Machinery and Intelligence. The test, which he describes as "the imitation game", was his attempt to set criteria for judging whether a computer really can think.

In the most modern version of the test, judges talk via a computer to two unseen correspondents - one a human, one a piece of software - and then have to decide which is which.

Turing expected that by the year 2000, computers would be able to fool one in three judges into thinking the machines were human after five minutes of conversation, and that this would mean "one will be able to speak of machines thinking without expecting to be contradicted."

So far, that prediction hasn't come to pass, but the software has got a lot closer to snatching victory in the last few years.

The test is now run annually as the Loebner Prize, and takes place this year in October at the University of Exeter, with a grand prize of $100,000 and a gold medal for the first computer whose responses are indistinguishable from humans'. "Such a computer", says the Loebner prize website, "can be said 'to think'."

Brian Christian took part in the Loebner Prize in 2009 as one of the human 'confederates' and was so convincingly human in his conversation with the judges that he won the 'Most Human Human' award that year.

His book The Most Human Human is about his experiences of taking part in the Turing Test itself, and how to win at it. Although it is subtitled 'A Defence of Humanity in the Age of the Computer', it is also a philosophical - and occasionally practical - exploration of what it means to be human in a world where machines are increasingly displaying skills that were considered to be the sole preserve of mankind.

Machines are increasingly encroaching on the human domain, evidenced by the fact that every development in artificial intelligence (AI) has led to a narrowing of the definition of what it means to be human, says Christian.

"This has been an ancient question of what...

...makes humans different and unique. The computer complicates that. Typically what happens is that with each breakthrough in AI, it means that we narrow our definition of thinking," he told silicon.com.

Author Brian Christian, the 'Most Human Human'Photo: Michael Langan

However, it's not that the Turing Test implies the ability to hold a conversation for five minutes is the best definition of what it is to be human, says Christian, but it is a simple and effective route into a very complicated area.

"The Turing Test has come to resemble chit-chat but Turing's original vision was more of a kind of interrogation. Part of why Turing chose language is because he felt it could encapsulate so many different kinds of tests. This is something that's a little closer to home for us because we aren't all professional chess players but we all engage in the act of talking."

For Christian, the Turing Test is as much about how we choose to define being human as it is about building smart natural language-processing software. For example, computers are very good at imitating but pretty terrible at innovating, at understanding nuance, context and ambiguity and these are the characteristics we should celebrate in humans.

Can you win if you don't understand winning?

According to Christian, there are a number of major differences between humans and computers that mean there is still a long way to go before we will be reduced to shining the shoes of our robotic overlords.

Perhaps the fundamental difference is that while a human player would be disappointed to lose or overjoyed to win, software doesn't have that same drive.

For when the Deep Blue chess computer famously beat Grandmaster Garry Kasparov, although it played a technically better game, can it be said to have 'won' without understanding the concept of winning and simply executing a program as requested?

As Christian puts it: "We have to decide what is important and set our priorities. The computer is in the opposite position. Deep Blue doesn't make up its mind to care about chess - it's just told to. They don't have...

...curiosity or drive - or boredom."

The changing definition of humanity

According to Christian, our definition of what it is to be human has, over the last 2,000 years or so, moved increasingly away from the body, to the brain, and then specifically to focus on the left brain - the rational calculating bit. Similarly, the failure of AI to replicate a human level of experience seems to suggest that having a physical body is key to being able to think.

The only robots we should fear are the ones we turn ourselves intoPhoto: Shutterstock

"One of the things we are learning at a philosophical level for AI is the cardinal importance of the body. Plato and Aristotle tend to regard the body as an ox the mind must be yoked to, Descartes is famously a brain in a vat. AI has shown us what's wrong with these conceptions. As human beings we really do mediate our experiences through our bodies and senses," he told silicon.com.

For computer scientists keen to bridge the gap between human and computer, the absence of the body is an issue that must be overcome, Christian says.

"At the moment there are just these disembodied systems, they just sit there awaiting orders. We are going to have to put it in the world. That's not the paradigm of our software and that would take a pretty major reworking."

Learning from experience

As well as the question of the body, another gulf exists between man and machine - the notion of experience. With a computer, "you have the program and the data and the program acts on the data and changes it. The data does not modify the program", Christian says. And part of the human experience is that our experiences change us, change our programming - for better or worse.

And that's the optimistic conclusion of his book - that even if sometime soon a chatbot manages to beat the Turing Test judges, for Christian this isn't the end of the line for humanity. "We think that this man-versus-machine battle is a once-and-for-all thing. Having the experience of losing will not be a line in the sand, it gives us a chance to come back and raise the level of the game."

At one point in the book, Christian talks about why jobs become automated and says: "There's a crucial intermediate phase to that process: where humans do the job mechanically". So blame us, not the computers.

"I prefer to think of the long-term future of AI as neither heaven nor hell but a kind of purgatory: the place where the flawed, good-hearted go to be purified - and tested - and to come out better on the other side," he concludes.

That's the real message of the book. Rather than fear the day computers are able to think, we should make sure we celebrate and use the talents and skills that make us most human - the ability to innovate, to engage with others and create new experiences.

The only robots we really have to fear are the ones we turn ourselves into.