Life as it could be

Matthew is based in the south wing of Building 303 at the Science Centre, situated on Auckland’s tree-lined Symonds Street. Nestled amongst the other computational sciences, and across the road from the philosophy department where he can often be found, the locale is most appropriate for Matthew, whose passion is collaborative, interdisciplinary research. “Computer Science is well suited to being a hub in interdisciplinary research because you can build a computer model of just about any system – you just have to figure out which model to build.”

Rather than focusing on making computers solve problems and ‘be smart’ like his counterparts in the field of artificial intelligence (AI), Matthew is interested in fundamental questions concerning the mind and life, and the relationship between the two. As such, he affiliates with the artificial life research community. Although connected to artificial intelligence, the goals and motivations are different.

“We look at life and take inspiration from it to create an artificial system that is life-like in some way. We then study the artificial system to learn about the biological phenomenon that inspired it.”

Recently, Matthew was awarded a Fast Start Marsden Grant for his research that will explore the role behaviour played in facilitating the very earliest stages of life’s evolution. For this body of research, Matthew is thinking about life as a dissipative structure; a concept from physics used to describe systems that use energy to maintain their highly ordered or organised state. 

“I use computer simulations and real-world experiments to investigate dissipative structures such as whirlpools or self-propelled oil droplets,” says Matthew. “These systems demonstrate fascinating, life-like forms of primitive self-preservation where they move toward environments that have the energy they need to persist and improve their chances of survival.” 

Matthew’s research aims to find out if this kind of behaviour enabled the actual origins of life. Because these dissipative structures are quite simple, evolution is not required to explain how they came to be. But despite their simplicity, they perform life-like behaviours that prolong their survival. Matthew wonders if this ability will help to explain a gap he sees in most modern explanations of the origin of life.

“The common assumption is, to get life going all that you need is some evolving chemistry, molecules that divide, replicate and evolve, and so on, and that Darwinian evolution will create increasingly complex life. Fast forward a few million years and you will then have us!” he says. “But we still don’t understand a major transition where evolving chemistry developed into integrated, behaving organisms.” 

As part of the artificial life community, he is impressed with the breadth of the topics at the conferences and workshops he attends. The diversity of topics include evolution, neural networks, and complex systems of all different kinds. The many different approaches by different disciplines to understand life is, he says, “phenomenal.”

In order to be a successful scientist in the world of artificial life, Matthew believes that being a good science communicator is key. “You’ve got to learn a lot and understand the basics in each different area and be able to communicate your ideas clearly,” explains Matthew.

In a second ongoing project, Matthew is collaborating with researchers and students in the Department of Mathematics, who are helping to analyse an artificially evolved neural network in an effort to understand the role of lag or delay in natural nervous systems. 

“In the brain, things are going on all at the same time, everything is overlapping everything else, and it is really complicated and messy. In addition there are substantial delays in the amount of time it takes a signal to get from A to B. This is different than in computers, where engineers use synchronizing clock signals and other techniques to eliminate the effects of lag.”

To better understand the role of lag in natural nervous systems, Matthew has simulated a “very simple robot” with only one neuron. The neuron has a recurrent and delayed connection to itself, where its output is fed back into the input, but with a delay.

“I used artificial evolution, also known as a genetic algorithm, to tune that one neuron to control a robot to solve a really simple task,” says Matthew. “The robot is placed in an environment where there are two stimuli, a target stimulus and distractor stimulus, and it has to move through its environment to find its target stimulus and stay there. And I know, because it only has this very simple, one neuron brain that the only way it’s going to be able to solve this problem is to somehow take advantage of its lagged recurrent connection. The artificial evolution has worked – we have the robot solving the problem, and now we are working on understanding how the problem is being solved.”

Subsequently, Matthew is mindful that people are so “awestruck” by what the brain is, that we use the latest and greatest technology as a metaphor for essentially describing what the brain does. But there is something really different between computers and us and that is our cognitive skills.

“All the time people are comparing brains to computers. For instance people will talk about ‘having their wires crossed.’ It’s ultimately an unfortunate way to perceive ourselves because if you’re ‘wired wrong,’ well then you’re just out of luck – there’s nothing that can be done. But we are not computers! There are things that we can do well – for instance natural language – that remain essentially impossible for computers,” Matthew enthuses. “I hope to find new ways to think about what cognition is – new non-computational ways – to think about what we are and how we work."