Alan Barker is a writer, coach, training consultant and academic proofreader. Find out more about his work here.

When it comes to artificial intelligence – or, indeed, any kind of intelligence – Margaret Boden probably knows as much as anyone does. She is Research Professor of Cognitive Science at the University of Sussex; at this year’s British Science Festival, in conversation with Jon Turney, she reflected on a career that spans almost the whole history of AI. She took her first degree in 1958, only a few years after Alan Turing’s pioneering work in the field. According to some predictions, she may well witness the Singularity: the much-anticipated moment when AI exceeds man's intellectual capacity and makes humans redundant. (Apparently, we’ve got ten years.)

In the early sixties, Ms Boden studied philosophy under Margaret Masterman, who ran the Cambridge Language Unit. Masterman wanted to create a machine language. “She understood,” said Ms Boden, “that it would need to do more than translate concepts or objects directly into code. Any sophisticated machine language would have to become adept at contextual sense-making.” Masterman’s work was fundamental to the development of modern search engines as one example.

“I didn’t quite get the point until a couple of years later,” Ms Boden told us. “I’m interested not just in reasoning, but in the nature of the human personality. I didn’t see how what [Masterman] was doing could relate to those issues.” And then, by chance, in a New York bookshop, she happened on a “whacky and insightful book” called Plans and the structure of behaviour. “Within five minutes, my whole intellectual life was changed.” Suddenly, she recognized the concepts we need to understand the human mind.

And those concepts are computational. Ms Boden believes that, unlike the jukebox or the telephone exchange, the computer is not just another metaphor for the mind. Computation, she suggested, is to psychology and neuroscience what mathematics is to physics. What we’ve witnessed in the last half-century is an explosion of computing power to allow those concepts to be implemented.

But understanding the mind is one thing; building a machine to mimic it is quite another. The apps on our mobiles use symbolic AI, which instructs machines in logical processes: do this, then do that. But another approach builds AI in terms of networks and systems, modeled very broadly on the brain.

From the start, the grail has been AGI: an artificial general intelligence that could successfully perform any intellectual task that a human being can. Ms Boden remains sceptical. “If there’s one lesson that we’ve learnt from our work on AI,” she told us, “it’s that the human mind is vastly more powerful and subtle than any theoretical psychologist could have imagined.”

In her opinion, there are types of problem-solving humans can do that AI will never do. Consider this, for example: the kind of question posed in Jeopardy, a popular US television game show. “What hotel heiress might live in a European capital?”  Finding the answer – Paris Hilton – requires a type of knowledge that’s hard to programme. 

Cue Watson, an AI business platform, and the latest Grand Challenge run by IBM Research. The predecessor of this AI platform, Deep Blue, famously beat Garry Kasparov at chess; beating a human being at Jeopardy was the closest IBM could come to attempting the Turing Test, in which AI tries to fool a human being into thinking they’re having a conversation with another person.

One Jeopardy question asked for the two names of Jesus’ disciples that both end in the same letter and are top-ten baby names. IBM’s Watson duly found the right answer: Matthew and Andrew. One of his human competitors, Ken Jennings, however, had been about to suggest James and Judas: “But I don’t think Judas is a popular baby name, for some reason...”

Watson, says Ms Boden, couldn’t have done that. It doesn’t, in her words, “have the realisation of the hopes and values being expressed when somebody chooses a baby’s name.”

What sort of cleverness, then, is unique to human thinking? Ms Boden insists that it’s not magic. The processes are, in principle, open to scientific understanding and could be simulated in a computer. But they are so complex that producing an AGI with that degree of understanding is very unlikely. Researchers are simply not studying the right sorts of problems.

We can welcome, then, the superintelligence that already processes big data to solve practical, domain-specific problems, like managing traffic flow or diagnosing different kinds of cancer (which IBM is currently developing Watson to do). But we never quite reached the Singularity. And Ms Boden assured us that we almost certainly never will. “The robots won’t take over,” she asserted confidently, “because we’re not going to give them the motivation to want to.”

Follow the British Science Festival on Twitter: @BritishSciFest