Monash

Discovery

A new theory is spurring scientists to attempt to build a machine with artificial consciousness

Dror Cohen

22.01.2016

In a nutshell: The most remarkable, puzzling aspect of the human brain is how it gives rise to the conscious mind. Could recreating consciousness in a machine help solve the puzzle?

A new theory is spurring scientists to attempt to build a machine with artificial consciousness

The Essentials

  • consciousness is the most puzzling aspect of the human brain with theories to explain it currently based on a smattering of data, and huge amounts of untested theory to tie them together. Some refer to it as the neuroscience’s Big Bang theory
  • Integrated Information Theory (IIT) is one theory trying to break the intellectual log-jam. Scientists are using it to attempt to build a machine with artificial consciousness
  • if they succeed it will raise a host of ethical issues, such as does such a machine get human rights

Humans, and some animals, have something machines do not – we are conscious. The most remarkable, puzzling aspect of the human brain is how it gives rise to the conscious mind. Colours, smells, textures and sounds are not only processed but experienced by us. Add to this our experience of memories, feelings, and so on and you get a sense of the phenomenological experience that is at the heart of being conscious.

Until recently, artificial consciousness received little attention compared to its popular cousin, artificial intelligence — the sort of computational grunt that means the world’s best Jeopardy! player is now an IBM computer called Watson.

But a new theory is spurring scientists to take a fresh look.

To cut to the chase, Integrated Information Theory (IIT) claims that conscious systems are systems that can be in many different states (high information), and in which the states interact in a particular, complex way (integrated). This is the reason that the human brain, with many billions of neurons that are intricately connected, is conscious. The overwhelming complexity of the human brain makes the modern digital computer look simple — which is why, IIT claims, Watson may be smart, but is not conscious.

IIT is not the first, or only, theory of consciousness, but its breadth and depth go beyond previous accounts. IIT, which includes a series of mathematical formulae, unifies several known facts about consciousness that have so far been difficult to reconcile — for example, why certain parts of the brain are critical for consciousness, and others are not.

Though IIT can be regarded as a leading theory, it’s not all smooth sailing. Validating IIT is difficult from the computational, theoretical and experimental perspectives, so it will be some time before its predictions are rigorously tested.

Despite this, IIT has spawned a new method for assessing levels of consciousness. It is based on quantifying the complexity of the neural response to an external Transcranial Magnetic Stimulation (TMS) to give an estimated proxy for integrated information. Although still experimental, the method is being used to assess depth of anaesthesia, and to work out whether people are in a minimally-conscious state, with some awareness remaining, or in a vegetative state.

IIT is also opening up new research directions for artificial consciousness. Because the maths underpinning IIT describes exactly what is required for a system to be conscious it is possible to predict which systems will be conscious and to build them.

Scientists based in Japan are taking this idea seriously. The team is led by Ryota Kanai at Araya Brain Imaging in Tokyo, joined by an Australian team led by CIBF associate investigator Nao Tsuchiya of Monash University. They are funded by the Japan Science and Technology Agency’s Core Research for Evolutionary Science and Technology (CREST) program to construct machines that will be assessed in terms of their abilities to set their own goals (autonomous ability) and adapt to changing environments (flexibility). These markers of consciousness point to the possible advantages of artificial consciousness over Watson-style artificial intelligence.

The research program got a boost from the Tsuchiya team’s recent development of a new way to measure how a system integrates information between its individual parts.

If ever more complex machines one day make artificial consciousness a reality will we be ready? Would we grant such machines human or animal rights? Would it be ethical to put them to work in the same way as we use traditional machines? These are some of the questions we need to start pondering.

And when Watson finally gives up its crown will it be to a conscious machine?

 

Dror Cohen is a PhD student studying the neuroscience of consciousness at Monash University.