11th August 2016

When Alan Turing published his, now seminal, paper Computing Machinery and Intelligence in 1950, he fully expected that by the end of the century, a computer would have been built capable of passing his 'Turing test' - whereby the computer is able to converse convincingly as a human.

Yet - although substantial progress has been made in the field of artificial intelligence in recent decades - there is still no sign that a computer will be able to pass for a human any time soon.

Aside from in the world of science fiction, machine learning has been notably one-dimensional. In other words, we have been successful in creating machines to do one very 'intelligent' thing at a time, eg become the best chess player or GO! player in the world - but the same computer can’t do both.

Computer scientists have a number of theories as to why our progress has been limited and, correspondingly, different takes on how to respond to Turing’s challenge.

On the one hand, those such as Steve Furber CBE, ICL Professor of Computer Engineering at the University of Manchester's School of Computer Science, argues that the reason we are still so far from creating a convincing copy of the human brain is that our understanding of what is happening in the brain is still so limited.

In other words, we still don't understand what natural intelligence is, so it should come as no surprise that we haven’t been able to model it in a machine.

Furber's SpinNNaker (Spiking Neural Network Architecture) project - aka the 'brain box' - is an attempt on a massive scale to build a computer that imitates the workings of the brain. Its artificial neural network is a massively parallel processing (MPP) system designed to incorporate a million ARM processors and improve our understanding of brain function.

At the same time, it seeks to discover more efficient parallel, fault-tolerant computation - which will be needed to enable a major scientific breakthrough in this field.

Furber is no stranger to scientific breakthroughs - having been one of the team behind the design of the BBC micro computer and the designer of the highly power efficient ARM microprocessor during his time at Acorn Computers (now ARM Holdings) - a breakthrough which puts him in the company of Bill Gates, Tim Berners Lee, Vint Cerf and Tom Kilburn for his contribution to the IT industry.

Nonetheless, there are those who argue that the venture of modelling the human brain is quite separate to the development of 'smarter' computers.

It’s still far from clear whether we will ever successfully replicate the brain. (Despite the scale of SpiNNaker, it models around 1 million neurons: just 1% of the human brain.) However, the argument goes that there is no need to let this get in the way of the development of artificial intelligence (AI).

That's because AI in the real world is not about making computers which can 'think', but about making computers which can 'learn'. It relies on highly sophisticated pattern recognition breakthroughs for specific tasks to produce commercially viable systems.

Following this approach, machines learn through a process where a task is defined, a mathematical model is created, a human supervisor gives feedback to the system and then a finished model emerges.

This relies on a combination of very fast computers processing vast amounts of big data, and the input of engineers providing feedback - tweaking the algorithms - until the machine has 'got it'.

In recent years, this has led to impressive technologies emerging, with one of the most exciting and topical examples being the pursuit of self-driving cars, which have so far clocked millions of miles on the roads with minimal collisions.

However, it is not the same endeavour as attempting to understand and model more abstract notions such as intelligence or consciousness - which would arguably be required if we truly wanted to make a machine that can think.

So while specific industries, job functions and areas of life, have been significantly disrupted by new technologies, this has so far been very much by human design.

And while we await with bated breath the next generation of computers to come out of Manchester, it is perhaps reassuring that the prospect of our being usurped by robots in the near future seems unlikely.